ABSTRACT HISTORICAL EXPLANATION AND VALUE - NEUTRALITY by James J. Leach recent emergence of philOSOphy of history as a criti- a, attention focuses on the notion of objective his- iation. Three problems are central: the meaning of {planation,' the relationship between scientific and glanation, and the kind of objectivity appropriate to ically, the issues turn on the covering law theory of ‘we defend this theory, prOpounded by K. POpper and C. at the diverse challenges of Weber, Collingwood, Lavine, Donagan and Scriven. Briefly characterized, the theory itific and historical explanations on the model of log- Lon: deductive or inductive. Rationally acceptable nust incorporate, as essential premises, empirically true (or highly confirmed) general laws. t the issues, we critically explore three interrelated ad by Max Weber and pivotal to the controversy: that luiry must be objectively value-free (Value-Neutrality objective cultural explanation, nevertheless, requires Lerpretative understanding beyond mere subsumption 1derstanding thesis); and finally that such understand- ;es a peculiarly cultural method of concept formation 1esis). Each faction to the covering law controversy 'I 0.“ James J. Leach accepts at least one of these theses. In rejecting all three, we attempt to resolve the issue. The covering law theory, though de- fensible against the latter two propositions, is shown to embody unnecessarily and unjustifiably the thesis of value—neutrality. Here it can be successfully revised. Subjective judgment, as espoused by 'empathy' theorists, thus proves important but mislocated. Its significance lies not in the explanatory force of arguments but in the rational acceptability of the relevant hypotheses. Taking our cue from the substantial insight of Verstehen theorists (Lavina, Dray and Scriven) on the historian's need to make value judgments, we argue, against Hempel, for the essential role of such judgments in any philosophic analy- sis of rationally acceptable explanation. we take this insight to be an additional pragmatic condition to the covering law theory, rather than a fatal weakness. Hence the denial of value-neutrality does not support the thesis of subjective understanding. The case against value-neutrality, accordingly, seems best argued on the basis of recent work in statistical analysis of rational decision making in the face of uncertainty. In particu- lar, from the fact that the scientist accepts or rejects corrigible hypotheses, and thus decides when the evidence warrants his accep- tance, it follows that he cannot escape making value judgments. This argument we unpack and defend against covering law theorists. But in such a way as to avoid both a behavioralist reduction of be- lief to action andeipragmatic reduction of truth to utility. Suf- ficient evidence is shown to be a function of such pragmatic factors l y A James J. Leach as the cost associated with the importance of making a mistake when acting on beliefs. The goal of science and history thus appears not merely as truth for its own sake, but as truth modi- fied by other criteria, :22. epistemic and pragmatic utilities. The import of this argument forces a reconsideration of the meaning of scientific and historical objectivity and of the re- lationships between theoretical, technological and policy making aspects of rational inquiry. The humanistic orientation of his- tory, stressed by varied 'empathy' theorists, can be preserved. Yet not at the price of abandoning history as a branch of the science of society. This, then, constitutes the main thrust of our thesis that the covering law theory survives the varied logi- cal criticisms of 'empathy' theorists, but only on condition that pragmatic and purposive elements be included essentially in a logi- cal reconstruction of explanations, and hence that the value-neutra- lity thesis be surrendered. HISTORICAL EXPLANATION AND VALUE - NEUTRALITY By ~,~'\. y. James J? Leach A THESIS Submitted to Michigan State University partial fulfillment of the requirements for the degree of - DOCTOR OF PHILOSOPHY Department of Philosophy 1965 AC KN OV'JIEOGFIEI‘ITS : my pleasure to note the following acknowledgments: :or.Alan Gewirth for so thoroughly and ably introducing history of philOSOphy; to Professor Richard Rudner 1couragement and willingness to engage me in instructive .c controversy, for thereby introducing me to the : study of contemporary issues, and for serving as a my committment to the philosophic enterprise; to John Lachs, friend and colleague, for valuable 1 and encouragement to publish parts of these last >ters; to Professors Harold walsh and Gerald.Massey, advisors, for critical comments and encouragement; e and children for more than I can say. CONTENT ‘DGI‘O:ENTSOOOOOO0.0000000...000. NTRODUCTION . . . . . . . . . . . . . . . . . . . . The Recent Controversy Three Theses of Weber Plan of the Present Work HE COVERING LAN THEORY OF EXPLANATION . . . . . . 0 K. Ponper's Early Formulation C. Hempel's Deductive Rodel Pseudo, Genuine and Acceptable Explanations C. Hempel's Probabilistic Model Complete Explanation and Approximations The Covering Law Account as a Theory of Explanation ECONSTRUCTIONS OF THE SU THESIS . . . . . . . . . . Idealism and the SH Thesis The Standard Covering Law Answer Recent Replies A. Schutz' Reconstruction of SU ILLIAM DRAY'S RECONSTRUCTION OF THE SU THESIS . . 0 Some Relations Between the SU and VN Theses The Rational Model of EXplanation Pragmatic Dimensions of Explanation Critique of the Rational Model Further Objections to Dray‘s Rational Model HEMPEL'S VERSION OE RATIONAL EXPLANATION. . . . . Rational Explanation and the Covering Law Model Rylean Dispositional Explanations Hempelian Analysis of Dispositions Critique of Hempel's Rational Rodel: Rationality and Tautologies Critique of Hempel's Rational Model: Rationality and Evaluations The Probabilistic Model and VN 23 96 135 202 RE PRAGMATIC DIMENSION OF EXPLANATION AND THE VN THESIS 251 Inductive Ambiguity or Inconsistency and Total Evidence Rational Credibility and Utilities Explanation and Value-Neutrality Value-Neutrality and Historical Explanation AHF; I” O O O O I C O O O O O O O O O O O O O O O O O . iv 311 CHAPTER I INTRODUCTION The Recent Controversy In recent years philosophy of history has again become a live and controversial subject annng empirically oriented philosophers in Eng- land and fliis country. But it has become so only as a critical rather than a Speculative discipline.l After long neglect, in fact generally since Descartes' expulsion of history from the domain of knowledge proper in Part I of the Discourse, during which time the subject be- came a veritable obsession with speculative continental philosophers, attention has again been directed to questions of explanation, pre- diction, interpretation and objectivity in historical inquiries. No doubt the two major exceptions among empirical philosophers to this by-pass of history are Karl Popper (in his series of essays dating from 1936 and now published as The Poverty 23 Historicism and his later 9222 Society 22g ltg Enemies) and Carl Hempel (in his now classic essay I'The Function of General Laws in History"). However, these works were not at all sufficient to arouse recent philosophers from their complacent slumbers. This seems due in part to their lucid and potent argument producing general acceptance among empiricists, and in part to a lack of clarity and persuasiveness in the counter-argument of the idealists causing their position to go by default. Only when recent empiricists of 1 W. H. welsh, Philosophy 3; History (N.Y.: Harper and Brothers, 1951). Chapter I. the analytical variety, attempting to revive some of the idealist doctrines by reconstructing them in linguistic and pragmatic guise, took issue with and more lucidly challenged the position of these essays did the subject emerge once again as the stage of a major philosophic controversy. This general lack of empirically-oriented philosophic interest in history, strange though it be in itself, becomes even more per- plexing in light of the frequent charges by historians that philo— SOphers have imposed their own methodological problems and rigorous scientific ideals of inquiry on the historian's practice, and hence have forced this practice onto some Procrustian bed. It seems, in- stead, that historians themselves, not philosophers, have fostered these problems. And they have done so largely because of the ex- tensive variability in their interpretations of historical events, actions and processes, a variability often embarassingly difficult to reconcile with claims of historical objectivity. In the phrase of one theoretically-inclined historian, Cushing Strout, "A spec- ter haunts American historians—-the concept of causality. After nearly a hundred years of passionate and dispassionate inquiry into 'the causes of the Civil War' the debate is still inconclu- sive."2 Moreover, the specter arose in the context of develop- ments which, by erecting a barrier between the historian's prac- tice and his theory, prevented him from even attempting a sys- tematic appraisal of his explanatory concepts and hypotheses. So 2 C. Strout, "Causation and the American Civil War," in History £13 Theogy, Vol I (1961), p. 175. ce has the situation become that many historians take scrupulously avoiding generalizations and theoretical tions. Yet in practice, of course, these same historians to transcend mere chronicling of facts by offering ex- 3 and interpretations of what they take to be significant y suggesting causes, motives and reasons for historical d actions. developments largely responsible for this gap between ice and theory of historians has been instructively at- by David Potter3 among others, to the dilemma produced rlier cult of scientific history and the relativist re- it. During the "scientific" era of history, under the of Ranke's anti-prOpaganda campaign, historians strove hard, neutral descriptions of fact and disclaimed any nterpret or explain or find meaning in the facts. Though med to know the truth of these events, it was a truth e only by the purity of descriptive inquiry. Any attempt n or interpret would be to underwrite purity with bloody judice and personal value committments. when it became evident that this strategy failed to lessen hility of historian's conclusions, that the data apparent- in varied tongues to even more varied investigators, his- inally discarded this creed and partly regained their in- l sanity by recognizing the necessity of going beyond the L7 Potter, "Explicit Data and Implicit Assumptions in Histori- ," in L. Gottschalk (ed.), Generalization in 3,113 Writing 9;; Chicago: University of Chicago Press, 19637, pp. 17 - 3. iterpret and explain it, to organize and unify the facts ;0 understand them. And when this strategy produced only Tusion and variability, they embraced, as a last resort, . relativism. Since one must interpret and explain, and 5 produces unwanted variability, they sacrificed in theory Livity of their inquiries. The assumption common to both ;hat objectivity is an invariant and value-neutral matter, ought to yield the same conclusions for all competent in- fostered a neglect of the theoretical aspects of historio- T the nature of explanation, interpretation and objectivity. a view, they could not attain objective truth without con- ;heory, and on the other they could not attain objective 1 if they did invoke theory.h In neither case was theory -le investment. Better to be "pragmatic", to rely upon :5" or is "successful" in practice, became the strategy >rking historians. However, the problems connected with )HS as explanation, interpretation, cause, motive and 2p interfering with such practical serenity until even- : problem of clarifying these notions is imposed upon the :r of history. .n the phiIOSOpher's hands the problem takes a different .3 not always recognized by the historian as his problem. :ears to him that the phiIOSOphers are forcing his practice 'rocrustean bed, that they are legislating criteria for 2 historical practice, criteria of proper explanations, 'ido , p0 1790 of causal imputations and of the objectivity of conclusions. At any rate, this situation serves to introduce some of the problems underlying the major controversies for contemporary philosophers of history: what is the nature of historical explanation? Does it differ fundamentally from scientific explanation? What relation is there between a phiIOSOphic model or theory of explanation and actual historical practice? Can historical explanations be ob- jective? Does objectivity presuppose a value-neutral appraisal of the acceptability of conclusions? What, if any, is the role of purpose and efficiency in the process of explanatory inquiry? Three Theses of Weber In what follows, I shall consider certain aspects of this con- troversy by a careful investigation of three interrelated theses advocated at one time or another by Max Weber, perhaps one of the most sophisticated and influential social scientists of our cen- tury. I shall argue that all three theses are unacceptable, but that they are so in extremely suggestive and important ways. In particular, they form the basis of most recent controversies about Hempel's "covering law" theory of historical explanation, at least in the sense that all sides to the dispute have accepted one or more of these theses while most reject at least one other. More- over, though these beliefs are unacceptable, I find Weber's method- ological writings in general to touch in an unsystematic but in- structive way on most of the conclusions to be advocated herein. In fact, there is little doubt that the major positions held in this controversy are all influenced by Weber's writings. This includes the naturalistic views of Nagel, Popper and Hempel, the non-naturalism of Lavine, Natanson and Schutz, and the analytic position of Gardiner, Dray, Donagan and Scriven. Before turning to Weber's theses, let me characterize briefly what has become known as the "covering law" (CL) theory of ex~ planation, since it is at the center of the recent controversy, and since Weber's theses can best be viewed in contrast to this "COVering law" theory. But since chapter two is devoted to a detailed examination of this theory, the present account will be to the following sketch: ‘Hhat is scientifically explained, in addition to laws, er a type of event or action nor a unique concrete phenom- t an aspect, property or description of an event or action. To explain scientificaLly and completely some event or 3, ideally, to provide an argument, deductive or inductive, 1g a description of the phenomenon to be explained as the an, and a statement of the appropriate general laws and 1t conditions as premises. To be adequate or rationally acceptable, a scientific ex- 1 must contain as essential premises or eXplanans, general :h are both testable and either true or highly confirmed alevant available evidence. Scientific eXplanation and prediction, two of the central T scientific inquiry, are structurally identical. They xly pragmatically, so that all adequate explanations have -al predictive force. , the three fundamental theses of Weber's methodology, which > use as a framework for considering the above controversy, >llows: _.) Thesis of Value-Neutrality (VN): The objectivity of 'ical explanations requires as a necessary precondition that Liner, qua scientist or historian, make no value judgments. .. Hempel and P. Oppenheim, "The Logic of Scientific Ex- 1," in Feigl and Brodbeck (eds.), Readings in the Philoso— :ience (N.Y.: Appleton-CenturyaCrofts, 19537, pp. 319-330. -ar he must remain evaluatively neutral when appraising ibility or correctness of his explanations. Hence objectiv- Lriant in the sense of yielding the same conclusion for ant inquirers. This thesis follows from the sharp logi- Lion of questions of fact and questions of value, and lbsequent distinction between descriptive statements and )roposals. In Weber's terms, "a systematically correct proof in the social sciences, if it is to achieve its lst be acknowledged as correct even by a Chinese";6 it "unconditionally valid type of knowledge“.7 This means at be acceptable to all, independently of variable per- 'ests, attitudes or values. "For scientific truth is rhat is valid for all who seek the truth."8 ) Thesis of Subjective Understanding (SU): Although Lble empirical explanations must be objective in this manner, adequate historical explanations of purposive ins cannot be attained by mere subsumption of the action 'ing general laws. Instead, the explanatory force of the connecting link between the action and its causes, reasons, is provided by some sort of subjective or in- re understanding. Our aim in the cultural sciences is :tanding of the characteristic uniqueness of the reality f Weber, The Methodology of the Social Sciences (Glencoe, Free Press, l9h9), p. 58: ;S°’ p. 63. ,d., p. 8b. we move...and the cultural significance of individual 9 we are, consequently, "concerned with psychological and tual phenomena the empathic understanding of which is I a problem of a specifically different type from those e schemes of the exact natural sciences in general can or 10 In other words, since the subject matter of solve." as a cultural science, consists of purposive human actions, only be adequately explained or understood from the sub- Joint of view of the agent. Hence, the mere subsumption actions under covering laws, from the external point of the observer, is insufficient to explain completely the 1's subject matter. SU, accordingly, entails the denial ; theory. Lii.) Thesis 0f.£§EEiTIXE§§ (IT): To eliminate "the stic prejudice that the goal of the social sciences must aduction of reality to tlawg',"ll and to achieve the re understanding required by historical explanations, the 1 must utilize "a kind of concept-construction which is and, to a certain extent, indispensable to the cultural ."12 Weber refers to the product of this peculiarly cul- 3hod of concept formation as an ideal-type or utOpia. generic limiting concepts, "purely analytical constructs Ibid., p. 72. ibid., p. 7b. ibid., p. 101. Ibid., p. 89. .- 159-3. 10 created by ourselves,"13 used as a standard to analyse historically unique configurations and to compare or measure the culturally significant components of human action. The statements containing these constructs are to be clearly distinguished from judgments of value ideals as well as from empirical or descriptive hypotheses or laws. 'Weber, however, is extremely vague about the nature of ideal-types, and hence about the exact way in which they consti- tute an alternative to the CL theory of explanation as subsumption under laws. But for our purposes we need only consider this thesis as advocating some alternative to the CL theory, an alternative using ideal-types and satisfying the requirements of the SU and VN theses. In the succeeding chapters, it will be shown that the non- naturalists (Natanson and Schutz) accept some version of all three theses, the naturalists (Nagel, POpper and Hempel) deny SU and IT but defend VN, while the analysts (Dray, Scriven and Donagan) deny VN and IT but defend SU in a revised version. Accordingly, in denying all three theses, I shall attempt to mediate this contro- versy, to synthesize the important denials of both the naturalists and the analysts. But while my synthesis will include a defense of the CL theory of explanation, I will also extend the theory in such a way as to incorporate within it the significant contribu- tions of many opponents of this theory, and hence in a way that none of these three groups would find especially felicitous. Such is my embarrassment. 13 Ibid., p. 96. 11 The important controversy recently aroused about historical explanations centers, as we have noted, on the CL model of explana- tion defended by Hempel, Popper, Nagel, Braithwaite and others of a naturalist persuasion, and challenged by such varied non-natura- lists as Schutz, Dray, Donagan and Scriven. With the defenders emphasizing the formal and logical aspects of eXplanation and the challengers stressing the epistemological and pragmatic aspects, the controversy has not always proved fruitful, since often they seem to be arguing past each other. The main issue at hand is whether or not historical explanations are best construed accord- ing to the CL model of scientific explanations. One of the most frequent objections of the challengers, ground- ed on the VN and SU theses, has been that a different kind of model is required for historical explanations because of the extreme ob- jectivity prevailing in the CL scientific model. Some suggest a wider use of the term 'explanation,‘ a more generic version common to all types of explanation; others advocate a peculiarly different and characteristically historical kind of explanatory model. But in both cases the epistemological and pragmatic dimension1h of ex- planation is emphasized so that, unlike Objective scientific cases, the role of the person accepting explanations cannot be ignored. Hence, both groups distinguish sharply between explanation 322 po- tential scientifically testable prediction and explanation 323 in- telligibility or understanding; and between the objectivity of 1h . . J. Yolton, Thinkin ‘gpglPerceiVing (LaSalle, 111.: Open Court, 1962), pp. 117-36. 12 testable and confirmable scientific explanations achieved inde- pendently of variable personal judgments and the subjectivity of historical explanations grounded in these valuational beliefs and personal judgments of acceptability. Some, however, have argued for "a scaling of explanations in 15 terms of their objectivity" on the basis of the latter aspect, of the relative independence from any one individual, instead of for a clear-cut dichotomy between an objective science and a sub- jective history. No doubt, this view requires emphasizing not the logical structure of explanation but the epistemic criteria of rationally acceptable explanations. Yet Yolton, representing most recent Opponents of the CL model, takes this question to concern the "explanatory force" of theories or hypotheses, and hence erron- eously, I think, views this force not as a logical matter but as one concerning conceptual schemes, general attitudes and empathetic understanding.16 But in this case it is not clear whether he is opting for intelligibility or understanding as providing a different kind of explanatory force in historical explanations, or as a pre- condition of all explanations. Yolton's objection rests on the belief that testability is appropriate for scientific explanations but not for historical ones. For in history, general attitudes and conceptual schemes, i.e., an element of Verstehen as subjective understanding not explanation, is more relevant even though there are no clear-cut tests or 15 Ibid., p. 122. 16 Ibid., pp. 12h—128. 13 criteria for determining the correctness of the understanding. The question then is whether or not a system of statements must meet the criteria of testability and deducibility in order to be explanatory. Yolton holds that it must meet these requirements only to be a scientific but not an historical explanation. The latter requires only the criteria of intelligibility or under- standing which "varies in kind with the difference of needs, of objective," of context, purpose, values and interests. In fact these are the preconditions, he maintains, of all explanation, while deducibility and testability are the ideal controls only of the sciences. But again the former can be arranged in a scaled order of degrees of objectivity and independence of particular needs, values and interest, and degrees of empirical testability. Now it seems to me that the challengers of the CL model are correct in advocating a widening of the notion of explanation so as to include these epistemological and pragmatic aspects. They have, nevertheless, failed to upset the CL theorists' claim that a logical relation, deductive or inductive, between laws or general- izations and the events to be explained provides the explanatory force in both scientific and historical explanations. Instead, the inclusion of pragmatic or purposive elements weighs heavily against the VN thesis, but not in favor of SU as they think. The notion of Verstehen as embodying criteria of intelligibility, to be effective against the CL theory, must be directed not to the logical explanatory force of an argument, but to the objectivity and acceptability of the explanans. The important questions such 1h challengers raise, however unclearly, are first whether or not historical explanations, not easily amenable to empirical confirma- tion, can be appraised as acceptable on inductive criteria or in- dices other than testability or confirmation. And secondly if so, whether such criteria support the denial of the VN thesis. I think an affirmative answer is due each question. But the challengers of the CL model either have not seen this or have unsuccessfully defended it, and either for two basic reasons. First, they have assumed that if there were other criteria than testability, they must be incompatible with, and hence replace, the criterion of deducibility which requires the event to be ex- plained to be deduced from the explanatory premises. This led them to lodge their attack against the logical aspects of the CL model, and hence to support SU as a thesis about the explanatory force of arguments, instead of against VN as a thesis about the acceptability of explanatory hypotheses. Some, in fact, have confused the two theses by conflating them, by suggesting that normative generaliza- tions provide the explanatory force or connection between antecedent conditions and eXplanandum. Secondly, they have emphasized, along with CL theorists, the deductive model almost to the total neglect of the probabilistic model. As a result, the usual arguments of those who do oppose VN have been misdirected and inadequate, and the arguments of those who defend the VN thesis have been swayed by the unfortunate notion of objectivity associated with the deduc- tive model. Consequently, I want to argue, against Weber's three theses, 15 that explanations in all the various empirical disciplines have the same eXplanatory force, which is adequately explicated by the CL theory as an ideal type or idealization. But, additionally, the objectivity of explanations in all disciplines, insofar as they take the form of inductive arguments, seems to require additional in- ductive criteria of acceptability beyond confirmation and testability. Moreover, one such criterion entials the denial of VN by requiring the making of value judgments about the costs associated with the possible mistakes of accepting or rejecting explanatory generaliza- tions. Finally, these considerations suggest that the various dis- ciplines can perhaps be distinguished according to how much weight must be placed upon this latter criterion, even though it is nec- essary in some degree in all disciplines, since some can more easily establish policies that are invariant regarding various goals and hence are relatively more value-free than others. 16 Plan pf the Present work My general plan then will be, first of all, to examine in some detail the CL theory of explanation as formulated by Hempel and Popper. This, the task of chapter two, will include a discussion of both the deductive and probabilistic nomological models of ex- planation, the conditions of adequacy for sound explanations, the status of the models as complete idealizations, and various senses of incompleteness or approximations to these two ideal models. Then, since the SU thesis entails the denial of the legitimate ex- tension of the CL models to historical explanations, I will examine various formulations or reconstructions of this thesis. In chapter three both the idealist formulation of SU and the standard reply of such CL theorists as Abel, Hempel and Nagel to this intuitive version of Verstehen will be considered. This early exchange leads to more recent reconstructions of the SU thesis, to alternative defenses of peculiarly ideographic historical explanations in contrast to the nomothetic explanations found in the sciences. In particular, the non-naturalist position of Natanson and Schutz will be considered as a rebuttal to the standard CL or naturalist answer to all empathy theorists. Schutz' reconstruction of Verstehen and the SU thesis raises the question of the status of the CL theory of explanation, of how it relates to explanations actually offered by historians. Hempel's answer to this question is that suCh philosophical theories are explica- tions or, extending Weber's notion of ideal-types, idealized models to be appraised in part on grounds of their usefulness in attaining 17 certain purposes or goals. This answer will receive special em- phasis since it fits one of my main contentions: that the major attempts to rehabilitate the SU thesis by some form of empathy turn on the inclusion of such related pragmatic, personal and purposive factors as decisions, interests and attitudes. However, we will argue that the necessary inclusion of these pragmatic elements fails to support the SU thesis. Instead, as noted above, they support a denial of the VN thesis, a point not sufficiently appreciated by the non-naturalists. But analytically-oriented philosophers (especially Dray, Scriven and Gardiner), on the other hand, do clearly recognize the bearing of these pragmatic factors on the VN thesis. ‘we will, consequently, turn in chapter four to a detailed critical examina- tion ofIDray‘s I'rational model. of explantions, viewed as an analytic reconstruction of the SU thesis. For Dray uses his model for the two-fold purpose of defending the SU thesis and of denying the VN thesis. He does so by substituting normative principles of action for Hempel's descriptive empirical covering laws as the source of eXplanatory force. As already indicated, however, we shall contend that Dray's criticism of the CL theory fares no better than the earlier critiques of the idealists and the non-naturalists. They all fall Short of their mark. To this extent will we attempt to defend the CL theory of explanation: to the extent of support- ing it against the SU and IT theses. In the last two chapters, we will examine the extension of the CL theory to cover historical explanations of purposive human actions. 18 Chapter five is devoted to Hempel's broadly dispositional analysis of human actions, while chapter six raises some problems for the probabilistic model of the theory. In chapter five we will consider Hempel's alternative CL theory of reason-explanations by contrasting it with Gilbert Ryle's version of diSpositional predicates, since the latter serves as the basis for more moderate criticism of the CL theory. The discussion at this juncture centers on the criticisms of Donagan, R. B. Brandt and Scriven to the requirement of includ- ing general laws as a necessary condition for adequate historical explanations. Here again Hempel's CL theory will be defended against attempts to disunify the empirical sciences, to contrast sharply the ideographic and nomothetic sciences. Finally, in chapter six, we shall examine some aspects of Hempel's probabilistic model of explanation, since the general laws required to explain historical actions will usually be statis- tical in nature, and since the inclusion of such laws bears heavily on the VN thesis. In particular, we shall argue that while Dray's argument against the VN thesis is unconvincing, his conclusion, the denial of VN, can be adequately supported on other grounds. These grounds relate closely to Hempel's probabilistic model and to the criteria of acceptability for statistical hypotheses. To this extent will we defend some of the varied opponents of the CL theory. The argument will try to show that the insistence of empathy theorists on purportedly non-experimental factors, which force the historian to consider pragmatic and evaluative aspects of inquiry, is cogent. And in a way which meets the standard or official CL answer by placing 19 the issue clearly in the context of the logic of justification. It is noteworthy in this context to notice the constancy of almost all CL theorists in their advocacy of value-neutrality and in their unwillingness to provide the pragmatic dimension of ex- planation with any important systematic function. While not deny- ing the existence of this dimension nor even its significance, they relegate it to a pre-systematic, non-theoretical or psycho- logical status concerning the discovery rather than the justifi- cation or confirmation of explanatory generalizations. They then regard the objectivity of scientific and historical eXplanatory accounts as independent of pragmatic or purposive considerations, and hence as supporting the VN thesis. Objective justification of explanatory generalizations involves, for them, only the require- ments of deducibility, testability and evidential or confirmatory strength. But this position also requires depicting the scientist as essentially a guidance-counselor of decision makers, not himself as a decisionmaker. It requires distinguishing sharply between the theoretical goal of achieving truth and nothing but the truth on the one hand, and the practical goal of deciding to accept or re- ject hypotheses or theories on the other. Accordingly, our defense of the Opponents of the CL theory turns on a criticism of this latter distinction, the notion of objectivity it supports, and on the tenability of widening the notion of explanation to include the pragmatic dimension. Lest this twofold defense appear paradoxical, however, it must be noted that the apparent paradox results from an assumption 20 shared in common by both sides of the controversy, as well as by most historians: that the CL theory of explanation (and hence a denial of SU) entails VN. If this entailment did hold, my twofold defense of the CL theory and of the denial of VN would indeed be inconsistent. Such, it will be argued, is not the case. But recognition of the fact that the entailment does not hold, that the denial of VN is compatible with and perhaps even required by the CL theory, has been obscured by the undue emphasis placed on the deductive model of explanation and the use of universal or deterministic laws as necessary ingredients in the explanans. Both sides are, I fear, partly responsible for the neglect of the proba- bilistic model and the subsequent lack of investigation of statis- tical generalizations, so important for social and historical ex- planations. Only when the latter model receives proper attention can it be seen how VN can be successfully denied and, at the same time, why this denial does not entail the affirmation of SU or the denial of the CL theory. The usual attempts to deny VN fail because, as Weber clearly saw, they locate the value element in the context of discovering or imaginatively constructing plausible explanatory hypotheses. Most CL theorists readily concede this point without damaging their VN thesis. For the latter thesis concerns not the discovery but the justification, corroboration or confirmation of explanatory hypotheses. Further, most defenders of SU, from Weber and the early ideal- ists to Schutz and Dray, fail to locate the value element in the context of justification for much the same reason that they object * ‘r,v‘--V~"-v 21 to the CL theory in the first place: because they believe this com- mits one to a crude form of behaviorism or pragmatism or both. Our task, accordingly, will be to show that this belief also lacks founda- tion. In other words, we will defend the CL theory of explanation by extending it to include a denial of VN, but in such a way as to avoid any committment to a completely behavioral account of the acceptance of beliefs or hypotheses, or to a completely pragmatic version of evidence or the rational acceptability of empirical hypotheses. All that is nec- essary to Oppose VN successfully, I think, is to show that the acceptance and acceptability of explanatory hypotheses entails some behavioral aspect and some pragmatic criterion of appraisal. It is not necessary to show that they are equivalent. That is, a denial of VN neither re- quires beliefs to be reducible to actions nor truth and confirmation to be replaceable by utility. The cogency of either of these latter theses we leave an Open question, though the latter surely seems less so than the former. Much of the point at issue amounts to the charge made by experi— l7 mentalists that philosophers as varied as Weber, Dilthey, Schutz, Dray and Hempcl fail to supply a broad enough model of scientific or historical inquiry. In particular, they tend to evade the issue which 17 E. A. Singer, Experience _a_n_c_i Reflectiog (Philadelphia: University of Pennsylvania Press, 1959); C.W. Churchman, Theory of Experimental Inference (N.Y.: Macmillan, l9h8), Prediction and QE- timalIDecisIOn’(Englewood Cliffs, N.J.: Prentice:Hall, 19617, "Sta- tistics, Pragmatics, Induction," Philosophy of Science, Vol. XV (July 191.8); P. Frank, Philosophy of ScienceTEnglewood Cliffs, N.J.: Prentice-Hall, 1957); R. Braithwai‘E'é’,"'s'Ei'e'nt'iric Emlanation (N.Y.: Harper Bros., 1953); R. Rudner, "The Scientist Qua Scientist Makes Value Judgments,“ Philosophy 2f Science, Vol. XX (1953). 22 experimentalists take as central: the theory of experimental ac- tion. For when such a theory is considered at all, its efficacy seems at best construed at the level of Weber's "subjective selec- tivity,"18 instead of at the level of experimentally controllable notions such as teleology, production, function and purpose. These latter notions turn on Weber's neo-Kantianism, particularly in his use of ideal types or limiting processes to relate observational data to theoretical and evaluative ordering structures. Further, these varied philosophers fail to supply a sufficient- ly broad model of rational inquiry largely because they tend to isolate questions of fact from questions of evaluation, questions of confirmation or evidential strength of beliefs from questions of purposes or application. Hence, by emphasizing the formal to the neglect of the purposive aspects of explanation, they fail to even consider the necessary conditions for a complete theory of ex- planatory inference of methodology, a theory for selecting the most reasonable explanations. Such a theory, it would seem, requires not only the semantical criterion of confirmation or agreement with facts and the syntactical criteria of consistency and simpli- city or economy, but also the pragmatic criterion of utility and efficient purposive behavior.l9 18 . ‘Weber, pp. Cit., p. 82. 19 Frank, pp. cit., Chapter 15. Mn ‘3'" CHAPTER II THE COVERING LAW THEORY OF EXPLANATION ,E‘ Popper's Early Formulation Before considering some of the issues surrounding the controversy about the applicability of the covering law theory of explanation to historical inquiry, let me pull together some of the various strands of this theory. It will be helpful, I think, to have a fairly full statement of the CL theory before us in order to see whether or not it can be fruitfully extended to cover historical, as well as scienti- fic, eXplanations, i.g. whether or not it can be defended against the various interpretations of the SU thesis. Most of the formulations of the CL theory occur in the context of natural science explanations, particularly of the causal 'variety. Consequently, much of the dis- cussion will be limited to those aspects of’the theory which bear most directly on the case of historical explanations of purposive human actions. This means that certain important aspects of the theory will receive more detailed treatment than others. In particu- lar, we must forego any but the briefest account of the relationship between explanation and prediction, i.g. of Hempel's structural sym- metry thesis. Additionally, we can but mention the difficult onto- logical problems about what it is that can be explained by the CL theory, except insofar as the question relates to the central notion of a complete explanation which will be discussed in some detail. The plan of this chapter, accordingly, is to outline briefly Popper's early formulations of the CL theory and then to look more closely at Hempel's recent systematic treatment. The theory will 23 214 be shown to include two formal models of scientific explanation, and various criteria or requirements for determining what is to count as both "an eXplanation" and "an acceptable or sound explanation." Different senses of ideal completeness along with various degrees of approximation or incompleteness will also be considered. Finally, in this connection we will raise the question as to the nature of this particular enterprise, i.e. of offeringaitheory or explication of the notion of explanation. Let me begin with Karl Popper's formulation of the CL theory, since he claims to be its author, having put it forth as a general theory of explanation as early as 1935 in Logik der Ferschupg, more recently translated as The Logic 2f Scientific Discovery; and again in the two works cited above with special reference to history. The central thesis of the theory, however, has historical roots in the comparable views of Weber, Campbell, Mill, Galileo and even Aristotle. In brief, Popper follows Weber in characterizing the explanation of natural phenomena as the subsumption of the many under the unity of the one, in the sense of subsuming what is to be explained under general laws. Hence explanation, unlike descrip- tion, takes the form of an inference or argument containing general laws as essential premises. To offer an explanation of some phenom- enon is to offer an argument, not merely descriptive information. For it is the logical or inferential connection between the general laws and the phenomenon to be explained which provides the explana- tory relevance of the former to the latter and assures the eXplana- tory force of the covering laws. Accordingly, POpper writes: "To 25 give a causal explanation of a certain event means to derive de- ductively a statement (it will be called a prognosis) which de- scribes that event, using as premises of the deduction some Epi— XEEEEE lgzg together with certain singular or specific sentences which we may call initial conditions."1 He then illustrates this pattern by reconstructing a causal explanation of the breaking of a given piece of thread found capable of carrying one pound only, but with a two pound weight put on it. The appropriate explanation will contain both kinds of constituent statements just mentioned, gig. two laws and two initial conditions. The two universal laws are: (L1) "For every thread of a given structure S (determined by its material, thickness, etc.) there is a charac- teristic weight W, such that the thread will break if any weight exceeding W is suspended from it," and (L2) "For every thread of the structure S, the character- istic weight W, equals 1 lb." The two initial c0rditions then are: (Cl) "This is a thread of structure 8.", and (C2) "The weight to be put on this thread is equal to two pounds."2 From these four statements, both kinds of which are necessary in- gredients of a complete causal explanation, we can thus deduce the K. Popper, The Logic 2f Scientific Discovery (N.Y.: Basic BOOkS’ 1959), p. 59. Ibid., p. 60, new footnote *1. 26 prognosis, conclusion or description of the event to be explained: (E) "this thread will break." The situations described by the initial conditions (Cl) and (C2) are then spoken of as the cause of the event in question, and the event described in the prognosis (E) as the effect. But while differing from description in this inferential way, explanation is also similar to description in at least one impor— tant sense for both Popper and Weber. All scientific explanations and descriptions of facts are highly selective; they are always theory - dependent and never occur in isolation. The reason for the impossibility of avoiding selectivity is, of course, the "in- finite wealth and variety of the possible aspects of the facts of our world,"3 and the finite limitations of descriptions. Thus our descriptions and explanations will always remain incomplete, a mere selection according to our interests of the facts available for description. The point is a result of what Popper calls his "searchlight theory of science," since description depends on our point of view, theories and interests; much as what a searchlight makes visible depends upon its position, our way of directing it and its intensity. There can be then no such thing as an actually complete description, no less a complete explanation, of any in- dividual event or fact in the world. Both require abstracting from and selectivity of the infinite subject matter. K. Popper, The Open Society and itg Enemies (N.Y.: Harper and Row, 19527: vol. II, p. 2ST: Q‘t. 27 In his later work Popper also derives three important con- sequences from this deductive model of causal explanation. Events, first of all, are causes or effects only relative to some universal laws covering them, not absolutely. There is little doubt of Popper's allegiance to a Humean view of causality which involves the denial of necessary connections between events, and instead emphasizes the connection in terms of empirical regularities. Yet his theory "differs from Hume (l) in that it explicitly formulates the universal hypothesis that events of kind A are always and every- where followed by events of kind B; (2) that it asserts the truth of the statement that A is the cause of B, provided that the uni- versal hypothesis is true."h In other words, in addition to Hume's events A and B, POpper establishes a third element, a universal law, with respect to which we can speak of a causal link, or even a "necessary connection." However, Popper readily admits, in a passage influencing some recent critics, that "these universal laws are very often so trivial (as in our own example) that as a rule we take them for granted, instead of making use of them."5 Secondly, he formulates loosely what has recently been labeled the "structural symmetry or identity thesis" concerning explanation, prediction, and confirmation or testing. "There is no great dif— ference between explanation, prediction and testing. The difference is not one of logical structure, but one of emphasis; it depends on our interests what we consider to be our problem and what we Ibid., p. 363. Ibid., p. 262. 28 do not so consider."6 Further, this pragmatic emphasis serves to distinguish three kinds of sciences, parallel to the three kinds of scientific interests, purposes or problems we may have. The "theoretical or generalizing sciences" (e.g. physics, biology, sociology) use the pattern to test and establish universal laws or hypotheses considered as problematic. The "applied generalizing sciences" (e.g. engineering) take the premises as given and use them as means for predicting the prognosis and hence deriving some new information. And the "historical sciences," by contrast, take the prognosis as the given explanandum and attempt to uncover the premises, initial conditions and laws, from which to deduce and hence explain the given particular event, instead of testing or predicting. Accordingly, Popper accounts for the oft-repeated view that historians are interested in explaining particular events, not in formulating or establishing universal laws. The laws are formulated by the generalizing sciences (E'E' sociology) and 'assumed' by the historian. However, he is careful to block the conclusion which many have inferred from this point, gig. that his- torical explanations need not utilize general laws. Finally, POpper's deductive model and the derived division of the sciences serves to eludicate his view concerning the role of theories, interpretations or points of view in history. Unlike the generalizing sciences, in history we have no "unifying theories; K. Popper, The Poverty 2; Historicism (London: Routledge and Kegan Paul, 19577: p. 133. 29 or, rather, the host of trivial universal laWS we use are taken for granted; they are practically without interest, and totally unable to bring order into the subject matter.“7 Some of these laws are indeed trivial, as Popper's case of explaining the defeat of Poland's first division in 1776 by appealing to the following law clearly indicates: "If of two armies which are about equally well armed and led, one has a tremendous superiority in men, then the other never wins." Yet Popper also endorses the historian's practice of appealing to selective principles, points of view or interpreta— tions which are merely "quasi theories," often preconceived notions as the great-man thesis or the causal priority of economic con- ditions, geographic conditions or moral ideas. Though such "his- torical theories" contrast sharply with scientific theories in so far as they are untestable (unfalsifiable) by facts independent of the preconceived theory itself, and hence as non-scientific though still cognitively and empirically significant, they are nevertheless given an important role and status as "inevitable" in historical inquiry. They serve as foci, "centers of interest" or working hypotheses for collecting additional facts and records, as well as being of topical interest by elucidating the problems of the day. We will want later to inquire whether such "theories" con- stitute constituents of proper or merely pseudo explanations when combined with appropriate antecedent conditions. We might note 7 Popper, Th: Open Society E29 TEE Enemies, p. 26h. 30 here, however, that not all such interpretations are of equal merit for Popper. In fact, he accepts some kind of continuum ranging from high-level interpretations to singular hypotheses serving as initial conditions, with "all kinds of intermediate stages."8 This would seem to mean either that there are other criteria for appraising general interpretations than evidential strength or testability, or that the criterion of testability must itself be weakened to one of degrees. In the latter case, inter- pretations would be taken as merely less testable or falsifiable than scientific theories or singular statements, instead of as untestable in principle. Let me reserve comment on the problems surrounding the notion of testability and empirical significance until a comparative analysis with Hempel's notion of incomplete explanation can be made. So far, then, we have seen that explanation in Popper's view requires a selective process, consists in the deductive sub- sumption of particular events under general laws or hypotheses, differs from prediction and testing only pragmatically but not structurally, and can often be accomplished in historical inquiry by substituting general interpretations or "theories" for scientific laws on the basis of selection and ordering. Before moving to Hempel's more detailed theory of explanation, two comments pertaining to Popper's later writings seem noteworthy for our purposes. In a new essay, "Facts, Standards aid Truth," Ibid, p. 266. 31 published in 1961 as an addendum to The Open Society and Its Enemies, Popper makes explicit two important and related points which were at best implicit in his earlier work. The first concerns indirect- ly Weber's IT thesis. For while Popper quite explicitly denies the thesis in Weber's sense, 1.3. in any way which would conflict with the CL theory, he nevertheless uses the notion of an ideal-type, limiting standard or regulative principle in characterizing the notion of the truth of explanatory hypotheses. As the essential element of a general attack on relativism or skepticism-— the view that the choice between competing explanatory theories is arbitrary-~ Popper clearly distinguishes between "knowing what truth means, or under what conditions a statement is called true" and "possessing a means of deciding-- a criterion for deciding-- whether a given statement is true or false."9 Fbllowing Kant and C.S- Peirce, he construes the idea of truth as a regulative ideal which can be approximated but not known to be achieved.10 Hence, though there is no general criterion of truth, there are criteria of progress toward truth. We can know when our theories are approximating to the ideal standard or meaning of truth, and when not. Now for our purposes his theory of truth is not so important as is his use of ideal regulating principles. For in the last sec- tion of this chapter, I will suggest that the basic tenets of the CL theory are best considered in just this manner. This is, I take 9 Ibid., p. 371. 10 Ibid., p. 376. 32 the CL theory to be a philosophic reconstruction or explication of some important and useful meanings of the term 'scientific ex- planation'. And the status of such a theory or explication I take to be that of a set of regulating principles, an ideal-type or standard for appraising and clarifying our ordinary explanations as approximations. In this way Weber's IT thesis, though incorrect- ly Opposed to the CL theory, serves to illumine one important as- pect of the issue at hand. Still, it is not entirely clear whether Popper would counte- nance such an extension of his theory of truth to apply to his theory of explanation. Though in general this extension might not be objectionable to him, the specific analysis of the CL theory which I will propose undoubtedly would. And this brings us to our second comment, which pertains to Weber's VN thesis. Unlike many supporters of the notion of value-neutrality in the acceptance of explanatory hypotheses, expecially those who appeal to a clean distinction between facts and decisions to defend the thesis, Popper has recently acknowledged the essential decisional aspects of accepting hypotheses as well as of proposing normative or value judgments. In other words, the VN thesis is often defend- ed on the Weberian grounds that the truth of value judgments depends on human decisions to adopt certain standards, while the truth of factual assertions or explanatory hypotheses does not. But Popper, in following out the consequences of his fallibilism thesis whereby no hypothesis is immune or exempt from error and criticism, finally rejects this position and concedes that we will have to decide when - w: 33 the evidence for such hypotheses warrants our accepting them. Hence, "in this sense, decisions enter into the critical method," i.e. in the sense of justifying "the tentative acceptance" of some theories as preferable to others.11 Nevertheless, POpper maintains his support of the VN thesis by shifting the grounds for it to a dualism of facts and policies or standards, instead of his earlier dualism of facts and decisions. We will, accordingly, pursue this kind of defense in some detail in Chapter VI. It might suffice for the moment to suggest a point of clarification concerning the VN thesis. The issue concerns whether or not an adequate explication of the notion of acceptable explanation would require a scientific or historical inquirer to make value judgments. It is not a question of value judgments being identical with or reducible to factual judgments; nor is it a ques- tion of the subjective or objective character of value judgments. For both of these questions, however important, are independent of the main issue. Hence, in addition to investigating Hempel's version of the CL theory, we will also attempt to see what bearing these two points concerning ideal-types and value-neutrality have had on it. 11 Ibid., p. 380. I 9%, 3h 9. Hempel's Deductive Model Overlapping the work of Popper for the past twenty years, Hem- pel's essays on explanation develop Popper's theory in a forceful, lucid and influential manner. Perhaps because of these very reasons, they have been subjected in recent years to serious critical in- vestigation which in turn has spurred Hempel to elaborate and also to modify his earlier position in important ways. To What extent such elaboration of detail raises additional difficulties, and whether the modifications amount to a retraction of the original deductive model are some of the questions to be treated below. Our main consideration will eventually rest with the question of how far the theory can be reasonably extended to include such non- natural science inquiries as history. Hempel's original essay on the theory of explanation, "The Function of General Laws in History,"12 generalizes the deductive model beyond Popper's strictly causal form, but remains essentially similar on most counts, while filling in the theory with a more de- tailed analysis of central aspects. Perhaps the prime motivation of both Hempel and Popper, clearly expressed in this essay, was and remains the rebellion against the earlier idealist tradition, arising in Germany and spreading to England and the Continent, which argued for the SU thesis and for a radical difference in kind between the explanatory methods employed by historians and those utilized in the sciences. This purported demarcation of sharp boundaries between 2 1 C. Hempel, "The Function of General Laws in Histony,' in P. Gardiner (ed.), Theories of History (Glencoe: Free Press, 1962), PP. 3hh‘3560 35 the different fields of scientific inquiry, and the consequent autonomous development of each field, was enshrined in the con- trasts between ideographic and nomothetic disciplines, unique and repeatable events, between "Geisteswissenschaft" and "Naturwissen- schaft," and between "Verstehen" and "erklaren" or "begreifen." In Opposing these basic contrasts, POpper and Hempel concur in advocating the methodological unity of all the empirical sciences. The influence of Comte, Mill and Buckle, as well as Hume, is clear. Their approach is to reform the social and humanistic domains by making them more scientific and subject to empirical controls. In our present case this reforming attitude manifests itself in their insistence upon assimilating historical explanations to scientific ones, in particular to the deductive or covering-law pattern as a prototype or model. To be sure, such an assimilation flies in the teeth of the multiple and varied arguments used by idealists and their recent analytic defenders: arguments from the uniqueness and complexity of data, from the presence of value bias and the need for empathy, from the existence of free will and self- fulfillment, from teleological causation, from the inaccessibility and non-physicality of the mind, and from the requirements of morality. Accordingly, we shall in later chapters look carefully at some of these arguments since they are persuasive enough to be revived by such critics of the covering-law model as Lavine, Schutz, Dray and Scriven. However, a good deal still remains to be said about Hem- pel's construal of the CL theory. 36 In addition to this concurrence with the thesis of methodo- logical unity, Hempel, in his early essay, indicates further doc- trines shared with Popper. Scientific explanation, for example, is said to be formulable in terms of a deductive argument con- taining general laws which are understood to be "statements of universal conditional form...capab1e of being confirmed or dis- confirmed by suitable empirical findings...and assumed to assert a regularity"13 between the initial conditions and the explanandum. Hempel and Oppenheim insist, with Popper, upon the strong logical relationship of entailment between explanans and explanandum, on the necessity of universal general laws as part of the premises, and on the empirical content or testability of these laws- all as necessary though not sufficient requirements for an explanation to be scientifically sound, and hence distinguished from both un- acceptable and pseudo explanations. In a later essay, "Studies in the Logic of Explanation," these are codified into the following four logical and epistemic condi- tions of adequacy for the soundness of any prOposed or potential explanation: (R1) The explanandum must be alogical consequence of the explanans. (R2) The explanans must contain general laws, and these must actually be required for the derivation of the explanan- dum. (But unlike his earlier essay and also Popper's 13 Ibid., p. 3&5. mm A; #9.? 7..- - “ 4’:- 37 account, Hempel no longer requires nonlawlike initial conditions. This allows explanation of laws or generali- ties as well as of particular events.) A (R3) The explanans must have empirical content, i.e. it must be capable, at least in principle, of test by ex- periment or observation. (Rh) The explanans must be true.lh And the schema for a sound scientific deductive explanation is then presented as (D) (1) L1 . L2 .... Lk (2) Cl 0 C2 coco Cm (3) E where 'Ll . L2 .... Lk' represent universal laws, 'Cl . 02 ... Cm' represent statements of initial or boundary conditions, 'E' the statement of the explanandum event, and (l) and (2) together as the explanans logically entail (3). Thus, Popper's example concerning the thread's breaking can readily be seen to fit pattern (D) since (L1), (L2), (Cl) and (02) serve as explanans and have (E) as a logical consequence. Another example, cited by Hempel, explains why the part of an oar which is under water appears, to an observer, to be bent upwards. The phenomenon is explained by means of general laws-- mainly the law of refraction and the law that water is an optically denser medium than air-- and by reference 1h C. Hempel and P. Oppenheim, "Studies in the LOgic of Ex- planation," p. 321. 38 to certain antecedent condflions-- especially the facts that part of the oar is in the water, part in the air, and that the car is practically a straight piece of wood. Thus, here again, the question 'Why does the phen- omenon happen?‘ is construed as meaning 'according to what general laws, and by virtue of what antecedent con- ditions does the phenomenon occur?l Further points of similarity with Popper are the structural symmetry or identity of explanation and prediction, i.e. that there is only a pragmatic difference of direction, interest or purpose between the two; the centrality of explanation and predic- tion as primary goals for scientific inquiry; and the ontological thesis that what is explained is not merely a type or kind of event nor a concrete event but rather an aspect, property or description of an event, 1.3. an event of a certain kind. Concerning the latter thesis, both Hempel and POpper are anxious to follow Weber by deny- ing the possibilfiy of even a complete description, not to mention a complete explanation, of a concrete individual event (such as the assassination of Huey Long) since this "would require a state- ment of all the properties exhibited by the spatial region or the individual object involved, for the period of time occupied by the event in question." Their intent of course is to undercut the idealist notion that the peculair function of history is to "grasp the unique individuality" of its subject matter by arguing that history can do this "no more and no less than can physics or chemis- try."16 We will return to this ontological thesis in connection 15 Ibid., p. 320 16 C. Hempel, "The Function of General Laws in History," p. 3&6. 39 with the distinction between complete and incomplete explanations later in this chapter. Meanwhile, pursuing the relationship be- tween explanation and prediction will perhaps help to clarify some of the point of explanatory arguments. Following Popper, Herpel maintains that the same formal analysis, including the four necessary con- ditions, applies to scientific prediction as well as to explanation. The difference between the two is of a prag- matic character. If E is given, i.e. if we know that the phenomenon described by E has occurred, and a suitable set of statements Cl, C , ...., C , L , L .... L is produced afterwards, we spea of an explanation of therphenomenon in question. If the latter statements are given and E is de- rived prior to the occurrence of the phenomenon it describes, we speak of a prediction. It may be said, therefore, that an explanation is not fully adequate unless its explanans, if taken account of in time, could have served as alpasis for predicting the phenomenon under consideration." It will be noticed, however, that in this passage two theses are really being propounded, a weaker and a stronger one. The stronger and hence more controversial thesis maintains a structural symmetry or identity between scientific explanation and prediction, while the weaker thesis merely holds that all adequate scientific explanations must have potential predictive force. We will refer to them as the "Symmetry Thesis" and the "Predictive Thesis" re- spectively. Both theses, however, concern only explanatory and predictive arguments, not statements that merely describe some past, present or future event. They refer to the logical deriva- tion or inferability of the explanandum from the explanans not to the mere assertability of the explanandum. This distinction is of 17 C. Hempel and P. Oppenheim, "Studies in the Logic of Ex- planation," p. 323. ho some importance since many attacks on the symmetry thesis seem to have misfired due to a failure to preserve Hempel's distinction. Many philosophers have pointed to clear cases of asymmetry between the assertability of scientifically predictive and explanatory statements, but such cases are irrelevant to a thesis about logical relations and inferability.18 Rather than become embroiled in the many controversies con- cerning this thesis, a task well beyond the scope of the present chapter, it might instead be more instructive for our immediate purposes to consider the Weaker predictive thesis. Some of the arguments recently used to oppose this thesis raise questions hear- ing heavily on our later inquiry about historical explanations. For example, if scientific explanations are taken inferentially as arguments instead of as descriptions, what then are they intended to show? What is the point of such explanatory arguments? How do they differ from other kinds of arguments? Or is it the case that all arguments are explanatory? To come by an adequate answer to these questions will require, of course, a fuller analysis of re- quirements (Rl)~(Rb). But a beginning can be made by examining some of the arguments against the potentially-predictive thesis recently proposed by I. Scheffler and J. Kim. In the course of defending the scientific legitimacy of ex 18 Cf. the instructive defense of this thesis by A. Grun- baum in "Temporally-Asymmetric Principles, Parity Between Expla- nation and Prediction, and Mechaiism Versus Teleology," Philosophy pf Science, Vol. XXIX (April, 1962). bl post facto explanations, Kiml9 follows Scheffler's20 lead in opposing Hempel's potentially-predictive thesis by distinguishing sharply between scientific explanation and prediction. In this way he attempts to vindicate 2x post facto explanations as legitimate scientific explanations, even though they lack any significant predictive power. While it is conceded that scientific eXplana- tions do indeed take the form of an argument or inference, it is nonetheless maintained that they do not purport to establish, support or prove their conclusions or explanandum-statements. They do not, in other words, "purport to show that the event to be explained actually took place or is taking place."21 But pre- dictive arguments, on the other hand, are intended to show just this. As attempts to gain knowledge of particular events or states by projection from known to unknown data, they are intended to substantiate or support their conclusions. Hence their premises do function as evidence for the predictive conclusion which in turn is dependent upon the premises for evidential support. Now, since POpper and Hempel concur with this view of prediction 19 "Ex post facto" explanations refer to those cases where the antecedent conditions have to be ascertained after the request for an explanations is made, and where our knowledge of the actual occurrence of the explanandum-event plays an essential evidential role in as- certain ng these conditions. 20 I. Scheffler, "Explanation, Prediction and Abstraction," Bri- tish Journal for the PhilosOphy 2f Science, 7 (1957). This entire—paper is included in revised and enlarged form in Scheffler's most recent work, Anatomy 9f Inquiry (N.Y.: Knopf, 1963), Part I. 21 J. Kim, "Inference, Explanation and Prediction," Journal pf Philosophy, LXI, No. 12, June, l96h, p. 362. b2 but still take scientific explanation to be potentially pre- dictive, the issue clearly turns on an alternative version of explanation. Scheffler and Kim point out, correctly I think, that eXplanatory and predictive arguments are not merely abstract kinds but are instead concrete arguments given at adefinite time, in a specific context, for a specific purpose. This is to say that they are "not argument-types or inference-types, but specific argument- tokens and inference-tokens." But from this pragmatic, context- bound view of arguments, they argue that explanations, unlike pre- dictions, are only attempts to systematize known events and states; not, as Hempel and.Popper suggest, attempts to establish, support or prove a conclusion, nor to show that the explanandum-event ac- tually took place. implanations are intended "merely to exhibit "22 logical relations obtaining between statements, in order to show the connection, mediated by laws, between the events described in the antecedent-conditions and the conclusion. But this conclusion is based on arguments that are at best misleading when taken as objections to the Hempelian theory of ex- planation. Hempel has, in many essays, insisted that scientific explanations must establish or support the conclusion of an in- ference, even if this is not their main task. In his most recent statement, he submits a general and necessary condition of ade- quacy for all rationally acceptable scientific explanations of a given event, 312. that "any such explanation... of the type 'why did X occur?‘ must provide information which constitutes good 22 Ibid., p. 362. 143 grounds for the belief that X did in fact occur."23 The predic- tive thesis then follows as a consequence of this condition of adequacy. Kim offers two arguments to deny this condition of adequacy. The first is that often the truth of the explanandum statement is actually known with greater certainty than that of the explanans statements. The second appeals to the fact that when we ask "why did X occur?" we presuppose or presume that X did occur. Hence, he claims, in providing an explanation we neither intend nor need to give proof or support or good grounds for our belief that X did occur. Now, I think we can concede both of these points as accurate descriptions of some ordinary and scientific practice of providing explanatory arguments. Yet neither point entails the denial of Hempel's condition of adequacy. And the underlying reason why they do not is because they concern the psychological-pragmatic aspects of explanation, while Hempel's theory is an explication or recon- struction of the logic of explanation. This is not to suggest that the pragmatic elements of explanation are unimportant or fruitless. To the contrary, we will argue in later chapters that such elements are central, even for Hempel's reconstruction. However, the point here is that recent analytic critics of Hempel's CL model have not always met the issue clearly. They have argued for the pragmatic 23 C. Hempel, "Reasons and Covering Laws in Historical Ex- jplanation," in S. Hook (ed.), Philosophy and Histogy (N.Y.: New York University Press, 1963), p. 1&6. his aspect of explanation by describing ordinary and scientific ex— planatory practices. But since Hempel is not likewise offering an alternative description, but instead a methodological prescrip- tion or explication, of these practices, there is no essential conflict between them and hence no refutation of Hempel's condi- tion of adequacy. Surely there is no conflict in explanations serving both purposes: systematizing known events and states; and also establishing, supporting or proving their conclusions. Further, while it is correct to say that actual explanatory arguments are always concrete or argument-tokens, it does not follow that one cannot profitably and accurately abstract from these con- crete cases some important logical structures and conditions or rational acceptability or adequacy as ideal types or idealizations. In fact, if one could not, it is doubtful how, or even whether, we could elicit any reasonable criteria upon which to appraise criti- cally such arguments as acceptable or not, as genuinely scientific or pseudo-scientific. In this sense the CL model of explanation, following Weber, is instructively compared to the concept of mathe- matical proof as construed in meta—mathematics. Surely all actual proofs are also concrete or proof-tokens. Yet this fact does not preclude the significant construction of a theory of proof as a theoretical account abstracted fronithe concrete cases where someone proves something to some other person at a definite time, in a specific context, for a certain purpose. Hence, it seems that the two reasons offered by Kim, represen- tative of many recent arguments concerning historical explanations hS to be considered later, are though true at best misleading When employed as objections to the CL model. Moreover, the question is not merely whether pragmatic elements enter into the analysis of scientific explanation. Indeed they do. But CL theorists have never denied that they do and their reconstructed model does not require that they deny it, so long as they are limited to the descriptive level. For in this case Hempel can reply that such objections miss their mark since they apply to his non-pragmatic or theoretical concept of explanation standards that are only proper for a pragmatic construal. The question of importance, in- stead, is whether or not the CL model itself, as an explication or recenstruction of ordinary scientific explanations, requires the inclusion of pragmatic elements. Since Hempel and POpper claim that it does not, a more reasonable objection to their theory would seem to be one showing that such elements are required for this task. Accordingly, the last chapter will be devoted to just this topic. So far, then, the CL theory emerges as an analysis of scienti- fic explanation which insists upon their status as deductive argu- ments, 1.3. as satisfying (R1). In addition, the explanans of such arguments serve to support evidentially, as well as to organ- ize and systematize, their explanandum-events, and hence have po- tential predictive power. h6 Pseudo, Genuine and Acceptable EXplanations Yet suppose the latter point to be conceded, gig. that the ex- planans of a genuine and of an acceptable scientific explanation must be capable of evidentially supporting the eXplanandum. Still, obviously not all deductively valid arguments are explanatory. Hence we need examine more closely Hempel's other conditions of adequacy, (R2)-(Rh), in order to determine what distinguishes ex- planatory arguments from others. More specifically, we will want to see how Hempel distinguishes between genuine and pseudo scienti- fic explanations, as well as between those genuine cases which are rationally acceptable and those which are not. To make a start in this direction, we might ask for a defense of (R2). Even conceding (R1), why and in what sense must the ex- planans contain essentially-occurring general laws in order to pro- vide adequate support for the explanandum? For surely any singular statement can be deduced from some set of premises, none of which are of the form of universal laws, i.e. of the form 'All A is B' or '(x) (Ax Bx).' Any defense of (R2), accordingly, must be in- dependent of the reasons for maintaining (R1) or deducibility. Moreover, a defense of (R2) will become important for our purposes in later chapters. For many critics of the CL theory, and proponents of alternative theories of historical explanation, rest their case on the inadequacy of (R2) and on the subsequent claim that scientific and historical explanations can be genuine, complete and rationally acceptable without containing general laws. The question at issue then is why Hempel takes (R2) to be a necessary condition of the adequacy of genuine and acceptable explanatory arguments. h? Hempel, unfortunately, has not explicitly formulated a de- fense of (R2). Instead, his writings reveal brief relevant comments to this issue, followed by mostly futile attempts to answer quite a different question: what is meant by a law or a lawlike hypothesis? But however important and difficult this question may be, an answer to it will surely not serve to defend (R2). ‘We will do well, con- sequently, to start from POpper's allegiance to Hume, mentioned earlier, an allegiance also shared by Hempel. Our earlier comments, when conjoined with Hempel's brief defense, suggest two primary arguments to support (R2) as a necessary condition of adequacy. The first is an argument from the meaning of 'explanation' and such closely associated terms as 'cause' and 'because! As our earlier reference to POpper indicated, part of what it means to say 'A caused B' is "that events of kind A are always and everywhere followed by events of kind B," i.e. that A and B are nomologically connected. In other words, the very meaning of statements used as evidence, reasons, causes or explanations is such that they are at least implicitly general, that they presuppose generalities or laws which serve to connect the events in question. For example, to say that some piece of thread broke because a two-pound weight was put on it is to say that the same kind of effect will be produced in all relevantly similar cases where the same kind of cause is present. For if one were to deny the latter generalization and still hold that the events and circumstances were relevantly similar, we would be puzzled as to what one meant by the 'because' in the for- mer statement. Hence, 'A caused B,’ 'A is a reason for B' and 'A h8 explains B' all seem to be incomplete or elliptical statements. They are elliptical in the sense that they are relative to or depen- dent upon the appropriate generalization.2h Much of the point of this argument, then, turns on the epistemological question of the nature of empirical explanatory or causal statements, and on the Humean answer that all empirical knowledge requires as part of its meaning an appeal to regularities or laws. The second argument, though closely related to the first, can be developed independently as an argument from challenge. It concerns the way one might defend, say, a causally explanatory statement of the form 'A caused B', if challenged. While a scientist or historian might not always mention a law in the explanation he offers, still, CL theorists argue, in order to defend such an ex- planation or causal connection against challenge, he would have to invoke some lawlike connection. Only in this way could he claim objectivity for his statement. Inability or unwillingness to specify the lawlike connection would mark the statement as subjec- tive and hence not an objectively genuine explanation at all. Another way of putting this is to say that a scientist's personal explanation makes a claim to being "an explanation," a genuine scientific explanation and not merely a pseudo one. It makes a claim to be more than just his personal explanation. Thus, any explanation will be genuinely empirical and scientific only if it is objectively defensible, and it will be objectively defensible 2h Cf. the instructive parallel treatment of moral terms in M. Singer, Generalization 2.“. Ethics (N.Y.: Knopf, 1961), pp. 3u-6o. D9 only if it presupposes the truth of some empirical lawlike generali- zation which warrants connecting the events in question. On the basis of these two arguments, finally, CL theorists maintain that not all deductively valid arguments, but at best only those containing statements of general laws, are explanatory. Yet, of course, not just any lawlike generalization will be genuinely explanatory either. For (R ) remains to be invoked. In addition 3 to containing essentially general laws, genuine and acceptable scientific explanans must also contain empirically testable or falsifiable statements, 1.2. statements with empirical content or import. However, the question of what constitutes an empirically testable or significant statement has of yet received no generally acceptable answer, not even among CL theorists. Hempel, in fact, wiuld be the first to admit that there is little likelihood of finding such a general criterion applicable to all or even most scientific ex- planations. ‘We might even have to learn to live with degrees of testability, with some explanatory systems having more than others.25 In any case, there does seem general agreement that some cases must be ruled out on the basis of (R3). And while we are unable to make a clear distinction in all cases, it surely does not follow that 25 Cf. C. Hempel, "Problems and Changes in the Empiricist Cri- terion of Meaning," in L. Linsky (ed.), Semantics and the PhiIOSOphy of Language (Illinois: University of Illinois Press, 1952); C. Hempel, “The Concept of Cognitive Significance," Proceedings of the American Academy of Arts and Sciences, LXXX ( l9Sl—Shvjjv6I; and I. Scheffler, TEE-inatomyjéi Inquiry, Part II. So anything goes, that any purported explanation is thereby genuine, or that the line between empirically significant and pseudo expla— nations cannot be drawn in particular cases. Instead of pursuing the complexities and problems of the issues surrounding (R3), let me turn to a related but more general issue. Having examined briefly Hempel's first three conditions of ade- quacy, it is still not clear what they are conditions of. Are they, §.g., conditions marking off genuine from pseudo explanations, or are they intended instead to distinguish between those genuine ex- planations that are rationally acceptable and those which are not. Unfortunately, neither POpper nor Hempel has been either explicit or clear concerning these questions. Popper, of.course, has for some time used the criterion of falsifiability to demarcate between genuinely scientific and non-scientific but still empirical expla- nations. But it is not clear whether he also intends this criterion to demarcate between acceptable and unacceptable genuinely scienti- fic explanations. Again, Hempel at times seems to take the CL theory as an ideal standard of genuine scientific explanation in contrast to pseudo- explanations; but at other times, particularly when commenting generally on the "conditions of adequacy" he is clearly trying to distinguish "sound" or acceptable explanations from inadequate or unacceptable, though genuine and not merely pseudo, explanations. For example, in describing a potential danger of motive or teleo- logical explanations as that of lending itself "to the facile con- struction of SE pgst facto accounts without predictive force," A“ w—w‘y i Sl Hempel never makes clear whether his objection to such cases, which violate (R3) and hence lack cognitive significance, is that they are pseudo, 2.3. merely "alleged motivational explanaticrs,” or that they are unsound or unacceptable while still genuine. Again, Hempel distinguishes in various essays between a po- tential explanation which satisfies only (R1) - (R3) and an actual explanation which in addition satisfies either (Rh) or its weakened version of high confirmation. An argument will be an actual explanation only if its premises are in addition true or highly confirmed, and hence actually do ex- plain its conclusion. Since we obviously can find many potential explanations for any given event, the problem is to find hypotheses which actually explain, which are either true or highly confirmed. Another way of putting this point is to distinguish between a merely valid argument and a sound or rationally acceptable argument. For then a potentid. explanation will be formally valid but not neces- sarily empirically sound, while a sound explanation will also con— tain rationally acceptable premises or explanans. However, the issue now concerns the ambiguity of the term 'actual' explanation. Clearly Hempel does not mean by this that any argument anyone ac- tually intended to be explanatory in ordinary affairs was thereby genuinely explanatory, for such cases might even violate either (R1), (R2) or (R3) and thus not even be potential explanations. But we are still left with our original puzzle: does 'actual' ex- planation mean 'an explanation,' a genuine explanation, or does it mean a good, sound, acceptable or better explanation? 52 Professor S. Barker has recently complained of the CL theory that "it precludes the giving of any real account of what it is for one explanation to be better than another," of how we are "to choose among competing explanatory theories."26 It would seem then that if Hempel construes "actual" explanation as "sound," "good," "scientifically acceptable" or "better than" others, he is obliged to meet Barker's demand, to indicate some criteria for the application of the latter predicates, as an essential task of his theory of explanation. And in this case (R1) - (Rh) could be construed as just such criteria or requirements, for (Rh) in parti- cular can be used as a criterion for such a choice. Hence, as Rudner has noted in reply to Barker's charge,27 nothing in the CL model in any way precludes such criteria for choosing the better among rival putative explanations. However, Rudner's particular defense of Hempel against this charge depends on our taking "actual" explanation not as good, sound, acceptable or better, which would require criteria for such, but as merely genuine, "not pseudo,“ "merely putative" or "an explanation," and hence not requiring such criteria. This becomes clear in his reply that "at any rate lack of a criterion for constituting a better explanation does not entail lack of a criterion for consti— tuting an explanation at all."28 26 S. Barker, "The Role of Simplicity in Explanation," in Feigl and Maxwell (eds.), Current Issues in the PhilosoPhy 2; Science (N.Y.: Holt, Rinehart and Winston, 1951}, p. 273. 27 R. Rudner, "Comments," in ibid., p. 28h. 28 Ibid. 53 This, of course, is true. But if the CL model is merely es- tablishing criteria of adequacy for an argument to be "an expla- nation," as Rudner squests, instead of being a sound, acceptable or better explanation; then Hempel's objection to ex post facto explanations of, say, motivations or reasons for an action must be that they are not even explanations, that they are merely pseudo- explanations, rather than that they are just unacceptable, unsound or not the best explanations. In other words, lacking predictive power, adequate confirmation or cognitive significance, for example, suffices to incriminate putative explanations as not actual and indeed as not even potential. We seem driven then to the same predicament as above. For, if "actual" explanation is taken to mean simply "an explanation," or a genuine explanation, the oppo- site of pseudo explanations, then we are forced to concede that most of what we take to be competing explanations of some phenome- non really are not explanations at all. Consequently, just as we will find Hempel finally acknowledging the "questionable merit" of his early defense of (Rh) on these grounds, so it seems we must reject Rudner's construal of "actual" explanation as "an explana- tion," and hence depict it instead as meaning "scientifically acceptable," or "better" or "sound." Though this does not preclude some such explanations being adjudged better than others. But then we are still free to take "potential" explanation as meaning "an explanation." In this case there can indeed be many competing genuine explanations of the same phenomenon, gig. all that meet criteria (R1) - (R3) or all potentially explanatory sets of Sh of hypotheses. Yet at the same time not all of these will be the best, correct or most acceptable, since not meeting (Rh)' Now, if the above argument is cogent, 1.2. if the CL theory of explanation requires criteria of rational or scientifically acceptable explanations, serious problems will arise in regard to Hempel's acceptance of Weber's value-neutrality thesis of science, based as it is on his sharp distinction between pure and applied science, especially when the CL model is extended to cover probabilistic as Well as deductive explanations. For if it can be shown, as I will attempt to do in the final chapter, that rationally acceptable statistical explanations require pragmatic criteria and the making of decisions and value judgments; then the CL theory of scientific explanation, as espoused by Hempel and Popper, will re- quire the denial of the value-neutrality thesis as an essential ingredient. And it is largely because this denial depends essen- tially on a pragmatic construal of the concept of explanation that we will analyze CIOSely the recent criticisms of the CL model and the subsequent reconstruction of Weber's position, made especially by William Dray. However, one major thesis of the present work will be that Dray's defense of the extension of the notion of explanation so as to include a pragmatic dimension, and his consequent grounds for denying the value-neutrality thesis, are misplaced and need to be redirected. His case for the SU thesis, for example, does not adequately support his accompanying objections to the CL theory of explanation. In other words, I submit that Hempel's advocacy of 55 the CL theory as presented in this chapter and Dray's denial of the value—neutrality thesis are not incompatible positions, as Hempel and Dray seem to believe. But much remains to be done before attempting a defense of this position. Let us therefore look more closely at Hempel's epistemic condition of adequacy for empirically sound or acceptable scienti- fic explanations. We will want to ask, in particular, whether (Rh) is necessary to distinguish between those genuine scientific ex— planations whichare acceptable and those which are not, or whether this condition is too restrictive and hence requires weakening to that of high confirmation. Clearly, the three logical requirements mentioned above, (R1) - (R3), are not alone sufficient to guarantee the soundness, rational acceptability or adequacy of a scientific explanation. Some sort of empirical or epistemic condition must supplement these three formal requirements. Not as evident, however, is what this condi- tion must be. Aristotle, §.g., stringently required that the pre- mises be true, be known to be true and be “better known" than the explanandum or conclusion. But as Professor E. Nagel clearly in- dicates in his recent illuminating book,29 the latter two require- ments are unacceptable as originally presented by Aristotle. That the premises be "better known" than the explanandum refers of course to Aristotle's metaphysical notions of "necessary" objects of scientific inquiry and.b Tirst" principles. The former, universals, 29 E. Nagel, The Structure of Science (N.Y.: Harcourt, Brace and World, 1961), ppt_h2-h6. Cf._Aristotle, Posterior Analytics, in W. Ross (Ed.), Complete Norks 2f Aristotle (N.Y.: Random House, l9hl), 71b10-72a10. 56 are contrasted with merely contingent particulars which are not proper objects of science; and the latter, first principles, are those better known by nature in contrast to what is better known to man from sensory perception. Of more immediate concern is the Hempelian position concern- ing the first two of Aristotle's requirements. That the premises be "known to be true" in order for an explanation to be scientifi- cally acceptable likewise presents difficulties. For, as Nagel argues, few if any accepted scientific explanations meet this con- dition and hence would be satisfactory according to it. At best scientific hypotheses or laws seem to be known with more or less degrees of confirmation or probability. Hence, if the requirement is insisted upon, it would simply lead in practice to the intro- duction of a new term to distinguish "known to be true" from "known with high confirmation."30 Nevertheless, these considerations suggest that the stipula— tion might be more acceptable in a weaker though perhaps more vague form instead of merely discarded altogether. In fact, this is the alternative Hempel opted for initially when he maintained that, along with a set of initial conditions, "the scientific explanation of the event in question consists of ... a set of universal hypo- theses, such that the statements of both groups are reasonably Eel; confirmed by empirical evidence...."31 Let us refer to this weaker 30 Ibid., p. u3. 31 Hempel, "The Function of General Laws in History," p.3h5. S7 condition as (Rb'). However, in a later essay the requirement was changed from this modified version of Aristotle's second condition to his first condition, 2.3. from the epistemic requirement of "reasonably well confirmed" to the much more rigorous requirement (Rh)’ that the explanatory premises or explanans be true. Now, the reason for this change deserves some attention. It is not that the weaker requirement is vague or that no precise and generally accepted standard is available for judging when an hy- pothesis is "reasonably well confirmed by empirical evidence," even though this is no doubt the case. Instead, Hempel defended (Rh) by arguing that (Rb')’ the well-confirmedness requirement, leads to "awkward consequences," viz. to a relativized concept of expla- nation. Suppose that a certain phenomenon was explained at an earlier stage of science by an explanans which was well supported by the evidence then at hand, but which had been highly disconfirmed by more recent empirical findings. In such a case we would have to say that originally the explanatory account was a correct ex- planation, but that it ceased to be one later, when unfavorable evidence was discovered. Furthermore, the awkwardness and hence the erroneous aspect of this temporal relativization consists in its counterintuitiveness, 1.2. in the fact that This does not appear to accord with sound common usage, which directs us to say that on the basis of the limited initial evidence, the truth of the explanans, and thus the soundness or the explanation, had been quite probable, but that the ampler evidence now available made it highly probable that the explanans was not true, and hence that the account in question was not-- and had never been-— a correct explanation. 32 p0 3220 Hempel and Oppenheim, "Studies in the Logic of Explanation," 58 The charge of counterintuitivity results from Hempel's belief that according to sound common usage the correctness of a given explanation is independent of temporal factors, just as is the truth of a given statement. Still, it remains unclear how, or even why, common usage or counterintuitivity serves as a relevant criterion for judging such cases. And this is not, it would appear, an unimportant consideration. Hence it is unfortunate that Hempel fails to elaborate upon the relationship between his theory or explication of the notion of scientific explanation and the ordinary usage(s) of the term. At any rate, while this issue has been pointedly pressed by recent critics and will require subsequent treatment shortly, M. Scriven helps to clarify the situation some- what by directing our attention to another facet of Hempel's de- fense of the requirement of well—confirmedness.33 A fundamental defect of this seemingly persuasive argument is the ambiguity discussed earlier concerning the term 'explana- tion.’ For Hempel's defense turns on a shift from an analysis of 'explanation' which admits of many legitimately competing explana- tions of the same phenomenon some of which are not well-confirmed, 1.3. as a possible potential or genuine explanation; to an analysis which does not countenance such competition, 1.2. as a "correct" or acceptable explanation or even in some cases as "the" explanation. 33 M. Scriven, "Explanations, Predictions and Laws," in Feigl and Maxwell (eds.), Minnesota Studies 12 the Philosophy of Science, Vol. III (Minneapolis: University of Minnesota Press, 1553), pp 0 190-1. 59 If the truth-requirement were upheld, surely an even more awkward consequence would result. For then what would the false explanation be, if not an ex- planation? What would an invalid argument be if not an argument, or a false proposition? Surely Hempel's proposal of (Rh) is at least as counterintuitive as the requirement of well-confirmedness (Rh')° Besides the tale that lies therein concerning the use of counter- intuitivity as a criterion for deciding such cases, have we not good reason to accept (Rh')? No doubt the correct or better ex- planation is obtained only when we have uncovered true premises, at least ideally. Still, the only way of discovering which genuine explanation is likely to be true and hence to satisfy (Rb) is by employing the notion of evidential strength and choos- ing the one with the highest degree of confirmation. As the result of such considerations Hempel, in his most recent and most complete analysis of the logic of explanation, "Deductive—Nomological vs. Statistical Explanation," acknowledges the "questionable merit" of his earlier defense of (Rh)’ the truth-condition, in the following passage. For in reference to explanations as well as in reference to statements, the vague idea of correctness can be con- strued in two different ways, both of which are of in- terest and importance for the logical analysis of science: namely, as truth R in the semantical sense, which is independent of any eference to time or to evidence; or as confirmation by the available relevant evidfince Rh' -- a concept which is clearly time dependent.3 3b Hempel, "Deductive-Nomological vs. Statistical Explanation," in Minnesota Studies 12 the Philosophy pf Science, Vol. III, p. 102. 60 He then proceeds to distinguish between true explanations and those that are more or less well confirmed by a given body of evidence. Accordingly, he defines a "potential explanation" or genuine scientific explanation as one meeting (R1) - (R3), 1.2. as one whose explanans need contain a set of statements, L1, L2 ..., Lm, which are empirically testable and also lawlike (1.3. laws except for possibly being false) instead of necessarily being laws and hence true. In turn, "true explanation" and'Well- confirmed explanation" are derivatively defined as potential ex- planations whose explanans satisfy'(Rh) or (Rh') respectively.35 The upshot then seems to be that agenuine explanation must satis- fy conditions (R1), (R2) and (R3). But in order to qualify as scientifically adequate or correct and hence to provide complete understanding of why something did or will occur, genuine expla- nations must also meet either (Rb') or (Rb)‘ 35 Ibid., pp. 102-3. 61 C. Hempel's Probabilistic Mgdgl So far we have attended almost exclusively to the deductive- nomological model of explanation, endorsed commonly by Popper, Hempel and all other CL theorists. But before considering the notion of complete explanation, and in what sense explanation 325 potential pre- diction constitutes an adequate pattern of a complete explanation, mention must first be made of another pattern of explanation more recently endorsed by Hempel, one with no little import and interest for historical explanations and the value-neutrality thesis. This pattern is of course a non-deductive, probabilistic, inductive or statistical systematization of explanation. And its only difference with model (D) lies in the fact that the lawlike statements in the explanans can be statistical. This requires a weakening of (R1) to the logical relationship of inductive probability between explanans and explanandum, which we shall call (Rl'). Lest this appear as a recent innovation or stipulative exped- iency, it should be remarked that in his initial essay on explana- tion, Hempel suggested that the deductive pattern was not the only ideally complete model of explanation. Many an explanation offered in history seems to admit of an analysis of this probabilistic kind: if fully and explicitly formulated, it would state certain in- itial conditions, and certain probability hypotheses, such that the occurrence of the event to be explained is made highly probable by the initial conditions in View of the probability hypotheses. And again in "Studies in the Logic of Explanation,I he and Oppen- heim refer to the subsumption of the explanandum under statistical 36 Hempel, "The Function of General Laws in History," pp. 350-1. 62 laws. But here they recognize that "Analysis of the peculiar logical structure of that type of subsumption involves special 5. ..37 problem Thus it is clear that while these essays are restricted to an analysis of the deductive causal type of explanation, Hempel makes no claim that this pattern constitutes the only kind of legitimate or genuine scientific eXplanation. This fact is of some import since many critics of the CL theory rest their objec- tions on this very claim, much to the detriment of their arguments. Perhqas such an oversight is excusable in a sense, however, be- cause of Hempel's failure to do more than mention the existence of a different kind of explanatory pattern in these early essays. His failure to elaborate its status and "peculiar logical structure," is, I think, partly responsible for some of the widespread mis- conception of his own views. Not until publication of much of this criticism did Hempel finally elucidate the probabilistic-nomologi- 38 cal or statistical pattern. Perhaps of even more importance for this misunderstanding is the fact that other defenders of the CL model, such as Popper and Professor M. Brodbeck, have still not 37 Hempel and Oppenheim, "Studies in the Logic of Explanation," p. 32b. My italics. 38 Hempel has elaborated this probabilistic version of the CL theory in greatest detail in "Deductive-Nomological vs. Statistical Ex- planation," but less complete accounts also appear in the following essays: "The Theoretician's Dilemma," Minnesota Studies in the Philo- sthy‘gf Science, Vol. II, 1958; "The Logic of Functional—Analysis," in L. Gross (ed.), Sym osium on Sociological Theory (Evanston, Illinois: Row, Peterson, l9§9;; "Inductive Inconsistencies," in Logic and Lan— guage (Holland: Reidel Publishing Co., 1962); "Reasons and Covering Laws in Historical Explanation;" and "Explanation in Science and History," in R. Colady (ed.), Frontiers 2f Science and Philosophy (Pittsburgh: University of Pittsburgh Press, 19627: 63 relinquished the exclusive claim of the deductive model. Unlike the deductive-nomological systematization (D), which contains laws and theoretical principles of strictly universal form, probabilistic or inductive explanations account for a given phenomenon nomologically by reference to laws of probabilistic- statistical form. Such statements usually assert that if certain specified conditions are realized, then an occurrence of such and such a kind will come about with such and such a statistical proba- bility, roughly with long-run relative frequency. The basic laws of genetics, the fundamental principles of quantum mechanics, and the laws of radioactive decay are examples of such probability statements used in science for the systematization of various em- pirical phenomena. As an illustration, Hempel suggests that the subsiding of a violent attack of hay fever in a given case might well be attributed to, and thus explained by reference to, the administration of 8 milligrams of chlor—trimeton. But if we wish to connect this antece- dent event with the explanandum, and thus to establish its explanatory significance for the latter, we cannot invoke a universal law to the effect that the administra- tion of 8 milligrams of that antihistamine will invariably terminate a hay fever attack: this simply is not so. What can be asserted is only a generalization to the effect that administration of the drug will be followed by relief with high statistical probability.... Hence the explanans will take the following form: John Doe had a hay fever attack and took 8 milligrams of chlor-trimeton. The probability for subsidence of a hay fever attack upon administration of 8 milligrams of chlor-trimeton is high. 39 Hempel, "Explanation in Science and History," p. 13. 6b Since the logical connection between this explanans and the explanandum, "John Doe's hay fever attack subsided," is clearly not deductive, the form of the logical transition not being uni- formly truth-preserving, the truth of the explanans makes the truth of the explanandum at best likely or "practically certain." The requirement of deducibility (R1) is thus weakened to that of probability (Rl'). Such an inductive or probabilistic-nomological systematiza- tion can be represented by the following schema: (P) Fi p (O, F) is very high Oi Here the explanandum, expressed by 'Oi', the fact that in this particular instance, i, (John Doe's allergic attack), an outcome of kind 0 (subsistence) occurred, is explained by two explanans- 1’ C2, ... Ck in (D), asserts that in case i, the factors F were realized. The second, sentences. The first, 'Fi', corresponding to C a law of probabilistic form, states that the statistical probability for O to occur in cases where F is realized is very high or close to 1. Finally, the double line represents the logical relation of inductive probability , high confirmation or likelihood, in contrast to that of deductive implication in (D). Hempel, following Carnap's account, also stresses the distinc- tion between the two kinds of probability statements, between the notion of likelihood and that of statistical probability, occurring in (P). Statistical probability concerns the long-run relative 65 frequency with which one occurrence (say F) is accompanied by another (say 0) and hence is a relation between kinds of occurrences. The former notion of likelihood, however, is a logical relation between statements and refers to the degree of rational credibility, evi- dential strength or of inductive support conferred upon the expla- nandum by the explanans. Or, in Carnap’s terms, it is the logical or inductive probability possessed by the explanandum relative to the explanans. The covering-law theory of explanation finally emerges, then, as two distinct patterns even though each refers to a certain kind of subsumption under covering law, statistical or strictly univer- sal. The difference between them lies in the character of the laws invoked, and hence in the logical relationship linking premise and conclusion. If it be asked whether patterns (D) and (P) are really or essentially distinct logical models, an affirmative answer can be supported by noticing the following quite distinct fundamental logical characteristics. The deterministic 'because' is a deduc- tive, either-or, unambiguous relation. The statistical one, how- ever, is an inductive relation, admitting of degrees, and exhibiting an ambiguity which calls for relativization to the total evidence available. Moreover, this very difference gives rise to many complex problems about statistical explanations. For example, one of the most compelling aspects of model (D) is the requirement that the explanans provide reasons for ruling out the possibility of the 66 explanandum-event failing to occur, and hence for showing conclusively why it actually did occur. In other words, (R1) requires the ex- planans to focus attention on precisely what was to be explained. But once the connection is weakened to (Rl'), a suspicion arises that the statistical laws abandon this focus on the particular case, since they are compatible with both the occurrence and non-occurrence of the particular explanandum-event. If this is so, then in what sense, we might ask, does a pro- babilistic explanation offer any explanatory understanding? What constitutes its explanatory force or import? To answer this ree quires showing how and why statistical laws lose their hold on individual case, in what way they are compatible with both E and non-E, and what additional requirements can be imposed on model (P) to eliminate this objectionable feature. Some aspects of this problem, which turn on the peculiar ambiguity or inconsistency of inductive explanations, will be examined in our final chapter, since they bear heavily on the value-neutrality thesis. From this brief characterization of model (P), however, it should be clear that for any probabilistic explanation to qualify as scientifically adequate it must, like deductive ones, satisfy (R2), (R3) and either (Rh) or (Rh') by containing empirically testable and at least high- ly confirmed general laws with potential predictive power. The question might arise, at this juncture, as to the nature and status of these two models or patterns, (D) and (P), particuw larly in view of the relationship between Hempelian explanation and prediction such that an explanation of form (D) or (P) is not 67 complete unless it is potentially predictive or might have served as a prediction. To be sure, seldom if ever are explanations in ordinary practice, historical or even scientific inquiry ever complete in this sense of the term. Seldom do our explanations satisfy the four conditions of adequacy stipulated by Hempel. The charge, accordingly, that such models are too rigorous or too far removed or abstracted from our usual explanatory practice to re- flect adequately such practice will surely be raised. After all, it seems not an unimportant fact that we cannot find a clear unam- biguous case of a complete explanation in Hempel's sense. It is not then a useless category? Would it not really be more scienti- fically and phiIOSOphically fruitful to replace Hempel's sense of "completeness" with one, or perhaps many different kinds, actually manifested in our scientific practice. This kind of objection, raised by both idealists and some followers of Wittgenstein and Ryle, usually receives an official reply from Hempelians, as indicated earlier. It amounts to the counter-claim that the task of the phiIOSOpher is not merely to record, mirror or describe the actual explanatory practice of working scientists or historians, but rather to construct a general theory in which these practices receive a systematic analysis, codification or rational reconstruction. Use of the term "model," Hempel suggests, reminds us that the two types of explanation as characterized by (D) and (P) constitute "ideal types or theoretical ..hO idealizations, and as such provide explications of certain modes ho Ibid., p. 15. V ' I ,' I' r ‘ - AV 3 h n 4.. __r i I O 1‘ h ‘ q . r , if N I ‘ - .' . . 'x a 68 of scientific explanation. In most illuminating and persuasive fashion, he compares them to the concept of mathematical proof as used in meta-mathematics, a concept also regarded not as a mere descriptive account of how mathematicians actually formulate proofs, but as a theoretical model or ideal standard to which the actual proofs only approximate. Such a theoretical model also serves additional functions, for actual explanations as well as actual proofs. It exhibits the rationale of explanations by revealing their logical structure, provides standards for a critical appraisal of any explanation of the kind governed by the model, and affords a basis for a theory of explanation, prediction, confirmation and related concepts. In sum, complete explanations are not attainable goals or objectives but rather ideals which, though unattainable in our actual expla- natory practice, may still be approached or approximated closer and closer. Hence, in this sense, an explication or theory of ex- planation is said to be treated theoretically as context-free. It is related to and respectful of, but not bound by, our actual usage of the term in actual contexts of application. This brief characterization or outline, barely indicated by Hempel, gives rise then to many additional problems. Can we speci- fy more precisely, §.g., in what sense the two models are context- free? And how they are related to actual contexts? Is it in fact possible to exhibit the rationale of explanations and to provide standards for their critical appraisal by the use of idealized models without introducing, as POpper suggests, pragmatic considerations 69 of intentions and purposes associated with actual explanatory con- texts? And if not, what sort of controls, if any, are there on the idealized models themselves? How, e.g., do we determine what the ideals of science are at any given time? Are they relative to time and our scientific evidence or are they somehow purely formal or perhaps philosophic matters for which empirical evidence is irrelevant? Moreover, can we specify some of the ways in which actual explanations offered in the sciences, and in history, fall short of the ideal of completeness? Is it possible to arrange them in some order of degrees of approximation to the ideal? And finally are there different senses of "completeness". In the remaining section of this chapter I want to consider two major aspects of these questions. First we will examine the notion of completeness, some of the kinds of approximations to ideal completeness and what it is that we can completely explain. The theoretical, context-free aspect of phiIOSOphic explication will then be discussed in this context, where POpper's pragmatic emphasis on explication will be further developed. l 1 'l I. . v 1‘ v‘. . l . . "r .. ‘ '. ‘ i g , 70 'l Complete Explanation and Approximations we have already suggested that according to the CL theory any pur- ported explanation which violates (R1), (R2) or (R3) fails as even ”an explanation"; it is a pseudo instead of a genuine scientific ex- planation. And since deducibility is largely where you find it, being independent of the presence of laws, while testability is built into the notion of an empirical law, much of the emphasis on whether or not an eXplanation is complete revolves around (R2),‘i.g. the in- clusion of empirical laws among the explanans statements. Moreover, there seem to be many important kinds of purported explanations, es- pecially those preposed by historians, which fall short of model (D) or (P) but which would not ordinarily or pre-analytically be considered pseudo explanations. Hence, rather than do violence to ordinary practice by construing all violations of (R1) - (R3) as pseudo expla- nations, Hempel has from his initial essay countenanced a separate category of incomplete cases. Such explanations fall between com- plete and pseudo ones. They can also be considered, I think, as de- grees of deviation from or approximations to the appropriate model as an ideal type, standard or regulating principle. we will limit our discussion to three different degrees of deviation from model (D), but many more could perhaps be specified. First of all, when presenting an explanation, a scientist or historian will often merely omit mention of some statements which he presupposes implicitly in his argument. Judged by ideal stan- dards, the argument will be incomplete in the inessential sense of being an elliptical or enthymematic formulation. When we explain, _ ll 71 §.g., that a car radiator broke because it was left in the cold and was filled with water, or that a small rainbow appeared in the spray of the lawn sprinkler because the sunlight was reflected and re- fracted by the water droplets, we tacitly assume certain general laws or particular facts, which we assume others can readily supply and which could be explicitly cited so as to yield a complete ar- gument. Hempel offers two reasons to account for why most explana— tions offered in history and sociology are thus elliptical: either the universal hypotheses are so familiar to everyone from ordinary experience as to be tacitly taken for granted, or it is too diffi- cult to formulate them explicitly with specific precision without loss of empirical content. It is conceded, too, that "in many cases, the content of the hypotheses which are tacitly assumed in a given eXplanation can be reconstructed only quite approximately."hl Another, more essential and important, degree of approximation deviates still further from the theoretical model. For often, even when we have reasonably reconstructed the implicit hypotheses assumed or taken for granted, they, together with the statements explicitly stated, explain the given explanandum only partially. This kind of deviation or incompleteness is of special interest in social, psy- chological, and historical explanations. Such explanatory arguments are referred to as partial and usually occur in cases of explaining particular events functionally, i.g. according to the function or role the item serves in the operation or maintenance of some larger hi Hempel, "The Function of General Laws in History," p. 350. 72 system. The explanans statements of such arguments, though pro— viding deductively conclusive evidence for expecting some item or member of a class of events K to occur, offer only inductive sup- port for expecting the occurrence of the particular explanandum item or event X as a member of K. In other words, while the explanans completely explains why some member or other of K had to occur, it only partially or inductively explains why the particular member was X instead of some other member of K. In case X were the per- formance of a given kind of action K, a partial explanation would consist in explaining deductively why some action or other of kind K had to occur, but only inductively why the particular action X did actually occur. The main point of partial explanations can be illustrated by Freud's account of a written "slip" in his "Psychopathology of Everyday Life." On a sheet of paper containing principally short daily notes of business interest, I found, to my surprise, the incorrect date 'Thursday, October 20th,' bracketed under the correct date of the month of September. It was not difficult to explain this anticipation as the expression of a wish. A few days before I had return- ed fresh from my vacation and felt ready for any amount of professional work, but as yet there were few patients. On my arrival I had found a letter from a patient an- nouncing her arrival on the 20th of October. As I wrote the same date in September I may certainly have thought 'X ought to be here already; what a pity about that whole monthl', and with gais thought I pushed the cur- rent date a month ahead. h2 S. Freud, The Basic Writings pf Sigmund Freud (N.Y.: Modern Library, Random House, l938):fp. 89; cited by C. Hempel, ”Explanation in Science and History", op. cit., p. 17. Q; 73 Freud's intended explanation is clearly incomplete in the sense of being elliptical since it inplicitly assumes but does not explic- itly mention any general laws or hypotheses connecting the slip of the pen 8 with the subconscious wish W and other antecedent circumstances. There seems little doubt, however, that his ex- planation relies on some such hypothesis as (1) "when a person has a strong, though perhaps unconscious, desire, then if he commits a slip of the pen, tongue, memory, or the like, the slip will take a form in which it expresses, and perhaps symbolically ful- fills, the giv n desire.""‘3 Yet even if the reconstructed hypothesis (1) is included in the explanans together with the appropriate singular statements, the resulting explanans still does not permit the deduction and hence the complete explanation of the explanandum. Since Freud's subconscious wish could, of course, be expressed and symbolically fulfilled by many other kinds of slips of the pen than S, the explanans at best permits deduction of the more indeterminate con- clusion that Freud's slip "would, in some way or other, express and perhaps fulfill his subconscious wish." In other words, the explanans does not imply, and thus fully explain, that the particular slip, say 8, which Freud committed on this occassion, would fall within the narrow class, say w, of acts which consist in writ- ing the words 'Thursday, October 20th'; rather, the explanans implies only that 8 would fall into a wider class, say F, which includes W as a proper subclass, and which consists of all acts which would express and sym- bolicallyhflulfill Freud's subconscious wish 33 some way 23 other. h3 Hempel, "Explanation in Science and History,“ p. 17. hh Ibid., pp. 17-18. Cf. also Nagel's instructive analysis of partial explanations in The Structure pf Scierxce, pp. 552-558. 7h Clearly then, this kind of incompleteness constitutes a much more serious form of incompleteness than an elliptical exphana- tion. But more importantly, we might note at this point that it also raises a question of the relation between Hempel's two models, (D) and (P). For, at least in examples like that of the "Freudian slip," Hempel seems to consider part of the incomplete- ness as due to the fact that Freud's reconstructed explanation is of the probabilistic form (P) and hence falls short of (D). If so, any explanation of the form (P) would for that very reason be an incomplete explanation. Hence instead of having two fundamen- tally distinct and equally complete but different kinds of ideal models, we would possess only one covering law model of complete explanation, gig. the deductive model (D). If so, it would appear that such critics of the deductive model as Dray and Scri- ven are not guilty of misrepresenting Hempel's proposed explica- tion. However, Hempel has explicitly denied that (P) is any less an ideal model than (D). But if this is the case, i.e. if (P) is an autonomous ideal model, then Hempel is surely required to elucidate, more than he has, in what way the probabilistic model can be complete in itself, i.e. what its completeness con- ditions are. Professor Brodbeck, in a lively and lucid defense of the de- ductive model, opts for the former position of only one model,b'S while Hempel's recent treatment of the subject would indicate a hS M. Brodbeck "Explanation, Prediction and 'Imperfect' Know- ledge," in Minnesota Studies, Vol. III. pp. 238-9. 75 preference for the latter alternative of two independent models and an attempt to codify some of the required conditions. And it is these conditions of adequacy for probabilistic explanations that raise themost serious difficulties for Hempel, as Chapters V and VI will attempt to show. Part of the confusion here results from Hempel's earlier concentration on the deductive model and his resulting lack of attention to the completeness conditions of model (P). It is this, I suspect, which leads him to contrast the two kinds of inference which occur in a partial explanation. For when juxtaposed in the same argument, the grounds for expect- ing S to be a member of F, which are conclusive, seem somehow to be better or more complete than the inconclusive but perhaps high- ly probable grounds for expecting S to fall within W, since the latter grounds do not strictly imply this explanandum. In fact, it is not at all clear why Hempel would characterize such explanations as "partial" unless he thereby meant to suggest that they do not fully or completely explain, according to the model (D). But since the reason for this claim, that the explanans does not deductively entail the explanandum, applies to all in- ductive arguments, it would naturally follow that model (F) is itself incomplete and not an ideal explication of a different kind of scientific explanation. At best, then, Hempel has marked a distinction in such functional explanations as Freud's between the conclusion that can be deductively inferred, "S is a member of F," and the one that can be only inductively inferred, "S falls in class W." But such a distinction is nevertheless of major 76 interest to the eXplainer. By using the terms "complete" and "partial" to so mark this distinction, however, he tends to con- fuse the issue of completeness, and with it his claim of the independent status of the two models and of the two kinds of ex- planation they represent. As a final case of deviation from the ideally complete pat- tern of explanation (leaving open for the moment whether this is (D) or (P) or both), there is what amounts to a lower limiting case in the continuum of approximations. Some ecplanatory accounts depart even further than elliptical or partial ones and in fact border closely on being untestable pseudo explanations. In an early essay Hempel labeled such accounts "explanation sketches" and described them as: a more or less vague indication of the laws and initial conditions considered as relevant, which needs 'fil- .ling out' in order to turn into a full-fledged explana- tion. This filling out requires further empirical £8- search for which the sketch suggests the direction. Consider two examples: first, we might explain that the Dust Bowl farmers migrated to California because continual drought and sand- storms rendered their existence extremely precarious, which seems to assume that populations tend to migrate toward regions with better living conditions. Secondly, a particular revolution might be explained by reference to the discontent of a large part of the population, together with certain prevailing conditions. While a general regularity is implicitly assumed as the connecting hé Hempel, "The Function of General Laws in History," p. 351. 7? link in this caSe also, it is most difficult to know to what ex- tent and what specific form the discontent has to assume, and what environmental conditions must be, to bring about a revolu— tion. Still, the sketch contains no empirically insignificant terms and does seem to offer direction for research into condi- tions which might tend to confirm or refute the more specific implicit statements of the explanans. Now we have so far considered, however briefly, all of the conditions of adequacy laid down in "Studies in the Logic of Explanation." (R3), that the explanans have empirical content, requires more consideration in this context, since the problems surrounding the distinction between pseudo explanations and genuine explanatory sketches turn on just this condition. For, while some might take the above description and examples of such a sketch as illustrating a pseudo explanation, others might charge that such cases are not merely sketches or incomplete in any way but rather are explanations of a different kind altogether and hence complete of their type. Such an issue turns largely on how rigorously the criterion of significance or empirical mean- ingfulness is employed, a problem which, as we noted earlier, is much too complex to be handled adequately in this paper. Nevertheless, independently of whatever criterion (or better, criteria) of significance turns out to be adequate, it remains the case that "an explanation sketch does not admit of an empirical test to the same extent as does a complete explanation; and yet there is a difference between a scientifically acceptable explanation 78 D7 sketch and a pseudo explanation (or a pseudo explanation sketch).’ The difference, simply put, is that pseudo eXplanations have no empirical content and thus are untestable in principle, thus vio- lating (R3). 80 wherever the line between empirical significance and non-significance may happen to be drawn, sketChes and pseudo explanations will occupy opposite sides. And this is because of the employment in the latter of empirically meaningless terms, which precludes even a rough indication of the kind of inquiry that might lead to evidence either confirming or disconfirming the purported explanation. Since Hempel's concern is not only to set a lower limit to the degrees of deviation from the ideally comphate model of ex- planation but also, by so doing, to distinguish legitimate or genuine explanations (in whatever degree) from pseudo ones, it might appear that his position diverges drastically from POpper's view, mentioned earlier, concerning historical interpretations or theories. You will recall that while POpper likewise contrasted such interpre- tations and scientific theories in so far as the former were un- testable, he nonetheless gave them an important role and status in historical and social inquiry as foci or working hypotheses for collecting additional information. In fact, their role and status was to be precisely'that.which Hempel assigns to explanation sketches. Yet Hempel considers them testable and hence scientifically acceptable, while Popper takes them to be untestable or unfalsifiable but still h7 Ibid. 79 scientifically useful and even appraisable in some sense. Part of this disagreement is merely apparent. And it is so largely be— cause of an obvious ambiguity concerning the terms "untestable" (or, to use Popper's terminology "unfalsifiable") and "scientific." Popper views such interpretations or sketches as empirically sig- nificant, but he disagrees with Hempel by giving them, for this reason, scientific status also. In other words, his notion of testa- bility is designed as a criterion of both empirical significance and scientific status. I am not suggesting that there are no important differences on this issue between Hempel and Popper, or that their disagreement is completely verbal, for indeed I think there are such differences};8 But for our present purposes, it suffices to show that they are agreed on taking sketches to be empirically testable, and hence satis- fying (R3), to be lower approximations to an ideal model of expla- nation, to be scientifically fruitful as heuristic hypotheses guid- ing their own development toward completeness or "filling in," and hence distinguishable from pseudo explanations. One point of im- portance emerging from the foregoing, then, is that neither Popper nor Hempel denies the existence of purported or potential explana- tions which do not conform completely to the requirements of the ideal models (D) and (P). Instead, it has been argued, their position regarding such purported explanations can be characterized as incom- plete approximations to the ideal models. h8 Cf. I. Scheffler's penetrating discussion of this issue in Anatomy pf Inquiry, pp. 137-50. H3 v‘\ ‘1- ‘u ‘~ ,M‘ «uh, BO If the above considerations are cogent, we can take the CL theory of explanation to consist in part of two ideal models of completeness and various degrees of approximation to these models or degrees of incompleteness. The notion of completeness however is still not as clear as might be. There seem to be various mean- ings or senses of the term as used in recent discussions, four of which I want now to examine. Let us refer to them as deductive, concrete, factual and descriptive completeness. One such usage, the deductive completeness alluded to earlier in our comparison of models (D) and (P), derives from Hempel's earliest essay on explanation, and has subsequently been the source of much debate and criticism, most recently by Dray and Scriven. In this sense, we have a complete explanation only when our explanans contains strictly universal laws and the explanandum is logically entailed by the explanans; in other words, only when we have a deductive-nomological explanation of the form (D). And since probabilistic-nomological explanations of the form (P) in- variably contain statistical or probabilistic premises in their explanans, which thus implies its explanandum not with deductive necessity but only with more or less high probability, they must be intrinsically incomplete in this sense. If this is the intended usage, then clearly for Hempel there would be only one ideal model of explanation, not two independent ones. In his most recent dis- cussion of this meaning}9 Hempel does little more than acknowledge h9 Hempel, "Reasons and Covering Laws in Historical Explana- tion," pp. 151-2. 81 its existence, not at all clarifying his own view as to whether or not it is a sense of completeness he intends or even finds acceptable. Three other uses of the term remain to be considered, each of which involves ontological questions as to just what it is we ex- plain, whether partially or fully. It should be noted that the first sense of com lateness just mentioned, that of deductive com- pleteness, applies only to the explanation of aspects of events described by statements not to the explanation of concrete events themselves. This point emerges more clearly perhaps by a consid- eration of the second sense in which completeness has been taken, what we shall call concrete completeness. Often 'complete expla- nation' has been used to mean something like Weber's "grasping the unique individuality of concrete events in their infinite variety or fullness." This usage is associated mostly, I suppose, with an idealist defense of the autonomy of history or the Geisteswissen- schaften generally. And for this reason both Popper and Hempel have, on various occasions, registered their objections to it. In this context, an individual event is typically characterized, as Hempel says, "by an individual name or by a definite description, such as 'the Children's Crusade,‘ 'the October Revolution,‘ 'the eruption of Mt. Vesuvius in 79 A.D.' ... and the like." The objection to this usage, 1.3. to talk of a complete explanation of such concrete events, is based on the grounds that Individual occurrences thus understood cannot be explained by covering laws nor in any other way; indeed it is un- clear what could be meant by explaining such an event. For any event thus understood has infinitely many aspects and thus cannot be even fully described, let alone explained .... 82 Evidently, a complete characterization, let alone explanation, of an individual event in this sense is impossible. Thus, Hempel wisely preempts the counter-charge that his own position implies a mechanistic view of man, society or historical processes, and employs "robustly materialistic language." More— over, another objection to the position of eXplanation as a rela- tionship between explanatory premises and concrete events, noted by Scheffler,Sl exhibits the thrust of Hempel's point. The force of the criticism turns on a consideration similar to that which produces the inconsistency or ambiguity of inductive inference. For if we talk of explaining the event b by providing appropriate statements A and L, having b's description, B, as a logical con- sequence (much as Popper52 does on at least one occassion), we are led to a contradiction. Suppose, e.g. we have a particular spatio-temporal chunk, k, described as blue by the statement, (I) Bk, and as hot by the statement (2) Hk. And suppose additionally we have the following set of premises: (3) (X) (y) (Wx . ny; Hy) (b) Wj (S) Rik where 'j' represents a spatio-temporal chunk or slice, 'W' a predi- cate applicable so such chunks, 'R' a relational predicate applicable 50 Ibid., p. 150. 51 I. Scheffler, Anatomy 2f Inquiry, pp. 58-9. 52 K. Popper, The Open Society and Its Enemies, Vol. II, p. 362, note #7. 83 to certain pairs of them, and (3) is assumed to be lawlike. If we further assume that (3), (h) and (5) are true, the deductive pattern will clearly apply to (2) as an explanandum. Roughly, then, we can say that (3) - (S) constitutes an ex- planation of k, or in Scheffler's symbols: (6)’ E(3) - (5), k'(to be read: (3) - (5) explains k). Yet, (3) - (5) just as surely does not explain k, since not yielding (l) as a logical consequence and hence not explaining the event described by it, 31E. : (7)' E (3) - (S), k' (to be read: (3) - (5) does not explain k). Accord- ingly, we end with a contradiction regarding (6) and (7), since (3) - (5) both explain and do not explain k. Furthermore, the reason appears to be that concrete events or spatio-temporal chunks are describable in alternative, logically independent ways, hence making it false to parade any given description of such a concrete individual as its unique description. As in the above example, one of these alternative descriptions may well be implied by a given explanans while others may not. It is perhaps clearer now why Hempel insists that deductive completeness applies only to aspects of events as described by statements and not to concrete individual events characterized by a definite description or an individual name. The above considera- tions, moreover, point to a way of avoiding the contradiction and, as a result, to a third or factual sense of completeness. Instead of depicting scientific explanation as a relationship between ex- planans and concrete events, Hempel opts for a relationship be- tween explanans sentences and “aspects 2f, 2; facts about, concrete \ |(L 8h : events,”3 3.3. facts about events or chunks—as-qualified-in-certain— ways. In our example, we should take (3) - (S) as explaining k-as- described-by—(2), not as explaining k. Consequently, we are a long way from any sort of "robust materialism." What is explained on this is not a concrete event or spatio—temporal individual chunk, but something of another sort associated with it, of which there are as many as there are logically independent descriptions of the chunk. These new entities (let us call them hereafter 'facts') are not themselves spatio—temporal entities: they are neither dated nor bounded. Nor are they identified with the descriptions themselves. They are abstract ('logi— cally intensional') entities, intermediate between chunk and descriptions, each such entity corresponding to some class of logically equivalent (true) descrijtions uniquely. 5h By the introduction of facts, then, Hempel apparently resolves the above—mentioned problem, since in place of the one entity or chunk, he now embraces two entities of a different kind associated with it: facts. In short, the contradiction is avoided since what is explained and also not explained by (3) - (5) is no longer the same thing, but two different things. The fact that k is hot is explained by (3) - (5), (8)' E (3) - (5), r(Hk)§ while it is instead the fact that k is blue which is not explained, (9)'~E (3) _ (5), f(Bk): Thus, this third sense of 'complete explanation,’ the only one in which Hempel countenances a covering law explanation of an in- dividual event, amounts to a complete explanation of a particular aspect or fact about a concrete event. S3 Hempel, "Reasons and Covering Laws in Historical Explanation," p. 1.50. 5h Scheffler, Anatomy pf Inquiry, pp. 59-60. .W- 4. :2 a' , / . If ’ . 1" ' i .o t l ' . ‘ i i I A 85 But Hempel also speaks of a fourth sense of 'complete ex- planation,‘ a sense in which one might speak of partial and more complete descriptions as well as explanations of concrete events. Instead of completely explaining one aspect or fact about a con- crete event, one set of explanations might explain more aspects or descriptions of it than does naother, and in this sense be said to be more complete than another set. A simple case would occur when the aspects explained in one set of explanations, 81’ each of which explains some description or aspect of a concrete event, forms a proper subset of those aspects explained in another set, S In this case S2 provides a more complete explanation of 2. the event than does 81' However, Hempel's appeal to an ontology of facts about or as- pects of concrete events, to replace particular individual events as the proper objects of explanations, itself needs defense against various charges. Two such objections will be considered briefly. One might charge, first, that Hempel's prOposal is liable to the same sort of difficulty as is the event-ontology, only on a new level. In other words, since according to a fact-ontology the same fact may still be associated with different descriptions, i.e. logically equivalent ones, it might be claimed that the same con- tradiction arises as to facts as arose for events. But the two cases are not parallel in this sense. The problem arose regarding events only because a set of explanatory premises implied one description of the event, Bk, but not another Hk, and hence both explained and did not explain k. In the case of facts, however, 86 since the different descriptions associated with them are logically equivalent, whatever explanans implies one such description must also imply the other as well, and hence cannot both explain and not explain the same fact. One might charge, nevertheless, that the successful avoidance of a contradiction such as that between (6) and (7) has been pur- chased at too high a price for anyone with a puritanical philosophi- cal conscience, 315' at the price of an alstract intensional onto- logy.55 The question thus arises as to whether or not we can con- sistently construe science as an abstractive and selective enter- prise with scientific statements explaining things without being driven to presuppose the abstractness of these things. Scheffler's most illumined discussion of this question indicates that it is in- deed possible to do so. His main strategy consists in rendering the entire analysis of scientific explanation explicitly concrete by assimilating event-explanation to the explanation of laws or generalizations in the sense of providing sentences as objects of explanation. That is, sentences are to be rendered, generally, as inscriptions or tokens (physical objects of certain shapes) instead of as abstract shapes or types. And explanations are to be ex- pressed in relational manner, but now relating sentences to other sentences, not to events or facts. While fulfillment of the CL theory requires connecting two sorts of sentence-strings, i.e. explanatia and explananda, this nominalistic interpretation omits 55 Ibid., pp. 61-76. 87 postulation of the intervening facts as objects of explanation and goes directly to the event-descriptions themselves. It also accounts for the importance of selectivity and abstractiveness, but without quantifying over, and hence without committment to, abstract enti- ties. 88 Th2 Covering Egg Account as 3 Theory'gf Explanation Now, for our purposes ther is no need to choose between these competing explications of what it is we explain scientifically, between Scheffler's inscriptional ontology and Hempel's intensional account. of more importance is the nature of the explicative or constructional method itself. So far in the process of analysing Hempel's version of the CL theory we have had brief occasion to defend his account against a common misunderstanding and ensuing objections. we will find need to extend this defense in later chapters when attention is turned to historical explanations. Our case was grounded, you will recall, on the interpretation of the CL theory as a philosophic explication or reconstruction of the important but inexact pre-analytic or ordi- nary notion of explanation, and of its systematic relation to such other notions as inference, empirical significance, laws and con- firmation. Since much was and will be made to turn on this method, it might be helpful to offer some additional comments of clarifi- cation concerning both it and its relation to the value-neutrality thesis, upheld jointly by Popper and Hempel. No doubt the main task of the CL theorists is to find an ade- quate definition of the concept of scientific explanation, one providing a basis for a theory of explanation. In this they follow the lead of Professor R. Carnap, who describes the philosophic task of explication as I'making more exact a vague or not quite exact concept used in everyday life or in an earlier stage of scientific or logical development, or rather of replacing it by 89 a newly constructed, more exact, concept."56 The defined or earlier term is referred to as the explicandum, and the term used for the proposed or defining concept as the explicatum. Some important cases where vague concepts have been explicated are Frege's theory of arithmetic based on the analysis of the number "two" as the class of all couples, Russell's analysis of definite descriptions as incomplete or syncategormatic expressions, Tarski's semantical version of "truth" and Carnap's proposal to analyze one sense of 'probability' along the lines of "degree of confirmation".57 Hence, the philosophic task is not merely to transcribe or duplicate the meaning of the explicandum, but to improve upon it by progressively "refining or supplementing its meaning,"58 in Quine's phrase. Nor is it a case of finding a synonymous expression for the explicandum, or even of exposing hidden meanings. In fact, Carnap suggests that the explicatum need not correspond very closely to the meaning of the explicandum at all. The philosoyher begins with some inadequately formulated concept, with a vague, ambiguous or incomplete explicandum which nevertheless serves some important functions. We have in ordinary or scientific discourse, in other words, 56 R. Carnap, Meaning and Necessity (Chicago: University of Chicago Press, l9h7), pp. 7— . 57 Cf. C. Hempel's discussion in Fundamentals pf Concept Formation in Empirical Science (Chicago: University of Chicago Press, 1952), Chapter I, especially p. 11. 8 S W. Quine, From 5 Lo ical Point of View (Cambridge, Mass.: Harvard University Press, 19E35, p. 23. 90 an expression or form of expression that is somehow troublesome. It behaves partly like a term but not enough so, or it is vague in ways that bother us, or it puts kinks in a theory or encourages one or another confusion. But it also serves certain purposes that are not to be abandoned. Then we find a way of accomp- lishing those same purposes through other channels, using other and less troublesomeggorms of expression. The old perplexities are resolved. It should be emphasized that "tightening up," i.e. making more exact and precise, the explicandum is not a sufficient con- dition of adequacy for the explicatum. Maximum precision is clearly not the only goal of an explication. But recognition of this fact has led some to suppose that the explicatum must also satisfy our intuitions in the matter. For, then, fifa proposed analysis such as Hempel's violates "sound common usage" or is counter-intuitive, it is abandoned as inadequate. Even Hempel, you will recall, invoked such a criterion in originally discarding (Rh') or well-confirmedness for (Rb) or truth. However, we are not told by those who invoke such a criterion what conforming to an intuition means, nor what constitutes a justifiable intuition, nor even why intuition must be a deciding factor at all. As a result, most constructionalists follow the view that the explicatum be usable in place of the original vague explicandum. As the above passage from Quins suggests, we fix on the purposes served by the unclear expression which make it worth bothering about, and then devise more efficient ways of achieving these same goals of functions. In place of a criterion of intuitive- ness or maximum precision we have one of efficiency, or significant, 59 w. Quine, Word fl Object (N.Y.: Wiley and Sons, 1960) p. 260. 91 relevant or usable precision, based on the parallelism of function between explicandum and explicatum. Yet it might be objected to such an enterprise, as many who invoke the criterion of counter-intuitiveness have done, that philOSOphic reconstructions such as Hempel's tend to neglect the immense richness and complexity of ordinary discourse and hence force it onto a Procrustean bed of "neat simplicities." But the other edge of this sword is, I think, even sharper. For, as Feigl and Maxwell have recently argued,60 this immense richness of ordi- nary language often turns out to be an embarrassment of riches, and hence requires selection, abstraction and systematization for philOSOphic as Well as scientific purposes. They suggest, in par- ticular, three reasons for the need to reform or explicate ordi- nary language: in order to analyse at all, since most interesting terms of ordinary language are systematically ambiguous; to abstract our invariants of such usage and to systematize general principles in order to eliminate irrelevancies, and to arrive at viable approx- imations; and finally to correct the implicit rules of ordinary language which reflect false beliefs.61 So, while ordinary langua e is indeed the "first word" and the groundwork of philosophic inquiry, to which our explications must in some sense correspond, it remains only something to be respected not to be bound by. The main correspondence of explicatum 60 G. Maxwell and H. Feigl, "Why Ordinary Language NBUdS Re- forming," The Journal pf PhiIOSOphy, 58 (August, 1961), p. h92. 61 Ibid., p. h96. 92 to explicandum is one of function or purpose, not necessarily of Synonymy, logical equivalence, extensional identity, nor even structural isomorphism. Even those who opt for the latter relation- ship recognize, with N. Goodman, that the opposition to construction— alism which greets, say, Hempel‘s proposed definition of a scienti- fic explanation with the protest that explanation is "Not Merely" such but "Something More" fails to grasp what Hempel is doing. For in defining 'scientific explanation' along the lines of the CL theory, "he is not declaring that a so-and-so is nothing but a such-and-such," and the "' =df' in a constructional definition is not to be read 'is nothing more than' but rather in some such fashion as 'is here to be mapped as,'"62 with the mapping to be appraised on grounds of efficiency in fulfilling specified pur- poses or goals. Now, if our account of the explicative or constructional en- terprise is adequate, two important consequences seem to follow, not both of which have been clearly recognized or acknowledged by constructionalist philosophers. First of all, since the adequacy of any explication depends on the efficiency with which the expli- catum fulfills the purposes not so efficiently served by the ex- plicandum and since the determination of purposes and efficiency of concepts is an empirical matter, it follows that the philosopher is not exempt from the controls of empirical science. He is not engaged in a totally different enterprise than the scientist, a 62 N. Goodman, "The Significance of Der Logische Aufbau Der Welt," in P. Schilpp (ed.), ills Philosophy 3; Rudolph Carnap (La Salle, Illinois: Open Court, 1963), p. 53h. 93 conceptual analysis of concepts, of eliciting meanings and estab- lishing conventions, as Ayer e.g. has suggested.63 Philosophic theses or definitions, in other words, are not merely analytic statements as opposed to the synthetic or empirical statements of the scientist. Though definitions and hence neither true nor false, they are subject to empirical criteria of adequacy connected with establishing purposes and efficiency. In this respect it is puzzling to find Popper disagreeing with Carnap's views on explication on the grounds that one cannot "speak about exactness, except in a relative sense of exactness sufficient £23 a particular 51232 purpose -- the purpose of solving a certain given problem."6)‘l For Carnap has consistently advocated just such a position. This is especially clear in his distinction between internal questions which occur within a linguis- tic framework and external questions about the acceptability of the framework or categorial principles governing it. While existential statements made within the system are analytic for Carnap, statements about the framework or system are construed pragmatically as practi- cal matters requiring decisions. That is, to offer a philosophic explication is much like constructing a linguistic framework or theory, and the question of its adequacy is largely a decisional or pragmatic matter of how efficiently the explicatum resolves the problems for which it was constructed. 63 A. J. Ayer, Language, Truth 32d Logic (N.Y.: Dover Publi- cations, l9h6), Chapter II. 6h K. Popper, "The Demarcation Between Science and Metaphysics," in The Philosophy 2f Rudolph Carnap, p. 216, footnote. 914 Thus, though the decision to accept the explication or an appropriate framework is not purely a factual or theoretical matter, it will indeed "be influenced by theoretical knowledge."65 Moreover, The acceptance or rejection of abstract linguistic forms, just as ... in any branch of science, will finally be de- cided by their efficiency as instruments, the ratio of the results aggieved to the amount and complexity of the efforts required. Hence, Popper's disagreement with Carnap and Hempel on this point seems based on a misunderstanding of their position. But once this agreement is seen, we are led to the second con- sequence of our account of constructionism, one that pertains in- directly to the value-neutrality thesis. Once it is conceded that our explications are to be appraised pragmatically, it becomes apparent that in most, if not all, contexts we will have a variety of purposes to be satisfied by any given explicatum, interpreted now as an instrument. This requires that some of the various purposes or goals be weighted as more important than others. Hence the decision to accept or reject a proposed explicatum will require the making of value judgments. This consequence applies of course to the philosophic Level of external questions about a linguistic framework. So far, at any rate, we have not argued for its application to scientific questions, assertions or explanatory hypotheses. But if such a case can be made, then the value-neutrality thesis of weber, Popper 65 R. Carnap, 'Empiricism, Semantics and Ontology,“ added as a supplement to Meaning and Necessity, p. 208. 66 Ibid., p. 221. 95 and Hempel will be unacceptable. This, as mentioned earlier, will be the main burden of our final chapter. Our more immediate task, however, is to see how far the CL theory of explanation can be extended, to examine whether the Hempelian account can serve as an adequate explication of historical, as well as of scientific explanations . CHAPTER III RECONSTRUCTIONS OF THE SU fldESIS Idealism and the §H Thesis In the present Chapter we will begin to consider how far the CL theory of explanation can be fruitfully extended to historical ex- planations. In particular, we examine two suggested reconstructions of the SU thesis. First, it will be shown that the idealist emphasis on the subjective element of intuition or an immediate grasp of his- torical reality is an inadequate reconstruction, one successfully warded off by CL flaeorists. Secondly, a more recent non-naturalis- tic reconstruction will be investigated and seen to offer important suggestions concerning both the VN and SU theses. Various arguments have been offered to support SU, to show that any extension of the CL theory to historical inquiries is impossible since history is in some sense autonomous. And arguments drawn from the historian's Special subject matter represent some of the most in- fluential and suggestive of recent attempts to prove history autonomous. Our main concern in this and the remaining chapters will be limited, how- ever, to a narrow range of cases, gig. to explanations historians give of purposive human actions considered important enough to be mentioned in historical narrative. Usually such explanations offer reasons why some individual person decided to act in a Specified manner under given cir- cumstances. The peculiarities of such historical actions have led many recent critics of the CL theory to rehabilitate Weber's SU thesis in the form of a traditional doctrine of idealist philosophers of 96 97 history: that the objects of historical inquiry, being these human actions, differ fundamentally from those of the sciences. This difference is then parlayed into the charge ”that the explanation of individual human behavior as it is usually given in history has features which make the CL model peculiarly inept.“l For, even if the CL model applies to natural events, by explaining them as sub- sumption under empirical laws, it would still be inapplicable in history because of the latter's peculiar subject matter. In other words, this difference and peculiarity, stimulating much sympathy with the idealist position, is used to defend the distinction not merely between the different sources and kinds of empirical laws but between different types of explanation. Historical explanation, it is claimed, requires a different kind of understanding and has a different kind of "logic' than does scientific understanding. Hence, to accept the CL theory as applicable to history would be to conflate the distinction between explanation types. Aristotle's comment in the Poetics, that poetry is of graver philosophic import than history, serves to introduce some of the issues between covering law theorists on the one hand and both past and recent critics on the other. The distinction between historian and poet...consists in this, that the one describes the thing that has been, and the other a kind of thing that might be. Hence poetry is something more philosophic and of graver import than history, since its statements are of the nature rather of universals, whereas those of history are singulars. By a universal statement I mean one as to what such and such William Dray, Laws and Ex lanation in Histogy (London: Oxford University Press, 193%”, p. 113. 98 a kind of man will probably or necessarily say or do..., by a singular statement2 one as to what, say, Alcibiades did or had done to him. Aristotle's view--taken seriously by such later idealist philo- sophers as Windelband, Collingwood, Oakeshott, Dilthey and Croce, and by historians as diverse as Butterfield, Beard, Trevor-Roper and Trevelyan-- led to'Windelbandfis widely accepted distinction between two different kinds of sciences. The nomothetic or generalizing natural sciences attempt to establish abstract general laws concerning pervasive, universal and indefinitely repeatable events. Ideggraphic or historical sciences, on the other hand, seek to understand what is special, singular, unique and nonrecurrent. Many covering law theorists concur in the marking of such a dis- tinction. POpper, you will recall, distinguished between the theo- retical and historical sciences on the ground that the former seek to establish general laws, while the latter assume these laws in order to establish warranted singular statements. Idealists, however, use Windelband's distinction to support two further claims: that the logical structure of explanation dif- fers essentially in the two kinds of sciences, and that historical explanations are sui generis. Consequently, they argue for a methodological disunity in the empirical sciences, since the his- torian, to explain his subject matter quite satisfactorily in his own way, need not appeal to general laws. A glimpse of the issue can be seen in Collingwood's declaration that history is not a 2 Aristotle, The Basic WOrks of.Aristotle, Richard McKeon (ed.) (New York: Random House, 19m), 1136 1-11. 99 spectacle. The scientist, subsuming events or actions under laws, patterns or regularities, remains essentially a spectator. The historian, on the other hand, adopts the standpoint of the agent, viewing events or actions from the "inside," not just externally or from the “outside." Hence he explains actions by appreciating the agent's problems, goals and beliefs, and by appraising the agent's responses to his problems. The central contention of The Idea 23 History is that history is an autonomous discipline with its own concepts and methods and with a unique kind of understand- ing. Understanding the thoughts of historical agents constitutes the primary task of the historian. By the 'outside' of an event Collingwood means everything be- longing to it which is describable in mechanistic terms of bodies and motion. By the ”inside” of the event he means I'that in it which can only be described in terms of thought: ‘Caesar's defiance of Republican law, or the clash of constitutional policy between himself and his assassins."3 The historian, accordingly, investi- gates not mere events, having only an "outside,' but actions, con- sisting of the unity of an event's "outside“ and I'inside." His main task is to think himself into the historical action, to discern the thought of the agent expressed in the event, and thus to achieve historical, as contrasted with scientific, understanding or in- telligibility. The scientist also goes beyond the events he encounters in 3 R. c. Collingwood, 313 Idea 2;; History (New York: Oxford University Press, 19h6), p. 213. ‘v..- .3 .‘su-L’) “N‘fl _‘.,,:—..:¢:_:r, ‘Itr-u‘“ __' ~H“ ‘- '-'_.;‘t.- r . ., . '-b;M-fl"u't§' _. .ar ' h 1‘ 4 1'3: _'__ ... . _ . .. ' - . O . C U I ‘1 - . . a . . u lOO inquiry, but only by relating events to others andthus subsuming them under general formulae or laws of nature. 30 Colingwood con- cedes the CL position when applied to science, since nature is al- ways and merely a phenomenon or spectacle presented to the inquirer. In history, however, events are never mere phenomena, mere spectacles for contemplation, They are what the historian looks past or through to penetrate the thought or idea within; they are purposive calcu- lated human actions. When one discovers the thought expressed in an event, one already understands the event. To know what happened is already to know why it happened. Such understanding results not from "merely" subsuming the event under laws, but is discerned instead by "re-thinking" the thoughts, by "re-enacting" the past, in one's own mind. By denying that human actions consist only of physical move— ments from which we can infer the motive or reason behind them, Collingwood concludes that thoughts must be known directly or im- mediately by a special kind of non-discursive or intuitive knowledge. All history, consequently, is the history of thought, of the plan or idea of human actions. In another idiom, historians explain in the way that art, not science, explains: by illumination instead of deductive inference, by revealing the universal in the particu- lar. Hence history cannot possibly be causal explanation or the science of human behavior. In opposing these further claims, CL theorists also try to eliminate the deficiency which Aristotle attributes to history in contrast to poetry. The ideal model they establish for historical 101 explanations of individual actions requires that the functions of history and poetry, viewed as distinct by Aristotle, be united. The historian must not merely describe the particular, whether from ”inside“ or "outside.“ He must additionally reveal the uni- versal which it embodies, but by subsuming it under general laws. Hence, in Aristotle's phrase, he raises history to the level of art and knowledge.h Accordingly, file CL position, along with that of their idealist critics, implies the denial of Aristotle's dictum that history has less philosophic import than poetry. The issue between them turns, instead, on how the universal is revealed in the particular: by subsumption under general laws or by an imagi- native, intuitive ‘re-interpretation‘ of the total context of the event. Further clarification of the issue arises from posing for the Idealist the question, 'What is the explanatory force, nature or logical structure of these allegedly distinctive historical expla- nations?“ In considering this question we hope to elicit and de- velop the more recent criticisms of the CL theory of explanation. But since the import of fliis criticism is best exhibited as an outgrowth of the earlier dispute, we will first examine briefly the idealists' reconstruction of the SU thesis. Then, having con- sidered the "official" CL rejoinder to it, we shall note how some recent philosophers rehabilitate and revive the older position. h E. Barker, IRational Explanations in History,I in 3. Hook (ed.), Philosophy and Histogy (New York: New York University Press, 1963) , In 179- Perhaps the main claim of the earlier idealist philosophers of history, already noted, is that the subject matter of historical inquiry differs fundamentally from that of the natural sciences, since concerning the thought and actions of humans. As a result, explanation by subsumption under empirical laws is considered singularly inappropriate in history. For even if individual human actions could be subsumed under law, this would not constitute understanding of these actions in a sense proper to the special subject matter. The additional intuitive factor required to achieve historical explanation proper, "empathetic understanding" or Verstehen, is usually contrasted with the allegedly superfi- cial knowledge gained through tests and statistics. With men con- sidered the ultimate unit of historical and social life, and the mind of man construed as a given immediate reality, understanding of the social world is founded on one's personal experience. One understands the experiences of other persons, especially in an historical context, only through "re-experiencing" or "re-living" these experiences. Butterfield, an historian exhibiting the influence of the idealists, summarizes much of their case in the following passage: Our traditional historical writing... has refused to be satisfied with any merely causal or stand-offish attitude towards the personalities of the past. It does not treat them as mere things, or just measure such features of them as the scientist might measure; and it does not content itself with merely reporting about them in the way an external observer would do. It insists that the story cannot be told correctly unless we see the personalities from the inside, feel- ing with them as an actor might feel the part he is playing-- thinking their thoughts over again and sitting h 103 in the position not of the observer but of the deer of the action.... the historian must put himself in the place of the historical personage, must feel his pre- dicament, must think as though he were that man.... Traditional historical writing emphasizes the impor- tance of sympathetic imagination for the purpose of getting inside human beings. We may even say that this is part of the science of history for it produces com- municable results-- the insight of one historian may be ratified by scholars in general, who then ive currency to the interpretation that is produced..../ But Max Weber offers perhaps the most influential account of this procedure in his postulate of subjective interpretation, which stresses the primacy of consciousness and subjective meaning in interpretations of social actions. The historian's primary task, claims Weber, is to attempt the interpretative understanding of social action in order thereby to arrive at a causal explanation of its course and effects. In 'action' is included all human behavior when and in so far as the agting individual attaches a subjective meaning to it. ‘Weber introduces the notion of "ideal types" as a device to explain concrete historical phenomena, such as the development of modern capitatism, in their uniqueness. Such understanding, con- cerning the individuality of a phenomenon, "is not a question of laws but of concrete causal relationships. It is not a question of the subsumption of the event under some general rubric as a representative case but of its imputation as a consequence of some S H. Butterfield, History and Human Relations (London: Oxford University Press, l9Sl). pp. 1&546. 6 M. Weber, Theory 9f Social and Economic Organization (New York: Oxford University Press, l9h77, p. 88. 10h constellation."7 To afford acceptable historical explanations, such causal relationships must be meaningful as well as "causally ade- quate" and "objectively possible." They must, in other words, be based upon those aspects of human behavior containing cultural sig- nificance, valuation or other motivating factors. These causal connections are expressed in terms of principles, classified as "general empirical rules," which convey knowledge derived from our personal experience. Weber also introduces, as a means of discov- ering such meaningful explanatory principles, the method of Verstehen or empathic understanding. The distinctive aim of the historian and social scientist thus appears as "understanding" social phenomena by using "meaning- ful" categories and imputing "subjective" states to human agents participating in social processes. This requires understanding the meaning an act has for the actor himself, not for the external ob- server of his actions, i.§. Verstehen. One seeks in such under- standing not a set of universal laws but the total intentional frame- work of the actor which clarifies the meaning of his specific act. Instead of subsuming the specific act under some set of covering laws, one refers it back to its intentional matrix which, as the ground of its meaning, helps to interpret it. Thus for a science which is concerned with the subjec- tive meaning of action, explanation requires a grasp of the complex of meaning in which an actual course of 7 M. Weber Methodolo Q; 222 Social Sciences (Glencoe: Free Press, l9h93, p. 78. ‘30:.- lOS understandable action thus interpreted belongs. In all such cases, even where the processes are largely effectual, the subjective meaning of the action, includ— ing that also of the relevant meaging complexes, will be called the 'intended' meaning. Weber thus resorts to two separate spheres of scientific cognition: we explain natural events and we understand human actions. We approach the former from without and the latter from within. Only as humans are we in a position to comprehend the subjective meaning an actor attaches to or intends by his action, and thus to formulate the general principles for understanding human actions. Fbr example, the conduct of a man about to be cut by a knife will surely be different depending on whether the knife-wielder intends a surgical incision or a mutilation. However, idealists modify SU so that interpretative understand- ing seeks the meaning of action in empathic intuition of a whole, of the realm of subjectivity which cannot be conceptualized but re- quires to be re—experienced or reproduced as a whole. "Conception is reasoning; understanding is beholding."9 Unlike Weber, they take historical thinking to be intuitive, not discursive. Howard Becker interprets Weber's SU thesis in the following way. Here, reduced to its barest, most obvious terms, is what is meant by interpretation, no more and no less. The interpre- ter puts himself in the place of the actor as best he can, and the degree to which he views the situation as the actor views it determines hifosuccess in predicting the further stages of the conduct. Weber, Theory pf Social and Economic Organization, p. 95. « L. von Mises, Epistemological Problems 2f Economics (New York: Van Nostrand, 1960), p. 13h. 10 H. Becker, Through Values 29 Social Interpretation (North Carolina: Duke University Press, 1950), p. 191. 106 And H. Cooley, the American sociologist, concurs. We understand human behavior by sharing their "state of mind" or intended mean- ings as a special kind of knowledge distinct from statistical knowledge which, without Verstehen, is superficial and unintel- ligent. The main point of the method of Verstehen common to the views of Weber and idealists, then, is that human actions are informed by meanings and purposes in a way in which only meaningful be- havior can be. Hence, to treat them as mere physical events in a causal nexus would prevent the observer from apprehending the action as historical or social, irreducible to non-social elements. More is involved than just the individual action in a behavioristic sense. One must also be familiar with the total social context of the action in order for it to be intelligible or understood, and no amount of general laws produces this understanding. At best they produce a causal explanation, but not understanding. But while Weber nowhere indicates unambiguously just what kind of knowledge empathic understanding provides the ide lists clearly opt for a non-discursive form of knowledge and hence sharpen the distinction between the nomothetic and ideographic sciences. 107 223 Standard Covering £3! Answer Unfortunately, however, neither Butterfield's summary nor the varied accounts of philosophers, social scientists or his- torians, including weber, succeed in clarifying either the nature of this empathic method of understanding, the logical structure of its corresponding kind of explanation, or what is peculiar about it. we are given no clear or unambiguous analysis of what the method of Verstehen amounts to in practice, nor what import to attach to the results of the method. Consequently, Theodore Abel, in a definitive essay drawing heavily upon Weber's own position as well as Hempel's essays, set himself the task of illustrating, analysing and evaluating "The Operation Called 'Verstehen'." His analysis reveals two particulars which are characteristic of the act of Verstehen. One is the 'internalizing' of observed fac- tors in a given situation [the stimulus and the response} the other is the application of a behavior maxim which makes the connection between these factors relevant. Thus we 'understand' a given human action if we can apply to it a generalization based upon personal exper- ience. We can apply such a rule of behavior if w are able to 'internalize' the facts of the situation. Abel illustrates the act of Verstehen as an explanatory tool in the following way. Although competent statistical research established a high correlation between the annual rate of crop production and the rate of marriage among farmers in a given year, we often feel we can forego statistical tests of such correlations 11 T. Abel, "The Operation Called 'Verstehen'," in E. H. Madden (ed.), Structure E; Scientific Thought (Boston: Houghton Mifflin, 1960), p. 161. 108 because we ”see” from our own experience of motivational patterns the connection as 'relevant‘. We I'understand ‘ why the rate of marriage in farming areas follows the rate of crop production. Our information is that failure of crops (C) materially lowers farmers' income (6'), and that one makes new commitments (E') when one marries (E). Then we internalize (c) into a 'feeling of anxiety' (Vc), and (E) into "fear of new commitments" (Ve) because of our own past personal experience. This gives us direct know- ledge or understanding of the bbviously" relevant connection or behavior maxim, “People who experience anxiety will fear new commitments" (Vc-Ve). The operation of Verstehen is thus based, according to Abel, on the application of personal experience to observed behavior. The explanation thus takes the following form: (a) The failure of crops (C) produces feeling of anxiety and insecurity (Vc). (b) we understand from our own personal experience that (Vc) leads us to fear new commitments (Vs). (c) (Ve) leads to low marriage rates (E). (d) Therefore, we can understand why (E) occurs under conditions (C). Another classic example of ”meaningful" explanation of a social phenomena is Max Heber's account of modern capitalism as due in part to the ascetic forms of “The Protestant Ethic.‘ Here the former (E) occurs under the complex conditions of the latter (C). But individuals participating in (E) are assumed to be subjectively committed to certain values (Ve), and those partici- pating in (C) to be committed to other values (Vc). And since 109 (Vc) and (Ve) are ”meaningfully" related, due to the motivational patterns we personally experience by Verstehen, we thus 'under- stand. why (E) occurs under conditions (C). The form of the explanatory argument follows: (a) Calvinistic Protestantism (C) developed individuals with subjective states, values or an attitude towards life called 'Protestant Asceticism' (Vc). (b) we understand from our own personal experience of motivational patterns that (Vc) is l'meaningfully" related to and leads to the attitude towards life called “The Spirit of Capitalism" (Ve). (c) Individuals with (Ve) were mainly responsible for the development of modern capitalism (E). (d) Therefore, we can understand why (E) developed under conditions (C) The main difference between such explanations and a CL one, then, seems to be that the antecedent conditions are connected with the explanandum-event not by a general empirical law but rather by a statement of what the historical agent valued or be- lieved just prior to his action. Such a statement of what was meaningful, significant or intended by the agent, the idealists claim, can only be known by imaginatively constructing or empa- thizing with his situation. However, from Abel's analysis of Verstehen, we see clearly the gross limitations of such an operation. At best it suggests plausible or possible explanations or hypotheses; it is a source of "hunches" or discovery. And Abel, Hempel, Popper and Nagel quickly offer the "official” CL answer to those advocating Verstehen as a peculiar kind of historical understanding. In 110 Hempel's terms: This method of empathy is, no doubt, frequently applied by laymen and by experts in history. But it does not in itself constitute an explanation; it rather is essen- tially a heuristic device; its function is to suggest certain psychological hypotheses which might serve as explanatory principles in the case under consideration...; but its use does not guarantee the soundness of the his- torical explanation to which it leads. The latter rather depends upon the factual correctness of the empirical generalizations which the method of understanding may have suggested. Nor is the use of this method indispensable for historical explanation. A historian may, for example, be incapable of feeling himself into the role of a paranoiac historic personality, and yet be able to explain certain of his actions;notably by reference to the principles of abnormal psychology. Thus whether the historian is or is not in a position to identify himself with his historical hero, is irrelevant for the correctness of his explanation; what counts, is the soundness of the general hypotheses involved, no matter whether they were suggeiged by empathy, or by a strictly behavioristic procedure. So, while Verstehen may serve an important heuristic role of suggesting or discovering explanatory hypotheses, it constitutes neither a necessary nor a sufficient condition for justifying or confirming them. Hence, it is not considered a serious competitor or alternative to the CL model of explanation. In fact, even Max weber, an advocate of non-intuitionist Verstehen, stresses the need to support aly given interpretation of subjective states with adequate observational verification. Otherwise: there is available only the dangerous and uncertain pro- cwweMtM'mgmuymWHMM'mRhmmmmin 12 c. G. Hempel, “General Laws in History," p. 1.67. 1 I 111 thinking away certain elements of a chain of motivation and working out the course of action which would thin probably ensue, thus arriving at a causal judgment. In "The Logic of the Cultural Sciences' he particularly warns against confusing the ”psychological course of the origin of sci- entific knowledge and 'artistic' form of representing what is known...with the logical structure of knowledge."1h He stresses no less, in his "judgments of objective possibility," the impor- tance of counter-factual conditionals to historical inquiry in general and to historical explanatory laws specifically. Popper further supports the CL answer by denying the unique- ness of the method of "intuitive understanding' to the social sciences or history. Even the physicist, though not helped by such direct observation, 1"often uses some kind of sympathetic imagination or intuition which may easily make him feel that he is intimately acquainted with even the 'inside of the atoms' -- with 15 even their whims and prejudices.” However, Popper also indicates our more direct knowledge of human actions than of physics. events. His major point, nevertheless, agrees with that of Hempel and Abel: any hypothesis resulting from intuitive understanding must be empirically testable to qualify as a genuine explanation. In other words, testability or confirmability (R3) is a necessary 13 M. weber, Theory 2f Social and Economic Organization, Po 9?. In M. weber, Methodology of the Social Sciences, p. 176. 15 K. Popper, Poverty gf Historicism, p. 138. 112 condition of an adequate explanation. And the method of Verstehen fails to provide an adequate method for testing or corroborating hypotheses. It is at best a method for discovering possible explanations, which must then be subjected to appropriate objective testing procedures. Nagel sums up this “official" CL or naturalist answer with three basic countercharges. He argues that there is not a differ- ent kind of knowledge involved in understanding social phenomena, that the method of Verstehen as an empathic response to or imagina- tive reconstruction of another person's motivation involves a fundamental subjectivism which renders it at best nonscientific, and finally that fine method offers, on its own, no criteria for testing scientific hypotheses regarding human actions. Some of the applied impact of this ”official" answer can be gleaned from a recent statement of a practicing sociologist, Professor D. Martindale. Speaking of the idealist use of ideal types, of configurations of 'meaningfulness" guiding the I're--living" of historical eXperience, Martindale claims that: such formulations have lost.their interest for modern students. Even the most tender minded of contemporary students is inclined to see science as all of one piece. The insights produced by intuition, empathy, or some method of verstehen are to the modern student mere untested hypotheses. The funeral oration of the verstehen point of view was gracefully and ceremoniopgly performed by Abel in his “Operation Called Verstehen." ' 16 D. Martindale, ”Sociological Theory and the Ideal Type,‘ in L. Gross (ed.), Symposium on Sociological Thought (New York: Columbia University Press, 1935), p. 82. 113 Thus, the "official" CL answer is often said to dismiss Verstehen as some sort of methodological dodge, as an inferior way of obtaining the same kind of explanation as can be obtained more reliably by direct subsumption of human actions under empirically testable and confirmed covering laws. 111; 332223 Replies This 'official" answer, nevertheless, fails to convince more recent non-naturalist observers of the issue. There is general agreement, I think, that this neo-Weberian CL answer exposes much mystery-mongering in the idealist position by emphasizing their cloudy mixture of psychological and methodological elements. Still, many feel that the CL argument does not cut as deeply as Abel, Hempel and Nagel assume. They believe that the CL theorists overlook flue important features about explanations of human actions in history, features stressed by Weber and the idealists. In other words, more recent philosophers feel that the idealists were after something significant, but couched it in misleading and vague terminology. As a result, they recognize the value of the CL answer as needed clarification, but then use this clarification to reconstruct the important residue of the Heberian position. This residue consists largely of the allegedly non-experimental elements in common sense, scientific and philosophic inquiry. With the general conclusion that the CL answer overlooks a signifi- cant element of the idealist case, the present auihor concurs. But, as I hope to show, the development of this element has not been convincing. It requires not affirming SU but denying VN. Such non-experimental elements often produce the anxiety of doubt in even the most thorough-going naturalists, who usually identify all knowledge with value-neutral scientific knowledge and opt for the universal applicability of the experimental method. J. H. Randall, for example, expresses just this anxiety in summing up a volume of essays by naturalists: 115 The idealists may have lacked scientific knowledge and tedhniques. But it is often hard not to feel that they have possessed most of the human wisdom...have the edge on insights, on the discrimination of values, on the appreciation of the richness and variety of the factors demanding organization.... Naturalistic philosophizing must become as rich as the idealistic philosophies by incorporating the facts and experiences thiy emphasized within its own more adequate framework.... Thelma Lavine, enlarging upon this anxiety of occasional naturalistic doubt, attempts to reconstruct the method of Verstehen along naturalistic lines, in order to include these important residual non—experimental elements in a wider natural- istic framework. "For naturalists do not so much seek to deny the fact of the various nonexperimental elements in inquiry as they fear the uncontrolled vagaries which are apt to result from acknowledging them. '18 The main problem of such a reconstruction, however, consists in properly locating such naturalistic safe- guards or controls for Verstehen. Miss Lavine suggests only a modification of the naturalistic emphasis on "a single intellectual method.“ Instead she stresses ‘a single intellectual criterion for whatever method may be feasible,"zig., the criterion of pertinent empirical checks or testability.19 But She nowhere makes clear just what the method of Verstehen contributes to the CL model of explanation or to the scientific 17 J. H. Randall, "Nature of Naturalism,” in Y. Krikorian, Naturalism and the Human Spirit (New York: Columbia University Press, 19HHT:—pp. 375-62 18 T. Lavine, "Note to Naturalists on the Human Spirit," in Natanson (ed.), Philosophy of the Social Sciences (New York: Random House, 1963), pp. 253:9. 19 Ibid., p. 259. 116 method of naturalism generally. Nor is it clear from her account what the method is or how far its scope extends. In fact, Nagel, in a direct response to her proposed reconstruction, resorts to his earlier distinction between the logic of discovery and the 10gic of validation in repeating the charge that even in its re- constructed form the method at best serves to generate suggestive hypotheses but does not suffice to verify or validate any. He thus remains constant in interpreting ‘subjective' in SU along the lines of private, personal and unverifiable judgment. In response to Nagel's criticism, Miss Lavine makes one telling point of major significance for our purposes. Yet it becomes pro- gressively confused when elaborated and hence is never successfully developed. Arguing that the scientific method of naturalism does not suffice as a general philosophic method, She contends that Undeniably, the principle of continuity of analysis does not bar the 'acceptance' of scientific conclusions. But the point I am making lies precisely here: what is en- tailed in the concept of 'acceptance of scientific con- clusions'? Further, what is the relationship between acceptance of scientific conclusions and the philosophy of naturalism? .... Surely in the most common usage of the term 'acceptance,‘ acceptance of scientific conclu- sions does not by itself entail any philosophicd_opera- tions whatsoegsr and is unworthy of being designated as naturalistic. So, though granting Nagel's claim that contemporary naturalists neither identify scientific method with overt experimental activity nor fail to recognize the importance of "non-eXperimental' elements, she argues that these elements remain residual and hence theoretically 20 Ibid., p. 267. 117 unexplored. Surely she is correct, at least in regard to the notion of "acceptance" of scientific hypotheses. Since I take this claim to be of primary importance, as the significant though confused insight of Verstehen theorists, it will be treated in some detail later. Let me comment here merely that some such naturalistic reconstruction of Verstehen is necessary to resolve the main issues debated by CL theorists and their recent ordinary- language opponents. Consequently, Miss Lavine's position deserves much more attention than Nagel's rather conventional reply. Having brought forth the important and difficult pragmatic concept of "acceptance" in this context, Miss Lavine proceeds to confuse the issue at hand by using it to defend the method of Verstehen as I'the sole method of philosophy.“21 The confusion in- volved in her naturalistic reconstruction of Verstehen is well marked by Natanson in a recent article in which he propounds an alternative non-naturalistic or phenomenological reconstruction. He complains that to provide ”naturalistic safeguards. for Verstehen, after placing the philosophic status of these very safeguards in question, is inconsistent and follows a step forward with a step back.22 This seems indeed to be the case. But instead of pursuing the purportedly non-experimental concept of 'accepting' scientific hypotheses, Natanson also moves in another direction. we will pursue this difficult and central concept further in 21 Ibid., p. 260. 22 M. Natanson, "A Study in Philosophy and the Social Sciences,” in ibid., p. 282. 118 Chapter 5. For the moment let us follow Natanson's argument. To remove the inconsistency, he denies the need to reinvoke naturalis- tic criteria as correctives for the method of Verstehen. Unlike Nagel, he complains not about macing Verstehen the essence of philosophical method, but only about her reverting to a Inotion of Verstehen in the narrow sense of method as a conceptual device." This he contrasts with the 'broad sense” which ''cannot be incorpo- rated into naturalistic methodology, because it is itself founda- tional.‘ Hence, Natanson's "way outI of the inconsistency is 'the transcension of naturalism in favor of a phenomenological stand- point... waich takes human consciousness and its intended meanings as the proper locus for the understanding of social action."23 Likewise, Natanson's reply to Nagel's objection, that Verstehen alone fails to provide any criteria for the validity of hypotheses about human actions, turns on the interpretation of Verstehen along quite different, more philosophic, lines. He denies that it was ever intended for such a purpose. Hence Nagel's objection, he claims, is simply misdirected. Verstehen, concerning only the 'conceptual framework within which social reality may be compre- bended“,2h is not intended to provide empirical criteria for deter- mining the validity of hypotheses. 23 Ibid., pp. 282-3. 2” Ibid., p. 281. 119 A. Schutz' Reconstruction 2f SQ Now, Natanson's identification of Verstehen with the "self- founding,” "presuppositionless" conceptual framework of phenomono- logy derives from another non-naturalist attempt to reconstruct the important residue of the idealist position, that of Alfred Schutz. Schutz' position bears investigation for a two-fold reason. On the one hand, along with Miss Lavine and Natanson, he finds the ”official” CL answer to the idealists based on a gross misunderstanding of Weber's postulate of subjective interpretation and of the SU thesis. His analysis, accordingly, serves to clari- fy further the misunderstanding and, atlthe same time, the nature of Verstehen. In fact, his reconstruction of Verstehen constitutes, in my opinion, one of the most complete and cogent accounts available; though it is not totally acceptable. On the other hand, his anay- sis also points directly to problems with the extension of ”testabil- ity" to apply to the purportedly non-experimental notions of goals, purposes and values. These are notions alluded to in the discussion of Levine's criticism of Nagel, notions to be developed later in connection with recent analytic-pragmatic criticisms of the CL theory of explanation. But while recognizing the importance of this point, he also fails to develop it as a serious challenge to the CL theory. In particular he fails to see how this tact requires denying VN. Sdhutz' defense of the non-naturalist Verstehen position is still particularly penetrating and enlightening. He disavows most of the previous vague and obtuse statements of the position in order to elicit the clear and important parts of it. Not what weber or 120 the idealists said but what they meant, or perhaps should have said, is his, as well as our, concern. He concedes at the start, for instance, that most “subjectivists” or non-naturalists had an erroneous view of the methods of natural science, usually depict- ing it in a most narrow and restricted manner. They were in- clined to generalize from the methodological situation in one particular domain, say history, to the situation of the social sciences generally. Instead, Schutz clearly opts for a unity of rules of scientific procedure, rules valid for all empirical sciences. In particular, he shares Weber's fear of private, un- controlled intuitions. The issue as he views it is not whether all empirical knowledge involves controlled inference, statability in propositional form, or observational verifiability. Nor does it concern the notion of "theory“, used to explain empirically as- certainable regularities, as applicable to history and the social sciences generally. On all of these points Schutz readily agrees with Nagel. Moreover, ...a method whidn would require that the individual scientific observer identify himself with the social agent observed in order to understand the motives of the latter, or a method which would refer the selection of the facts observed and their interpretation to the private value system of the particular observer, would merely lead to an uncontrollable private and subjective image in the mind of this particular studegg of human affairs, but never to a scientific theory. As a result of these disavowals, the important questions at issue are how to grasp subjective meanings scientifically, and how 25 A. Schutz, nConcept and Theory Formation in the Social Sciences," in ibid., p. 235. 121 to develop methodological devices for obtaining objective and verifiable explanations of a subjective meaning structure. For example, Schutz objects to the identification of experience with sensory observation and to the view of subjective uncontrollable and unverifiable introspection as the only alternative to con- trollable objective sensory observation. Unlike the idealists, he does not interpret Verstehen as providing a different kind of know- ledge than the natural sciences, nor as an unscientific empathic response to or imaginative reconstruction of another person's moti- vation in social action. Consequently, Schutz takes Nagel to be whipping a straw man, because of his failure to understand Max weber's SU thesis and his postulate of subjective interpretation. This seems dubious. But of more interest for our purposes is his reconstruction of Heber's postulate and of the Verstehen position generally, to which we now turn. Schutz' explication of Verstehen occurs in the context of de- fending three propositions: A) That the “primary goal of the social sciences is to ob- tain organized knowledge of social reality," 1.3. of "the sum total of objects and occurrences within the social cultural world as ex- perienced by the common sense thinking of men."26 It is not a pri- vate but an intersubjective world common to us all. B) That identifying experience with sensory observation and the experience of overt action l'excludes several dimensions of 26 Ibid., p. 236. 122 social reality from all possible inquiry,"27 3.3. such non- experimental elements as the observing scientist, the meanings of actions to performers, I'negative actions” and beliefs. C) That "all forms of naturalism and logical empiricism simply take for granted this social reality.... and assume, as it were, that the social scientist has already solved his fundamental prob- lem before scientific inquiry starts.“28 This point deserves special attention, for Schutz here, and elsewhere, elaborates on a point stressed earlier by Miss Lavine, viz. the residual and un- explored element of "acceptance" of scientific hypotheses. Schutz' criticizes a naturalistically oriented social science, one explaining human behavior in terms of controllable sensory db- servation, since it stops short before the description and explanation of the process by which scientist B controls and verifies the observational findings of scientist A and the con- clusions drawn by him. In order to do so, B has to know what A has observed, what the goal of his inquiry is, my he thought the observed fact worthy of being observed, i.e. relevant to the specific problem at hand, etc. This knowledge is commonly called under- standing. The explanation of how such a mutual under- standing of human beings might occur is apparently left to the social scientist.... This means ... that so-called protocol propositions about the physical world are of an entirely different kind than Bgotocol propositions about the psycho-physical world. All of these dimensions of social reality, he contends, re- quire not just the CL model of explanation but Verstehen as a 27 28 Ibid., p. 237. Ibid., p. 236. 29 Ibid., pp. 236-7. 123 process, technique or method of understanding. But while both critics and defenders of the process of Verstehen agree that it is ”subjective," they unfortunately use this term in different senses. CL theorists suggest that subjective understanding of huxnan motives depends on the private, uncontrollable, unverifiable intuitions of the observer. Social scientists such as Weber mean by "subjective" that the goal of Verstehen is to uncover the actor's intended meaning in his action, to determine what the agent meant instead of what the act "means" for an observer. Schutz accord- ingly sets himself the task of clarifying the meaning of Verstehen. F or , the failure to distinguish clearly between the various levels 0f Verstehen causes confusions in the CL answer to Weber's postu- late of subjective interpretation. The three different levels of application of Verstehen, accord- ing to Schutz, are: (1) “the experiential form of common-sense knowledge of hmnan affairs” whereby men in daily life do understand and inter- pret each other's actions by grasping the meanings, motives, atti- tHades and purposes intended by others. This he takes to be the Primary meaning of Verstehen. (2) "an epistemological problem“ of intersubjectivity, of hOw such understanding or Verstehen is possible, a problem that points to a clear distinction between the objects of knowledge of the natural and social sciences, but also one taken for granted in Our common-sense thinking. In this sense Verstehen is a metaphilo— sophical or categorial analysis of philosophical procedures, and 12h a conceptual framework for comprehending social reality. (3) "a method peculiar to the social sciences"30 whereby the concern is with second-order constructs or typifications of inter- pretation found in common-sense. Verstehen in this sense involves a theoretical system suitable for the clarification of the inter-. pretative understanding of the ordinary man in daily life. Let us look at these three levels of Verstehen in more detail. Let us consider them, however, not merely as a reconstruction of Max weber's postulate of subjective interpretation, but mainly as possible support for SU or as possible objections to the CL model of explanation and a naturalistic attempt to account for historical human actions. Following the neo-Kantianism of weber and Georg Simmel, Schutz stresses the importance of theory-laden facts. All facts are inter- preted facts. They are always selected by an activity of mind in accord with our purposes and interests. Hence there are, strictly speaking, no pure, simple or given facts. All knowledge involves a set of constructs, abstractions, generalizations or idealiza- tions. But as Weber, Popper and Hempel point out, this does not imply our inability to grasp the reality of the world, those relevant to our purposes and interests. Nevertheless, the thesis of "theory-laden facts" indicates a crucial difference between the constructs used by natural scientists 30 Ibid., p. 2&0. 125 and those used by historians (or social scientists generally). Since relevance is not inherent in nature itself but results from the interpretative and selective activity of man observing nature, the data or events explained by natural scientists are merely those within his observational field. They do not I'mean" anything to the atoms or molecules. They have no meaning or unity, in Simmel's phrase, prior to that given them by the inquiring scientist. I'The unity of nature emerges in the observing subject exclusively.“31 The subject matter for the historian or social scientist, on the other hand, consists of events and data of quite a different sort. The social world is not essentially structureless or “meaning— less" prior to the inquirer's observations. It already contains a unity, meaning and relevance for the human beings acting and thinking therein. For they have already preselected and preinterpreted this world by their common-sense constructs and idealizations. These very constructs or thought objects help them “get on' in their environment by determining their behavior, goals and means. As a result, the con- structs used by the historian and social scientist are not first-order constructs about uninterpreted data, as are the natural scientist's, but are instead second-order constructs. They are ”constructs of the second degree, viz. constructs of the constructs made by the actors on the social scene, whose behavior the scientist observes and tries to explain..."32 31 32 A. Schutz, "Common-Sense and Scientific Interpretation of Human Action," in ibid., p. 305. G. Simmell, "How Is Society Possible2", in ibid., p. 7h. .- V ’ “I ~ I l | =-..... _ ‘ - ----- '1 are." 126 But in order to explain social realityvv-in the sense of human conduct and its common-sense interpretations and systems of pro- jects, motives, relevances and constructs---the historian's second- order constructs must include reference to the subjective meaning which actions have for the actors, to the purposive behavior of historical actors. This is what Max weber intended, Schutz claims, by his postulate of subjective interpretation, which must be understood in the sense that all scientific explana- tions of the social world can, and for some purposes must, refer to the subjective meaning of the actions of human beings from which social reality originates. 33 Now, it is not clear how far Schutz wants to extend this generally cogent construal of Weber's postulate. But if he intends it to ex- haust all social inquiry, he limits such inquiry arbitrarily and un- necessarily. For all explanations of historical developments of social institutions, to take only one example, would thereby be ex- cluded. Such explanations cannot be made from the standpoint of an historical actor, from the standpoint of subjectivity, but must be made from the viewpoint of the historical observer. For our pur- poses, however, we need not take Schutz' thesis as extending beyond clear cases of purposive human behavior which are informed by mean- ings and which become intelligible to an observer when he understands the presuppositions of social action in the subject's community. Even with this limitation of scope, Weber's postulate still leaves us with the central question of whether or not it establiShes the SU thesis and hence refutes the claims of the CL theory. I think 33 A. Schutz, "Concept and Theory Formation in the Social Sciences. p. 2’45. 127 it does not. But to show why requires asking how the scientist's second-order constructs, ideals or theories are related to the first-order constructs of common-sense; and how it is possible to grasp by a system of objective and verifiable scientific knowledge the subjective meaning structures of human behavior. This was the very question at issue in the dispute between Nagel, Miss Lavine and Natanson, the question concerning methodological devices for attaining objective and verifiable knowledge of subjective meaning structures, the question of establishing controls for Verstehen. Schutz' answer follows from his distinction between the two levels of Verstehen or interpretative understanding, and between the first- order constructs of common-sense and the historical ideal typical constructs of these constructs. The latter are by no means purely subjective or arbitrary but accord with the "procedural rules valid for all empirical sciences“ and are: objective ideal typical constructs and, as such, of a different kind from those developed on the first level of common-sense thinking which they have to supersede. They are uleoretical systems embodying testable general hypotheses in the sense of Professor Hempel's definition. Schutz' answer turns on the construction of models of rational action which suggest the importance of teleological explanations in history and the social sciences to supplement causal explanations. His argument leads, it seems, to an empirically-oriented science of teleology in order to devise adequate methods of selection for resolving problems or achieving goals. He clearly recognizes the I 3’4 Ibid., p. 21.6. 128 dependence of meaning and truth in social inquiry upon the purposes of the inquirer. What he calls “understanding" might then be achieved not by the 'intuitionist“ construal of Verstehen but by a theory of experimental teleology, which takes explanatory hypotheses as means for achieving the objectives of the inquiry, and hence appraisable on grounds of efficiency. However, Schutz makes two additional assumptions, each more controversial and at best misleading. One concerns his view of the particular attitude of the historian or scientist to the social world; the other, one of his criteria for appraising these second- order theoretical systems of constructs: the postulate of adequacy. Following the value-neutrality thesis of Max Weber, adhered to by Nagel and Hempel, Schutz finds the proper attitude of the historian and theoretical social scientist to be the same as that of the natural scientist: the historian must be a mere disinterested observer of the social world. The theoretical scientist Iqua scientist," not qua human being: is not involved in the Observed situation, which is to him not of practical but merely of cognitive interest. It is not the theater of his activities, but merely the object of his contemplation. He does not act within it, vitally interested in the outcome of his actions, hoping or fearing what their consequences might be, but he looks at it with the same detached equanimity with which the nggural scientist looks at the occurrences in his laboratory. There is, of course, a sense in which this view amounts to sound methodological procedure, especially useful in controlling or 35 A. Schutz, “Common-Sense and Scientific Interpretation of Human Action", p. 336. 129 eliminating various biases of the inquirer. But in a deeper and more important sense, it is most misleading since neglecting the pragmatic dimension of inquiry by suggesting that the scientific enterprise (natural, social or historical) is value-free. It indi- cates that the scientist qua scientist can and must decide to accept or reject hypotheses independently not only of the specific observed situation, but of all cost, decisional and value considerations. To attain objectivity in scientific results, to keep them under control, Schutz requires the scientist qua scientist to be "governed by the disinterested quest for truth in accordance with preestablish- ed rules, called the scientific method.“36 First-order common-sense constructs are formed from the per- spective of the actor within the world, which depends upon the actor's biographical situation and in turn determines his motives, attitudes and purposes. The historian, however, considers his position within the social world as irrelevant to his explanatory undertaking. He replaces his personal biographical situation with his inquiring situation. What is taken for granted in daily life may be a subject of inquiry for him, and vice versa. What seems relevant on one level may not be on another. Again, if this means only that the inquirer Operates with a different framework, set of presuppositions, purposes and system of relevances than does the common-sense man in daily life, then Schutz is clearly correct. But to the extent he suggests that either 36 Ibid., p. 337. 130 the social scientist or historian operates in a closed system with a fixed set of procedural rules, to this extent is his thesis erroneous. And the use of "objectivity" to mean I'value-free' or "independence of the inquirer's attitudes or values‘ implies that one can decide to accept historical, or other statistical, hypo- theses on the basis of purely logical considerations independently of the very purposive dimension he emphasizes. In a later chapter we argue against this view. Rather, the pragmatic element, and the subsequent making of value judgments, constitutes an intrinsic ingredient in such decisions. Hence ”objectivity" must be interpreted in a manner wide enough to allow for them. Neither the historian nor the scientist can really es- cape his own biographical situation. Nor can the historian avoid being ”vitally interested in the outcome of his actions,n for the relationship between believing and acting, it will be argued, is much closer than Schutz, weber, Nagel or Hempel seem to recognize. In fact, this relationship constitutes precisely what I take to be the important insight of Verstehen theorists. That is, one important difference between the various empirical sciences, a difference in degree, concerns the extent to which the acceptability of hypotheses necessitates the use of criteria other than evidential or confirma- tory strength, criteria involving the denial of the VN thesis. If this case can be made, perhaps the peculiarity of the his- torian's subject matter, intentional meanings, can be accounted for as merely requiring a heavier reliance on these other criteria than do the natural sciences. To see what bearing these elements have 131 on historical explanations of human actions, we will direct our attention in the next chapter to Dray's normative reconstruction of Verstehen and SU. His subsequent criticisms of the CL model turn on a rejection of VN. Meanwhile, we need comment on the second assumption of Schutz' position, one related to the 'epistemological' level of Verstehen. Although the constructs of the historian and social scientist are removed from and refer to the constructs develOped at the common- sense level, they are, according to Schutz, by no means arbitrary. They are appraisable according to three postuhates: logical con- sistency, subjective interpretation and adequacy. The latter de- serves our attention, since the second has already been examined and the first obviously needs no elaboration for our purposes. The criterion of adequacy is designed to assure “that the thought ob- jects of the social sciences... remain consistent with the thought objects of common sense, formed by men in everyday life...."37 That is to say, Each term in a scientific model of human action must be constructed in such a way that a human act performed within the life-world by an individual actor in the way indicated by the typical construct would be understandable for the actor himself as well as for his fellow-man in erms of common-sense interpretations of everyday life. This postulate or criterion of adequacy, however, needs further clarification, In what sense these human actions must be explained so as to be "understandable to the actor himself" in common-sense 37 Ibid., p. 3h2. 33 Ibid., p. 3h3. 132 terms, Schutz fails to specify. It is clear though why he feels compelled to postulate some such condition of adequacy. His pri- mary concern here is with the twofold naturalistic claim: that human behavior should be studied and explained as the natural scientist explains his object, and that the goal of history and the social sciences is to explain ”social reality. as experienced by man living his everyday life. For these two claims he takes to be incompatible with each other. This incompatibility results from the fact that the more fully refined and developed the abstract system of second-order constructs becomes, the furthm nmmved it is from the first-order constructs of common-sense in terms of which men experience their own and others' behavior. Thus, to avoid this difficulty Schutz postulates his condition of adequacy. In addi- tion, he advocates the use of particular methodological devices, models of rational action, controlled by the postulate of sub- jective interpretation. Taking the two naturalistic claims to be incompatible, and compelled to accept some sort of consistency between historical or scientific explanations and common-sense un- derstanding,he denies the possibility of explaining human behavior in the same manner as the objects of natural science. Thus his de- fense of Weber's SU thesis. Close analysis of Schutz' writings, however, reveals no serious reasons why human behavior cannot be explained in terms of the CL theory. That is, granting that the theory countenances the con- struction of models of rational actions, and that the laws appealed to in the explanations are not merely mechanistic or universal. 133 But then neither Popper nor Hempel restrict the kinds of laws usable in the CL model so as to preclude teleological, functional or sta- tistical generalizations. Hence, if the incompatibility between the two naturalistic theses depends on explaining human actions on a purely mechanistic level, Schutz is certainly correct in denying this and opting for the consistency between common-sense and his- torical understanding. Two further comments seem necessary at this juncture. First, we will have to pursue in more detail than does Schutz the question raised above: in what sense and to what extent must historical ex- planations be consistent with those of common-sense and understandable to file historical agents or actors. But since this question relates in important ways to William Dray's pragmatic interpretation of Verstehen along the lines of reconstructing the agent's rationale, to be considered in the next chapter, it will be more instructive to discuss the question in that context. Further, Dray's early work depicts the notions of historical explanation and understanding so as to make the compatibility of these notions with those of common usage an important, if not necessary, condition for the adequacy of a theory of historical ex- planation. Schutz' criterion of adequacy, in other words, correlates closely with the recent analytic notion of ”counterintuitivity“ whereby a theory or meaning is considered at least prima-facie un- acceptable if it violates, or is not in accord with, sound common usage. Even Hempel in his defense of requirement (Rh)’ you will recall, resorted to some such criterion. we argued in Chapter II 13h that such a criterion was neither clear nor acceptable as a condi- tion of the adequacy of philosophic explications or theories, and that it might be replaceable by a better one, one related to file purposive and pragmatic aspects of explanation. Hence, we will also pursue this aspect of Schutz' criterion in connection with Dray's early work. But, secondly, before accepting the extension of the CL theory to teleological or motive explanations, we will also con- sider whether there are other reasons than those adduced so far for rejecting the CL theory as an adequate model for explanations of human behavior. In particular, we will pursue Schutz' sugges- tion that such explanations require the construction of models of rational action controlled by flue postulate of subjective inter- pretation. This requirement serves as the basis of recent de- fenses of weber's SU thesis. Accordingly, some such reasons, and alternative models of explanation, proposed by recent analytic philosophers of history, will be examined in succeeding chapters. we open our discussion of this question in the next chapter with an analysis of Dray's extreme attack on Hempel's CL model, and of his novel alternative model of rational explanation. Then, in Chapter V, we pursue the more moderate criticism of the CL model offered by Gardiner, Donagan, Brandt and Scriven. The main thrust of both the extreme and moderate critics, it will be argued, agrees with that of Lavine and Schutz. The attack should be directed not at the CL theory but instead at the VN thesis. In the final chap- ter this thesis will be defended. CHAPTER IV WILLIAM DRAY'S RECONSTRUCTION OF THE SU THESIS In the last chapter we reviewed some criticisms of the covering law theory of explanation when extended to history. The discussion centered on explanations of human actions, as one kind of historical explanation. we considered various formulations and defenses of the SU thesis and of an alternative type of explanation or under- standing. Yet the result seems to be that none of the reconstructions of weber's position, though offering important suggestions, seriously damage the claims of CL theorists. So far, at any rate, we have found no substantial reason for upholding SU or IT, for suspecting that the insights of Verstehen theorists concerning the peculiar subject matter of history makes the CL model "peculiarly inept.” Nor that historical explanations have a "different kind of logic“ than that of scientific explanation. That is, so long as purposive elements like human motives, attitudes, goals and purposes can be adequately fit into the model. But this is precisely the point that requires more attention, since much recent criticism of the CL model tries to show the model's ineptness because it cannot account for human actions and dispositions. The inability to explain adequately the rationale or subjective meaning of human actions is still offered as the single greatest obstacle to the CL model. These and related charges receive one of their most cogent and 135 136 persuasive statements in a recent book by William Dray, Laws and Explanation in History. It is to Dray's attempted rehabilitation or reconstruction of weber's SU thesis that we now turn.1 The doctrine under consideration, you will recall, begins with the premise that the objects of historical inquiry, human actions, differ fundamentally from the objects of the natural sciences. It concludes that such objects cannot be explained or understood merely by subsuming them under empirical covering laws. Hence the cover- ing law theory is adjudged inappropriate in history, which is taken as an autonomous discipline with a peculiar "logic“ of its own. Upon brief survey of both the idealist position and the "official. covering law answer to it, Dray decides that the latter evades the main thrust of the former's doctrine. The residue left out of their answer he depicts as not a psychological matter of discovery but something that "should properly be taken into account in a logical analysis of explanation as it is given in history.“2 Consequently, he attempts to rehabilitate or “makes sense“ of what weber and some of the idealists said about historical understanding, and to do so in such a way as to make the SU thesis immune to the charges of the I'official" answer. His analysis also develops some of the issues raised in the last chapter by Lavine, Natanson and Schutz. There is, however, one major point of interest, All of the critics of the covering law theory considered in the last chapter See especially Chapter V, ”The Rationale of Actions.‘ Ibid., p. 121. 137 agreed with Hempel, Popper and Nagel that the historian qua his- torian, as well as the scientisthua scientist, makes no value judgments. Most critics and defenders alike assume historical analysis and etplanation to be a value-neutral activity, the re- sults of which ought not to depend in any essential way upon the attitudes, interests or values of the inquirer. It ought instead to yield a set of identical conclusions for all competent inquirers "governed by the disinterested quest for trufli in accordance with preestablished rules called the scientific method."3 Dray, correctly I think, opposes this uneasy alliance. His reconstructed version of ”empathic understanding” depends essentially upon denying the value-neutrality thesis. But again, as with those considered in the last chapter, Dray erroneously takes this position to conflict with the main tenets of the covering law theory. For the historian, in order to explain some human action, according to Dray, must ap- praise or evaluate the action in its context as appropriate and rational, thus appealing for explanatory force to normative principles of action rather than empirically descriptive covering laws. Dray attempts to reconstruct the 'empathy' position, then, by negating one crucial tenet of Weber's position, viz. the value-neutrality thesis of empirical science. This is especially noteworthy since Dray rejects Hempel's CL theory as an unsatisfactory reconstruction of weber's position, even though Hempel agrees with weber regarding value-neutrality. ‘Wé then 3 See footnote #36, p. 129 138 have the puzzling situation whereby Dray, with Weber, insists on the SU thesis and on reason explanations as autonomous, but only at the price of giving up the value-neutrality thesis; while Hempel dis- agrees with Weber's insistence on autonomy, but agrees with him re- garding value-neutrality. I submit the thesis that weber is wrong on both counts, and hence that Hempel and Dray are both partially correct. Hempel correctly rejects autonomy and SU in favor of the covering law model, but Dray is equally right to reject the value- neutrality thesis. Moreover, I will argue that these two theses are not only compatible but that an adequate reconstruction of the covering law theory includes the denial of value-neutrality. Thus, if the two theses were incompatible, as Dray suggests, we would in- deed have to choose between them. But Dray's criticism of the covering law model on the grounds that it requires value-neutrality is misplaced. Value appraisals enter the domain both of historical and scientific explanations. Yet not in the way Dray suggests, i.e. not as the explanatory force. They enter instead, as Lavine and Schutz both indicate, in the inquirer's use of judgment to determine the acceptability of his explanatory hypotheses. Accordingly, our investigation of Dray's case centers on three central questions. First, does Dray's reconstruction of the idealist doctrine sufficiently establish the denial of value-neutrality? If not, secondly, is there any other way of securing this denial by modifying or redirecting Dray's argument? Can we, in other words, discover the weakness of his reconstruction so as to improve upon it? Finally, if an affirmative answer to either of the other questions 139 seems advisable, i.e. if the value-neutrality thesis can be success- fully attacked on any grounds, how would this bear on the covering law theory of explanation? Would, e.g., the denial of VN entail the SU thesis? we might note here that the most important relationship between these two theses seems to concern only the probabilistic model (P), not the deductive model (D). This point becomes increasingly im- portant when we see to what extent Dray's criticism of the covering law model generally is directed only to the deductive part of it. Hence, this chapter will be devoted largely to our first question, while succeeding chapters deal with the last two questions. To elicit in their fullest form Dray's proposals about the connection between historical value judgments and explanations neces- sitates a consideration of the continuing debate stimulated by Dray's book and related essays. In particular, we will consider Hempel's response to Dray's charges, which occurs in the context of a new attempt to extend the covering law theory to include within its scope explanations of purposive human actions. We propose in this chapter, then, to clarify both the covering law theory and Dray's alternative model of explanation in order eventually to answer our first question. lhO The Rational 5295; gf Explanation As a particularly clear and representative example of how historians explain individual purposive human actions in terms of motivating reasons or beliefs, Dray cites G. M. Trevelyan's account of the successful invasion of England by William of Orange. In response to the question, "Why did Louis XIV make the greatest mis- take of his life in withdrawing military pressure from Holland in the summer of 1688?" Trevelyan explains that: He was vexed with James, who unwisely chose this moment of all, to refuse help and advice of his French patron, upon whose friendship he had based his whole policy. But Louis was not entirely passion's slave. No doubt he felt irritationwith James, but he also calculated that, even if William landed in England, there would be civil war and long troubles, as always in that factious island. Meanwhile, he could conquer Europe at leisure. 'For twenty years', says Lord Action, 'it had been his desire to neu- tralize England by internal broils and he was glad to have the Dutch out of the way (in England) while he dealt a blow at the Emperor Leopold (in Germany).' He thought, 'it was impossible that the conflict between James and William should not yield him an opportunity.‘ This cal- culation was not as absurd as it looks after the event. It was only defeatefl by the unexpected solidity of a new type of Revolution. Such accounts Dray labels ”rational explanations,‘ because they reconstruct the 'agent's calculation of means to be adopted toward his chosen end in light of the circumstances in which he found himselfI in order to display "the rationale of what was done."5 In so doing they constitute a distinctly different kind of explana- tion than subsumption under empirically verifiable laws and initial or antecedent circumstances, since they employ a quite different. h Trevelyan, The English Revolution, pp. 105-6; quoted by Dray, ibid., p. 122. 5 Dray, ibid., p. 122 and 12h. lhl criterion of intelligibility from that formulated by the covering law theory. And.they employ a different criterion because the ”goal of such explanation is to show that what was done was the thing to have done on such accasions, perhaps in accordance with certain laws.”6 Since the infinitive ”to do” functions, for Dray, as a value term, he claims that ”there is an element of appraisal of what was done in such explanations; that what we want to know when we ask to have the action explained is in what way it was appropriate.”7 Accordingly, the reasons an historian offers to explain in this rational manner must be ”good reasons” from the agent's point of view, must be such that ”if the situation had been as the agent envisaged it..., then what was done would have been the thing to have done.” Hence, since rational explanations need not be covered by general empirical laws of either a universal or probabilistic type, weber's SU thesis is vindicated. The distinction to be drawn at this juncture is, of course, that between a cause and a reason, since the expression, ”An actor A did.X because of Y,” is ambiguous. ”Because” can serve to in- dicate sometimes a cause and at other times a reason. When one says, for example, ”Louis withdrew military pressure from Holland because he was vexed with James", one has offered a causal ”because” and hence a causal eXplanation that can easily be fit into the Ibid., p. 12h. Ibid., p. 12h. Ibid., p. 126. 1h2 covering law model. But when one says "Louis withdrew military pressure from Holland because he thought he could conquer Europe at leisure,” one offers not a cause but a reason for the action, one produces a rationale of the action, that which tends to either justify or excuse what was done. In such a case one offers what Dray calls a ”rational explana- tion” in order to make intelligible from the agent's point of view, his grounds for so acting, to make sense of his action. Only by reconstructing the agent's calculations or reasons, ”by putting yourself in the agent's position can you understand why he did what he did.”9 The whole purpose of Trevelyan's explanation, according to Dray, lies in showing that Louis' unfortunate action, even though based on miscalculation, was appropriate to the envisioned circumstances. In this way, Dray claims, we begin to see the point of the Weberian and idealist insistence on Verstehen, of behavior maxims and of the ”projection” metaphors, which covering law theo- rists dismiss as merely psychological or l'methodological dodges.” Collingwood's inner—outer dichotomy, for instance, becomes trans- formed into the cause-reason distinction and its corresponding causal-rational explanation distinction. The covering law model is accordingly claimed to be irrelevant to historical actions since we want to know not how actions could have been predicted in advance, but the reasons why people did the things they did. To ascribe causes to human actions apparently 9 Ibid., p. 128. 1&3 commits some sort of category mistake, in Ryle's terms. In rational explanations it suffices that what provides the agent with a reason for acting be a rationally necessary condition Unat he had no reason to do what he did otherwise. It is not essential to show that he gguld net have acted in the same way without having that reason. But Dray requires an additional characteristic for rational explanations, because the conceptual connection between understand- ing an agent's action and discerning its reasons or rationale is neither deductive nor probabilistic. Subsuming the action logically under suitable empirical laws is, he argues, neither a necessary nor a sufficient condition of explaining. Hence, he refuses to concede the point we emphasize: all of the above amounts merely to recognizing an additional pragmatic condition historical expla- nations must satisfy, a condition tacked on to those of the covering law model. Instead, Dray condemns the covering law model as essentially inept in accounting for human actions. Thus he finds it in need not of additional conditions but of essential modifica- tion. Rational explanations require not the connecting bond of general descriptive empirical laws but of normative ”principles of action”, standards of appropriateness and rationality. Such practi- cal principles ”express a judgment of the form: 'When in a situa— tion of type Cl...Cn the thing to do is X.'”11 To explain a persons 10 Dray, ”The Historical Explanation of Actions Reconsidered,” in 3. Hook (ed.), Philosophy'gf History (N.Y.: New York University Press, 1963), p. 129. ll Dray, Laws and Explanation.in History, p. 132. a} l' g,“ _ r; L 5 ' "u 1 ‘. 1m behavior, then, one must represent it as the reasonable thing to have done in the circumstances. One appeals to the general knowledge expressed by rules or principles of behavior instead of empirical generalizations, to knowledge of what to do rather than of what is usually done in such circumstances. Although Dray does not dwell on the kind of circumstances referred to in these principles, it seems clear from his general analysis that they must include at least reference to the agent's goals or purposes, his beliefs about the empirical circumstances of his action and al- ternative courses of action, and his own moral standards or principles of conduct. Consequently, rational explanations provide answers to questions of the form, ”Why did agent A do act I?” by offering the following type of explanatory argument: (A) 1. Agent A was in a situation of type Cl...Cn (1.3. C) 2. In a situation of type C the thing to do is X. 3. Therefore, agent A did X.12 The first part of model (A) specifies certain antecedent conditions as do the covering law models (D) and (P), But the second part, the connecting link between reasons and action which gives the argument its explanatory force, the principle of action, replaces the general empirical or descriptive laws of (D) and (P). Hence, as Dray claims, (A) clearly differs from the covering law model. It constitutes a distinctly different type of explanation and employs a different cri- terion of intelligibility, because it contains an element of appraisal. Hempel, ”Reasons and Covering Laws in Historical Explanation,” in 5. Hook, pp. 213., p. 15h. 11.5 Because the historian must appraise or make value judgments about the appropriateness of the agent's action to his reasons, historical explanations of human actions are sui generis or irreducible to the CL models. Rational explanations remain essentially different from those offered in the natural sciences even though the determining motives, beliefs or reasons can be classified among the antecedent conditions of motivational eXplanations. In this way Dray defends Weber's SU thesis by attacking VN. In emphasizing the importance of appraisals or principles of action as the explanatory force in rational or motivational explana- tions, Dray follows the recent lead of Patrick Gardiner. In general Gardiner advocates the covering law model in historical inquiries. But the looseness or vagueness of laws used in historical explanations, which allow considerable width of interpretation compels him to modify or refine the theory. Surely historians offer many eXplanations without committing themselves to any covering laws. But covering law theorists suggest that such cases are only "explanatory sketches“, partial explanations or enthymemes. They require conpletion in the sense of eliciting the governing laws in order to be defensible if challenged. ‘Without such a defense the explanation would be at best a pseudo one. Gardi- ner agrees that explanations must ultimately rest on warranting generalizations. However, he is impressed by the fact that historians seldom conform to this pattern or defense in practice. Instead of appealing to general laws, the historian often completes his explana- tion by filling in details about the situation under consideration, lh6 by telling more of the story. Further, the historian relies in his defense on personal decisions and judgments, since his interpreta- tions are so loose and porous. Hence, Gardiner, led by the actual practice of historians, claims that they, ”like the general or statesman, tend to assess rather than to conclude”; and that ”there is, indeed, a point in terming (for example) the explanations provided by the historian 'judgments.'” But these assessments or judgments are not just incomplete sketches as Hempel suggests, They are not "made, or accepted, in default of anything 'better': we should rather insist that their formulation represents the end of historical inquiry, not that they are stages on the journey towards that end." If so, however, it is not clear why Gardiner's account consti- tutes merely a refinement of the covering law model, as he claims. For, in suggesting that these decisional explanations are not ”half- way houses” or incomplete sketches to be filled in by explicitly formulated generalizations, but represent rather the ”end of histori- cal inquiry,” Gardiner surely argues for one version of SU. If such explanations are complete of their own kind, the covering law model is simply inapplicable in such cases. That these judgmental explana- tions constitute a sui generis category, not reducible to the covering law model, is further attested when he persistently claims not to be implying that they are 'subjective' in any vicious sense. For the word 'judgment' must be regarded as being ... simply that the criteria for assessing the validity of any 1 . . 3 Gardiner, The Nature 2f Historical Explanation (London: Oxford University Press, 195?), PP. 95-6. 1h? given eXplanation in history are, in general, different from those appropriate to the assessment of explanatiifls as they occur in certain branches of soientific inquiry. As already noted, Dray correctly holds that the logic of rational explanation requires showing the presence, on the occasion of the action, of the antecedent conditions. The circumstances surrounding the action, the agent's beliefs and goals, and the available alternative courses of action would be considered such determining conditions of the given action. But, of course, how- ever necessary showing this and the thing to do in such circum- stances might be, an historian must still establish which, if any, of these factor was in fact the reason for the agent's action. Surely without producing some evidence to this effect the historian's explanation of, say, why Louis withdrew military pressure from Holland in the summer of 1688 would be either incomplete, dogmatic or perhaps even a pseudo-explanation. For, though some factor may have been present when the action was committed, it might indeed have been causally inoperative. For example, in Trevelyan's expla- nation of Louis' action, Louis might have believed there would be civil war in factious England if William landed.there. Still, this fact does not establish this as the reason why Louis withdrew pressure from Holland, any more than the fact that he was vexed and irritated with James proves that Unis was the reason. How many persons are known to have hated J.F. Kennedy's civil rights stand enough to have assassinated him? Yet this kind of hatred was apparently not directly 11‘ Ibid., p. 95. 1&8 responsible for his death, as many at first assumed. Even if the assassin was so motivated, it would still remain to be established that the assassination occurred because of this hatred, instead of for countless other causes or reasons. But if an historian's proof that a certain factor was present does not establish this factor as a reason why the agent acted as he did, then how can an historian support a claim for the causal efficacy of any given factor? In what way, in particular, does a principle of action provide support for such a claim? To clarify Dray'w own position it will be helpful to dwell for a moment on the position he rejects, the solution offered by Gardiner, Nagel and others. Covering law theorists generally resort at this juncture, of course, to empirical laws or generalizations to the effect that whenever such conditions occur, events of this sort result. With Nagel they would contend: The historian can justify his causal imputation by the assumption that, when the given factor is a circumstance under which men act, they generally conduct themselves in a manner similar to the particular action described by the imputation, so that the individual discussed by the his- torian presumably also acted the way he did because the given factor was present. To this Gardiner adds that the "because" in motive explanations, as "John hit you with a hammer because he is bad—tempered," represents an "instance of how he can in general be expected to behave under certain conditions. It sets John's action within a pattern, the 1% I E. Nagel, The Structure Bf Science (N.Y.: Harcourt, Brace and World, 1961), p. 555. 1&9 pattern of his normal behavior."16 And such patterns are represented by covering laws or generalizations. Hence, these explanations are not essentially different from psychological and sociological inves- tigations of why people behave as they do on given occasions, from what in current social research is referred to as "reason analysis.” Yet, as noted above, Gardiner is unhappy with the appeal to vague generalizations about human behavior as a defense of the his- torian's causal imputations. He turns for a replacement to judgments, decisions and assessments. Most covering law theorists, on the other hand, argue that such judgments ought to be replaced by inferences by inferences from covering empirically validated laws or generaliza- tions; otherwise the historian's defense of his explanation would not be certified as rationally acceptable. No doubt such cases of historical reason explanations are in- complete when judged according to the ideal covering law models. Sometimes they are indeed enthymematic, containing implicit statistical generalizations which do not deductively entail a singular conclu- sion but only support it with some degree of probability. At other times they amount to mere I'explanation-sketches." But often flney are partial explanations, described in Chapter one. Nagel's recent analysis of the general structure of partial explanatory arguments remains the most instructive: Let A1 be a specific action performed by an individual X on some occasion t in order to achieve some objective 0. However historians do not attempt to explain the performance 16 Gardiner, pp. 233., p. 125. A'/ I, {i f' ‘ é =..‘Il' . 5 . _V i 150 of the act A, in all its concrete details, but only the performance by X of a type of action A whose specific forms are act A1, A ...,An. Let us suppose further that X could have achieved the objective 0 had he performed on occasion'tany one of the actions in the subset A1,A2...Ak of the class of specific forms of A. Accordingly, even if a historian were to succeed in giving a deductive explana- tion for the fact that X performed the type of action A on occasion t, he would not thereby have succeeded in ex- plaining deductively that X performed the specific action Al on that occasion. In consequence and at best, the his- torians explanation shows only that under the assumptions stated, X's performance of Al on occasion t is probable. Apparently, then, historical explanations of individual actions are at best interpreted, according to the covering law theory, as eases fitting model (P), because of the essentially statistical character of generalizations about human motives, reasons and con- duct. At any rate, the heart of this view is that only empirical laws can serve the logical function of producing explanatory force by connecting the antecedent conditions (reasons, motives or causes) to the explanandum-event (physical events or human actions) either deductively or probabilistically. Let us see, now, why Dray objects to this analysis, why he re- sorts to principles of action as a substitute for empirical generali- zations, and how well they respond to empirical evidence. Dray seems most concerned with the claim that historical events and conditions are unique, and hence require to be accounted for by characteristically historical explanations. He contends that the historian, in his attempt to explain the French Revolution say, is 'just not interested in explaining it as a revolution," as Nagel's account would suggest. Nagel, pp. cit., p. 558. 151 Instead the historian is almost invariably concerned with it as different from other members of its class... that is to say, he will explain it as unique, (not absolutely but) in the sense of being different from others with which it would e natural 23 group it under a classification term... Consequently, treating such events or actions as instances of anything, subsuming them under classification generalizations or laws, is to abandon historical inquiry for scientific. Such accounts leave out, for Dray, the most important ingredient of judgment. The missing element is surely a 'law' or 'rule' which would inform the historian when such a group of 'pre- disposing' conditions become sufficient.... The conclusion that the revolution or unpopularity could reasonably have been predicted....would be reached by an excercise of the historian's judgment in the particular case.... Collating a number of conditions, including supporting laws, is not applying a further covering law, perhaps in a vague way. It is doing something quite different and much more difficult.19 The ”something different" now appears, however, to be something like Lavine's decision to accept the hypothesis that the weight of the evidence suffices to warrant our belief, i.e. a weighing of a set of evidential factors. And Nagel's attempt to represent this judgment in simple, formal terms is considered mere I'prejudice." Another supporter of this position, M. Scriven, sees the "great truth in the Verstehen theories” to consist in their badly conceptualized formulation of 'the indispensability and efficiency of the historian's capacity to respond to the cues in a well-de- scribed situation, so that he may with justifiable confidence 18 Dray, Laws and Explanation in Histogy, p. h7. 19 Ibid., p. 55. ,9 l’. ‘2 m ‘\ ) )7 .z .1 ' ‘ .5 I : i :I I 152 .20 accept or propose a particular reason-explanation as correct.... With this view we have repeatedly concurred, yet we also stress the failure of the accompanying reconstructions to explicate and develop this ”great truth" sufficiently enough to count as a serious chal- lenge to the covering law theory of explanation. If our interpretation is correct, it would seem that Dray and Scriven have confused three different questions: what constitutes an historical explanation? and how can one justify or defend an historical explanation? and what constitutes an acceptable his- torical explanation? Dray's argument began by trying to show that unique historical events could not profitably be subsumed under covering laws. But it ends by showing the need for additional judgments in order to defend and collate the covering laws that do profitably enter into historical explanations. we submit that by clarifying the above confusion, and by re- lating the appraisal element to the notion of justification and the rational acceptance of hypotheses or explanations, Dray's case can be transformed into a serious challenge to the covering law theory on its probabilistic side. In other words, Dray's argument fails to support the conclusion he wants: that historical explanations can be fully warranted or rationally acceptable without covering laws of any sort. Nevertheless, it suggests a related criticism of the covering law theory: that the historian qua historian, in his explanatory practice, must make value assessments or judgments. 20 M. Scriven, I'New Issues in the Logic of Explanation,'I in S. HOOK, 5220 EEO, pp. 358-90 153 And make them not in lieu of but over and above, and especially about, his covering laws. Such judgments concern the extension of the logic of explanation to include model (P), and hence the rational acceptability of the statistical generalizations employed in, for example, Nagel's account of partial explanations. This is a point barely touched on by Lavine, Schutz and now Dray, but nowhere de- veloped to its full potential by any of them in their reconstructions of the SU thesis and the notion of Verstehen. Accordingly, we examine it in greater detail in chapter six, after analysing Dray's position. But it must also be noted that Dray has in no way established even unis point. Indeed he merely asserts it to be the case. He sets out to establish, as we have seen, the necessity for rational explanations to contain assessments, in the sense of principles of action substituted for empirical generalizations. Precisely this move confuses what we take to be his important but undeveloped in- sight. The issue is not whether laws, as opposed to judgments, pro- vide explanatory connection and force, but how to determine which laws are rationally acceptable as explanans. This, we suggest, as a case of practical judgment, casts suspicion on the VN thesis without in any way supporting SU- And such suspicion Opens for inquiry many important questions. But before considering the tenability of either the "rational“ model of explanation as a rehabilitation of the "empathy” point of view, or of its peculiar emphasis on the normative element of ap- praisals, or even of the specific objections contained therein to —‘*‘.-‘..-"I.. ‘VI' . 1‘. - I I w p v ' .5 p . p .. j w, I . £ . .- It 15h the covering law model, two further aspects of Dray's model need elaboration. One concerns the essentially pragmatic analysis of explanation offered by many critics of the covering law theory. The other relates to Gardiner's fear of introducing viciously sub- jective elements into historical explanations as a corollary to the entrance of appraisals or judgments. This, of course, raises once again the major problem of testability and pseudo explanations, faced by all advocates of Verstehen. Dray recognizes that "To allow the legitimacy of empathy appears to many of its Opponents as the granting of a license to eke out scanty evidence with imaginative filler."21 His case, in other words, needs completion by showing in what sense "rational" explanations are logically, not just psy- cholOgically, different from the covering law model; and in what sense they are responsible to inductive evidence and do not go beyond the controls of empirical inquiry. Let us consider the for- mer point first. 21 Dray, pp. 313., p. 129; Laws and Explanation in History, p. 129. "‘ "‘—"' .w {-9 \ g L ' . - :7. I 4' lSS Pragmatic Dimensions 2f Explanation From our earlier account of the covering law theory, it is clear that its proponents provide a formal analysis of 'explana- tion,’ as showing something to be subsumable under or deducible from general laws, mainly in order to achieve some objective cri- teria for what counts as sound or rationally acceptable explanation, and also as a genuine scientific theory of human action. They have been most reluctant, consequently, to countenance the pragmatic as- pects of explanation, no less the element of empathy, as little more than a psychological peculiarity. It will be essential, then, to consider Dray's complaint that CL theorists mistakenly take 'explanation' to be a term of formal logic instead of mainly a pragmatic term. For Dray proposes to deal not with the psychology but the logic of historical thinking. Yet he uses the term 'logic' not in the narrow sense of formal logic but in the much broader sense made popular by contemporary pragmatic and analytic philoso- phers. There is, he claims, in the broad sense "an irreducible prag- matic dimension to explanation,'22 which helps to bring 'the analysis of the concept more into line with the way the word is used in the ordinary course of affairs.“23 This broad interpretation of explanation provides it with greater scope by making it applicable to such ordinary ways of talk- ing as "explaining my meaning," "explaining the use of a tool," 22 Ibid., p. 69. 23 Ibid., p. 75. 156 "explaining my point of view,‘ and ”explaining my purpose.” In this way his version also correlates with Schutz' criterion of adequacy, mentioned in the last chapter. By thus broadening the notion of explanation, Dray, along with Schutz, hopes to show the implausibi- lity of the CL claim that its restricted formal meaning can apply to historical explanations. This aspect of his pragmatic interpretation, and of Schutz‘ criterion of adequacy, we find unconvincing and misleading. For to achieve this goal, they show how the CL model departs drastically from the ordinary meaning of the term 'explanation.' But this is to object to the CL theory because it "prescribes a sense of the term, rather than calls attention to one already accepted,"2h and moreover, prescribes in the sense of importing into historical cases a special, technical sense of the term designed for narrow scientific uses. Apparently Dray wants a description of ordinary usage in order to revive the earlier idealist distinction between generality and explanation on the one hand, and intelligibility and understanding on the other. In Scriven's terms: Explanations are practical, context-bound affairs, and they are merely converted into something else when set out in full deductive array.... Explanation when dress- ed in its deductive robes becomes a proof or a justifi- cation of an explaggtion (and usually no longer explains but demonstrates). Hence, Dray and his supporters stress still other aspects of 2" Ibid., p. 79. Scriven, ”Truisms as the Grounds for Historical Explanations," in P. Gardiner (ed.), Theories 2f Histogy (Glencoe: Free Press, 1962), p. 1850. 157 their pragmatic version of explanations. First of all, explana- tions are “context-—bound affairs.“ They occur not in abstraction but in concrete cases. Provided at a definite time, in specific circumstances, and for a specific purpose; they are, as indicated in Chapter II, inference-tokens not inference-types. Accordingly, to grasp Dray's intended rationale for his rational explanations of historical actions, it is important to see the varied contexts in which we can and do ask for an eXplanation. He assumes that the demand for explanation arises out of a genuine puzzlement, that we can only offer an explanation in definite contexts where there is some particular gap in our knowledge, some particular perplexity or puzzlement, ppg. "when from the 'considerations' obvious to the investigator it is impossible to see the point of what was done."26 To clarify this view, Dray introduces the notions of an explana- tory "scale“ and a kind of "logical equilibrium.” The simplest or complete case of rational explanation occurs when an agent's act is perfectly intelligible, when he did exactly what the inquirer would have done in similar circumstances. From this complete case, rational explanations are then scaled along a continuum, I'depending on the amount of 'foreign' data which the investigator must bring in to complete the calculation.“ The 'foreign' data consists of the agent's beliefs, principles and purposes which differ from those the inquirer might have employed. Such explanation attempts to "match" an action with its calculation, i.e. to achieve a kind of "logical 26 Dray, pp. 213': p. 125; Laws and Explanation.ip History, p. 125. 158 equilibrium." Since the demand for these explanations only arises from one whose equilibrium is upset, the function or purpose of explanation must be to fill in the gaps and produce understanding by resolving the particular puzzlement and restoring the equilibrium to the given case. This is achieved by uncovering the agent's reasons or cal- culations for his acting as he did. For, when we uncover, say, Louis' reasons for withdrawing pressure from Holland, we see the point and make sense of his action, ‘we see that given his particu- lar beliefs, goals and grasp of the circumstances, he had reasons for doing what he did. But what aspect of his calculation requires nfilling in" depends on the particular gap in the understanding of the person to whom the explanation is directed. Thus, the correct- ness or appropriateness of the explanation is relative to the con- text in which it occurs, to the needs or perplexity that produced the demand for an explanation in the first place. Understanding re- sults from the ”perception of relationships and hence may be con- veyed by any process which locates the puzzling phenomenon in a system of relations," but I"we deduce nothing; our understanding comes because we see the phenomenon for what it is...."27 Dray's position appeals strongly to many practicing historians. One, a participant in the New York University SympoSium on History and Philosophy, suggests he would even call Dray "the historian's 27 Scriven, "Explanations, Predictions and.Laws,” in Minnesota Studies i2 the Philosophy of Science, Vol. III (Minneapolis; Univer- sity of Minnesota Press, 1552), p. 193. 159 philosopher” were it not for the suspicion that this would be 'the kiss of death."28 Another, however, accounts for such popularity on the grounds that Dray's rational model produces “simply the common, garden variety of understanding presumably possessed by each of us,” while Hempel's covering law model has “ominously threatened the historian with all sorts of generalizing sciences which he would have to understand."29 Now although this psychological account may or may not be ade- quate, it at least points to the apparent disparity of levels at which Hempel and Dray are operating. Dray claims to be describing the logic of historical explanations as they are actually offered by practicing historians, and hence emphasizes their pragmatic di- mension. Hempel, on the other hand, claims to be offering a pre- scriptive philosophic explication of the logic of historical ex- planations. Hence he abstracts from the concrete contextual situa- tion of any given explanation in order to codify principles and con- ditions concerning the syntactical and semantic aspects of the soundness or rational (not just personal) acceptability of such ar- guments. This disparity, however, is only apparent not real. For Dray, when pressed, concedes that he is not describing historical explanatory practice any more than Hempel. His quarrel with Hempel and other covering law theorists is not that they misrepresent or overlook the 28 L. Krieger, "Comments on Historical Explanation in History,u in S. Hook, pp. EEE°’ p. 137. 29 B. Mazlish, “On Rational Explanation in History," ibid., p. 282. 1 (; \.~ 3 { 1 . -. . t . . . x I <. t“ , D . ‘ f t , 160 content of rational explanations, but rather Uiat they misinterpret the form or logic of such explanations. But only recently, and often after much persistent argumentation on the part of covering law theo- rists, has Dray recognized the important philOSOphic task to be not merely describing or duplicating what historians actually do, buta n'rational reconstruction' which may not, in every instance, coin- cide exactly with what a practicing historian does.“30 In other words, Dray's rational model of explanation occurs on the same level as Hempel's covering law model. Both are explications not descrip- tions. They are therefore serious competitors and need to be ap- praised as such, i.p. to the extent each constitutes an adequate codification of ordinary historical explanatory practice. The issue does not concern which is "closer to" or nbetter duplicates" such practice, as Dray's earlier writings mistakenly suggest. But, as already noted, one main goal of formulating a theory of explanation, which motivates covering law theorists, is that there be some objective way of determining what counts as a rationally acceptable explanation. Consequently, Dray's rehabilitation of the "empathy‘I position must meet the charge that rational explanation goes beyond the scope and control of empirical inquiry by introducing viciously subjective elements into historical explanations, thus making them not just "explanation sketchesu but pseudo-eXplanations, and giving the historian ”a license to eke out scanty evidence with imagi- native filler.“ He must guard against the view that anything which O . . . . - 3 Dray, "The Historical Explanation of Actions Reconsidered," p. l07. 161 relieves perplexity can count as an adequate rational explanation of an action. Dray of course acknowledges this danger. He nonetheless re- jects the attempt to dismiss "empathy” as a mere psychological or Inethodological dodge, and to discount its counterpart ”rational explanation” as a poorer method of obtaining the same results as can be achieved.more reliably by subsuming actions under empirical Ilaws. Moreover, he defends his model against the above charge by .arguing that it does have "an inductive, empirical side, for we 'build up to explanatory equilibrium from the evidence," from his- torical documents, letters, speeches, rather than from.scratch. In this way controls are placed on an inquirer who might let his imagination run riot. Hence, Dray avoids any metaphysical appeal to self-evidence for rational explanations, such as those usually associated with idealist pronouncements about intuitively understand- ing an action by "an immediate leap to the discovery of its 'inside', without the aid of any general laws”31 or of any empirical reasoning at all. In fact, Dray readily concedes that mistakes are possible in the inductive reasoning of the calculation and that new infor- rmation may be uncovered to upset the calculation. Nevertheless, he claims the procedure to be "self-corrective" and subject only to the ”normal hazard of any empirical inquiry."32 Unfortunately, he fails to elaborate upon this empirically 31 Dray, Laws ppp Explanation ip History, p. 129. 32 Ibid., p. 130. ., . ‘ I " b . 162 half-corrective" theme. As a result it is unclear how he intends it, how it might be implemented, in what way it is scientific even in a broad sense of that term, Perhaps whatever empirically self-corrective aspects there might be to rational explanations are due to the degree it conforms to the covering law model. Dray fails to Show, in other words, that his model is susceptible to empirical controls at the points where it diverges from the covering law model. For having denounced Hempel's view of rational expla- nations as incomplete sketches in need of filling in, yet still in- dicating the direction of a better, more completed historical ex- planation—he seems compelled to give up Hempel's method of con- f irmation. But then no other method of confirmation is discussed, much less opted for. Dray seems driven to the same position as Gardiner who claims that “the criteria for assessing the validity of any given explanation in history are, in general, different from those appropriate to the assessment of explanations as they occur in certain branches of scientific inquiry."33 But this sharp con- trast between criteria appropriate to scientific and historical ex- planations has not been established, largely because defenders of the Verstehen position uncritically accept the covering law theorists claim of value-neutrality in scientific inquiries. ‘And‘it has been this error, I think, which has prevented Gardiner, Dray, Schutz and other defenders of Verstehen from sustaining a successful attack against the covering law model of explanation, and also from developing 33 Gardiner, pp. _c_i_t.., p. 95. the important insight of Verstehen theorists. Nevertheless, Dray may intend, as Gardiner does, that rational explanations require a "scientific" defense in a wide sense, gig. in requiring tests not limited to confirmation or evidential strength, in the same sense in which Lavine and Schutz earlier spoke of Verstehen. Clardiner mentions only one other method, that of practical success. One of his critics, Alan Donagan, promptly replies that such a test "may be employed in judging the assessments of generals and states- Inen" but "plainly does not aoply to those of historians," and hence that historical explanations as judgments or assessments are viciously subjective.3h Donagan's replg however, is particularly harsh. Unless of course he assimilates all assessments and judgments to matters of personal taste, in which case he would clearly be correct, but at the cost of misinterpreting both Gardiner‘s "judgments" and Dray's "principles of action." Another approach will perhaps be more fruitful. Suppose, instead, we pursue further the assimilation of principles of action to empirically confirmable generalizations, which would alleviate some problems of testing principles. We will do this in more de- tail in the next section as a serious alternative and criticism of Dray's model (A). But for the moment let us pursue this assimilation only to elicit more clearly Dray's position, and to dispel some mis- ‘understandings about it and the related topic of testability of principles. Dray leaves no doubt that the employment of rational explanations 3h ... A. Donagan, "Explanation in History," in Gardiner, Theories SEE History, p. h32. 16h contains an element of implicit generality, that "reasons for acting" no less than "conditions for predicting" have a kind of generality or universality. Suppose, then, despite the distinction between reasons and causes, we still contend that a sound, acceptable or complete rational eXplanation must treat the data of the agent's calculation as antecedent conditions, As empirical data from which ‘what was done could have been predicted, they are to be connected 'with what was done by a covering, empirically confirmable generaliza- tion. We might, in other words, concede a difference in content of "cause" and "reason" statements, yet argue for their similar logi- cal function in explanatory arguments. If we said, for example, "Disraeli attacked Peel because Peel was ruining the landed class," ‘we might mean, Dray agrees, that anyone sufficiently like Disraeli in relevant respects would have done the same thing in a situation sufficiently similar in relevant respects.35 And, generally, "if 'Y is a good reason for A to do X, then Y would be a good reason for anyone sufficiently like A to do X under sufficiently similar cir- cumstances.”36 But Dray objects to this assimilation of principles of action to empirical generalizations on the grounds that the universality in the two cases is sufficiently different to make the assimilation Inethodologically hazardous for the historian. This kind of procedure ‘would commit the historian who offers a rational explanation to the 3S Cf. the treatment of this locution by J. Hospers, Human Conduct (New York: Harcourt, Brace and World, 1961), pp. 320-22. 36 Dray, Laws and Explanation in History, p. 132. I 165 truth Of a corresponding empirical law or generalization. And this is precisely what it takes, according to the covering law theorist, to make such explanations responsible to empirical evidence. But Dray insists they are responsible to evidence in different ways, since empirical laws are disconfirmable (falsifiable) while prin- ciples of action are not. If we founda.negative instance of each, he maintains, "the law itself must be modified or rejected," while the principle of action "would not necessarily be falsified", and this because the former is descriptive and the latter prescriptive.37 This, at best, confuses a number of complex issues. On its face Dray's defense is clearly incorrect. Surely an empirical law, any more than a principle of action, need not be rejected or even modified on the basis of one negative instance. Surely the argu- ments of Duhem, Quins, Hempel and Pepper concerning falsifiability deserve more consideration than this. In fact, Dray's case suffers generally from an insufficient analysis of empirical scientific methodology, of description and confirmation, and of the role of decisions and judgments in.empirical inquiry. At any rate, if falsifiability on the basis of one negative instance is not sufficient to support the distinction between empirical laws and rational prin- ciples, Dray has only two alternatives. He can take principles to be unfalsifiable in principle or he can view them, along with laws, as partially falsifiable though in a more complex way. But here he faces a simple dilemma. If his principles of action 37 Ibid., p. 132. 166 are unfalsifiable, then of course they cease to be empirically sig- nificant statements about the world, which he claims them to be, since they are no lonfler amenable to empirical control. But then there is no sense in which they are "self-corrective" short of some metaphysical or a priori procedure, and hence the fear of Gardiner and others would be borne out: they become viciously subjective or "self-evident." If, on the other hand, his principles are at least partially susceptible to empirical refutation, as are laws, then finding some large number of instances, e.g. finding that most people do not act in accordance with them, would at least create a strong presumption against the claim of the principles about the thing to do in the given situation. But at this juncture Dray makes a peculiar move. Having cor- rectly noted that no amount of empirical evidence compels the with- drawal of the principle, he then claims that "if it was not withdrawn, the explanatory value of the principle for those actions which HEEE in accordance with it would remain."38 The consequence of this move is twofold. First, the principle of action loses its generality or universality since now applying only to those few cases to which it applies, i.§. it would not be "the thing to do" in the circumstances generally but only what some particular person or persons would do. Hence it seems more like a dispositional statement about the persons in question. Secondly, as the last comment indicates, the principle of action ceases to be 8 3 Ibid., p. 132. 167 an aopraisal or value judgment about what is rational in a given set of circumstances. Instead, it is transformed into a descriptive or predictive empirical statement about how some person or group will act in a given situation. Dray states that "the connxion between a principle of action and the 'cases' falling under it is thus intentionally and peculiarly 39 loose." Perhaps so, but only because according to this dilemma his principles of action are either appraisals which are not sus— ceptible to empirical control but are viciously subjective, or else they are not universalizable and not appraisals but limited dis- positional descriptive statements which are falsifiable. And since the second alternative would be compatible with the covering law model, Dray seems led to the first in order to differentiate his model (A) from the covering law model and to defend SU. Let me illustrate with a common garden-variety example offered by one of Dray‘s defenders, Kai Nielsen. “In 'Mrs. Finkbine had an abortion because she had good grounds for believing her baby would be deformed,‘ her warrant or the explanatory force for her action might have been 'Don't bear and rear deformed children'"ho Put into the form of his model it appears as: (A1) (a) Mrs. Finkbine believed on good grounds that her baby would be deformed. (b) Whenever there are good grounds for believing that a child will be born deformed, the thing to do is not to bear and rear the child but to have an abortion. (0) Therefore Mrs. Finkbine had an abortion. 39 Ibid., p. 133. hO K. Vielsen, "Rational Explanations in History," in 5. Hook, pp. cit., p. 308. 168 Now, when such examples are spelled out, it appears, first of all, that the rational model (Al) offers not so much the explanation of an action as the solution of an ethical problem. We seem puzzled why the conclusion (c) is descriptive rather than a prescriptive statement. We should expect from premises (a) and (b) the conclu- sion (cl) to follow, viz. "Mrs. Finkbine ought to have an abortion," or "The thing for Mrs. Finkbine to do is to have an abortion." But when (c) is offered as the conclusion we begin to see more Clearly some of the confusions surrounding what Dray is after, what he means by saying that rational explanations attempt to make sense of a person's action or to ”understand why such a person should do such a thing," without trying to predict what he did. Surely from (a) and (b) we could not predict (c). 30 the question-of importance is "In what sense do (a) and (b) explain (0)?" dhat, in particular, is the role and defense of principles of action such as (b)? We have already seen the dilemma to which Dray's use of "principles" leads him. Much of the difficulty turns on the ambiguity of "rational action" and "justification" implicit in the following passage. In the ordinary course of affairs, a demand for explanation is often recognized to be at the same time a challenge to the agent to produce either justification or excuse for what was done. In history, too, I want to argue, it will often be found impossible to bring out the point of what is offered as eXplanation unless the over-lapping of these notions, when it is hupan actions we are interested in, is explicitly recognized.‘l Dray here, and Nielsen in the above example, assume that an action, when provided a warrant or rationale such as (b), qualifies as a hi Dray, Laws and Explanation_ip History, p. 12h. 16-9 rational action. But obviously we must distinguish between a rational act in the sense of a reasonable or acceptable action and in the sense of an act merely done for a reason (reasonable or not). And also between the justification of an act in the same two senses. Accordingly, three corresponding interpretations emerge in addition to that offered above in the example of hrs. Finkbine. One stresses the ethical aspect of a reasonable and acceptable action, hence would really be an evaluative argument of the following kind: (A2) (a) Mrs. Finkbine believed on good grounds that her baby would be deformed (b) Whenever there are good grounds for believing that a child will be born deformed, the thing tg‘dg is to have an abortion. (90 Therefore hrs. Finkbine ought to have an abortion Another interpretation stresses, instead, the purely descriptive as- pects of an act actually based on some reason, without appraising that reason, hence would take a different form: (A3) (a) Mrs. binkbine believed on good grounds that her baby would be deformed. 01) Whenever there are good grounds for believing that a child will be born deformed, Mrs. Finkbine believes that the thing to do is to have an abortion. (c) Therefore Mrs. Finkbine had an abortion. As one can readily see, the only differences in the three for- mulations are, first, that (A2) contains an ethical conclusion (cl) in place of (c) making it an ethical argument and committing its author to the truth or moral acceptability of the warranting principle (b). Since Dray clearly intends not to commit the historian or any other author of rational explanations to such a genuinely ethical 170 appraiefiil, which would be hazardous indeed, he would surely reject (A2). Rut when we recall the other horn of our dilemma, derived .from Dray's insistence upon the empirically falsifiable aspect of (b), it seems that he must reject (b) as well as (Cl)’ in which case he is left with model (A3) which differs from (Al) only by containing (bl) in place of (b). For, you will recall, Dray wants to keep the ex lanatory value of principles like (b) even in cases vehere most people agreed it was not the thing to do, so long as “there was some one case, say with Mrs. Finkbine, where it was be- lieved to be the thing to do. In this situation the principle loses its universality and its appraisal quality, since limited to one case (or a few) and merely describes the agent's appraisal or belief.~ Nonetheless,ihis alternative frees the historian from any commitment to a principle of action, while still allowing him to explain an action as appropriate from the agent's point of View. But it now begins to appear, if our appraisal of Dray's dilemma :is correct, and if beliefs can be interpreted at least partially as dispositions, that reason explanations are perhaps dispositional in nature and have a form similar to (A3). Especially since what is (described by (0) seems related to what is described by (bl) as an iristance or manifestation. In this case they would be similar to IRyle's general analysis of "mental conduct concepts," motive expla- ruitions and "lawlike" statements in Ehe Concept 9f filfld, and hence vwauld still need modification to accord with the covering law model. However, I do not think Dray would be happy with any of the al- -ternatives we have so far suggested, i.e. (Al), (A2) or (A3). Despite 171 the dilemma mentioned above Dray clearly wants to keep premises like (a) and (b), and hence would reject (A3). But he would also reject both (c) and (cl) as the proper conclusions. Since he is neither offering an ethical argument nor willing to countenance the structural symmetry of rational explanation and prediction, he more than likely ‘would prefer a conclusion something like (c2), "Having an abortion ‘would be rationally possible or seem 'all right' to Mrs. Finkbine,“ and hence an overall argument such as: (Ah) (a) Mrs. fiinkbine believed on good grounds that her baby would be deformed. (b) Whenever there are good grounds for believing that a child will be deformed, the thing to do is to have an abortion. k?) Therefore having an abortion would be rationally possible for Mrs. Finkbine. The main reason for rendering his analysis thusly turns on his per- sistent denial that rational explanations allow of predictions or proof that the action did in fact occur. Let us now turn our attention to some of the objections brought against Dray's rational model of explanation. Our critique will lead us to consider some of the alternative models mentioned above, which serve to introduce Hempel's covering law, dispositional account of reason explanations. ThiS, finally, will raise our two major in- quiries: how reason explanations of purposive actions can be fit into the covering law model, and whether pragmatic appraisals or value judgments are essential in such a reconstruction. w u _ w , I q .. .l...: . .. .‘i " truer!“ ‘thylkrwmmhflmu 4‘ .Lf ’I. 172 Critioue of the iational hodel In this section we consider some objections raised against Dray's defense of the SU thesis, against his rational model of explanation and his attack on the CL model of Hempel and Popper. This critique will help us to see how, and with what modifications, the UL theory can account for explanations of historical actions and perhaps overcome some Cf the deficiencies of Dray's model. Having already alluded to some of these criticisms earlier in this chapter, discussion here is limited to five major points. But to facilitate and clarify the discussion, let me first intro- duce an illustration drawn from a recent symposium on related tonics. Consider, as an example of an historical eXplanation, Professor Gershoy's biographical account of a pivotal moment in the revolutionary career of Bertrand Barere, a little-known figure during the French Revolution. The action to be exolained is why Barere, in the Thermidor crisis of July, l79h, alligned himself with a loose and unsavory coalition of anti-Robespierrists in what developed as a successful rebellion against the Incorruptible, even though he had earlier publicly praised RobeSpierre as "a great republican." The common y accepted explanation of Barere's action held that he joined the plotters at the last moment, not because he shared their views, but only because of expediency since he realized they would be successful. This usual view, eSpoused first by Lord Macaulev, connects the action with the reason by the more general interpretation of Barere as "a cowardly ,173 opportunist.and a trimmer sucked into a fierce power struggle in which he displayed a skill amounting to genius in jumping unerr- ingly on the bandwagon of the stronger." However, Gershoy, sus- picious of this intemperate characterization yet convinced that such actions must be "interpreted and explained both in the context of his personality and the circumstances," offers a conflicting eXplanation of Rarere as a moderate mediator forced to a decision of how best to serve the desirable objectives of the Revolution. "Refore he took his final stand," Gershoy claims: Barere had tried to mediate a bitter diapute within the Committee of Public Safety in order to maintain its unity and effectiveness of action. He had also endeavored to ward off attacks that other avowed enemies of RobeSpierre outside the Committee were making on him. Reluctantly, because he convinced himself, little by little, that breaking Robespierre‘s hold over the Committee was more desirable than having the Incorruotible dominate it, he opted for the former. Once he made his decision, he joined with the opposition and played an active and important part in PobeSpierre's downfall.’ Now, to get the logic of the case clear let me formulate Gershoy's eXplanation, according to Dray's model (A), in the following way: (I) (a) At time T Barere was (Cl) a moderate mediator who believed (Cg) that his attempt to mediate the diSDute between RobeSpierre and other members of the Committee of Public Safety was futile, (C3) that the Committee's unity and effectiveness of action were desirable goals, (Oh) that RobeSpierre had a strong hold on the Committee, resulting in unity but also in domination, (Cg) that breaking this dominating hold would produce more unity and effectiveness of action in the Committee, and (C6) that actively joining the 1,2 - “ L. Gershoy, "Some Problems of a working Historian," in S. Hook, on. cit., p. 66. A —— 17h opposition would break Robespierre's domination of the Committee. (Let us refer to (C1) - (06) as (Cm).) (b) In a situation of type (Cm), the appropriate thing to do is X, to actively join the opposition. (0) Therefore after T Barere did X, i.e., he actively joined the opposition. And let me also, for comparative purposes, formulate the commonly held explanation, squested by hacaulev, in the following sketchy way. (2) (a) At time T Barere was (C7) a cowardly opportunist who believed (C8) that the only important goal was safety for himself, (C ) that the plotters would be successful, and (810) that his safety was most reasonably assured by actively joining the opposition to Hobespierre. (Let us refer to (C7) - (Clo) as (00).) (b) In a situation of type (Co), the appropriate thing to do is X, to actively join the opposition. (0) Therefore after T Barere did X, i.e., he actively joined the opposition. Now it must be remarked, first of all, that Dray's case is at least incomplete in the sense that while he emphasizes the need to show that an act was "the thing to have done for the reasons given," that it was "the appropriate thing to do" in the situation, he nevertheless offers no normative criteria of appropriateness or rationalitv. His main thesis, of course, does not require that he do so. Yet in the absence of any consideration of what rationality might mean in the context of historical actions or even generally, Dray is led to an over-simplified and dubious twofold assumption. He misleadingly assumes both that there is some one clear and uneouivocal sense in which an action, in a given set of circumstances, 175 is the apprOpriate or rational thing to do, and also that there is exactly one course of action in the given situation that is appro- priate or rational or the best way of achievine a goal in this sense. In cases (1) and (2), e.s., there is obviously not even an implicit criterion of appropriateness indicated by the historian. And this makes it appear, misleadingly, both that historians and others are agreed as to what "approoriate thing to do" means in situations like (Cm) and (CO), and that in such situations there really is only one apprOpriate thing to do, viz., X. Surely, however, in situations like (CO) it might be just as appropriate not merely to join the opposition but to lead it, so as eventually to replace RobeSpierre. And, more generally, as Hempel indicates in a brief review of recent developments in the mathematical theory of decisions, these assumptions clearly do not hold. For, first, even when the decision situation is of a kind for which one definite criterion of rational choice may be assumed to be available and agreed upon-~e.g., the principle of maximizing expected utility-~then that cri- terion may qualify several different courses of action as equally rational. Secondly, there are various kinds of decisions...for which there is not even agreement on a criterion of rationality, where maximin Opposes mafimax, and both are Opposed by various alternative rules."3 These various conflicting criteria of rationality, it should also be noted, reflect not merely differences in the evaluation of the goals available, but rather different inductive attitudes h3 Hempel, "Rational Action," Proceedines and Addresses of the American Philosophical Association, Vol. XXKV (Yellow Springs, Ohio: Antioch Press, 196?), p. 10. "H .fl 176 toward the world, attitudes for example of optimism or pessimism, of venturesomeness or caution. ‘We will have occasion to elaborate in more detail upon this point in our last chapter since the notion of "inductive attitude" bears heavily on Hempel's probabilis- tic model (P) and on his defense of heber‘s value-neutrality thesis. It suffices for the present simply to succest that Dray's account of rational eXplanation needs a good deal of bolsteriny from recent developments in decision theory, in order to avoid any rationalist myth of some one thing to do in any eiven circumstances as rationally necessitated. A second objection relates to the fact that in many historical actions there is no conscious deliberation or rational calculation leading to the agent's decision. As a result, Nowell-Smith argues,Ml any rational eXplanation in Dray's sense would be falsified in cases where the agent was found not to go through the relevant calculation. As a result, Dray's model is overly intellectual- istic by making human actions appear more rational than they are. Dray concedes that not all actions are performed deliberately in the sense required by his model. Yet he resists the temptation to say that in such cases there is no calculation to be reconstructed by the historian. He contends that "in so far as we say an action is purposive at all, no matter at what level of conscious deliberation, there is a calculation which could be constructed for it: the one the agent would have gone through if he had had hh Nowell-Smith, "Review," Philosophy (April, 1959), pp. l70"2 o 177 time, if he had not seen what to do in a flash...etc. And it is by eliciting some such calculation that we eXplain the action."MS Doubtless there is a calculation that could be constructed. But the point at issue concerns the explanatory significance of such a fictitious set of reasons or rationale. The question here is not even the more complicated one, indicated in an earlier section and to be considered later, of whether or not the agent's reason was causally operative in the action. Instead, it is how or in what sense a fictitious calculation can be said to eXplain an action. This seems dubious at best, since if the agent did not calculate his decision, then considerations of rationality or appropriateness clearly were of no force in his decision. To explain his action by reconstructing what he might have, but didn't, calculate clearly runs the dancer of over-intellectuali- zation. One could not on this basis, for instance, distinguish between a deliberate and nondeliberate action. A calculation could in all cases be supplied: in deliberate and nondeliberate cases alike. It is all too easy to construct a pattern where there is nomeand hence to distort rationalistically an historical agent's actions. 1:6 Dray and Nielsenh7 reply to this line of criticism by admitting that such rationalistic distortions can, and do, creep 1:5 Dray, Laws and Exganation .122 History, p. 123. A6 Dray, "The Historical Explanation of Actions Reconsidered," p. 111. m Nielsen, 4L‘ cit., p. 310. 178 into narrative history written on this model, yet they insist that this difficulty pertains only to the particular content of an explanatory argument and in no way affects the correct £23m of a rational eXplanation. His account is a philosophic eXplication, Dray insists, of what makes an action understandable rationally. And this understanding results not just from a set of propositions which an agent consciously calculated or recited to himself, but from our perception of a rational connection between an action and the agent's beliefs and motives, from the ordering or reconstruc- tion of such ingredients in the form of a practical calculation. This reply, however, fails to recognize that the objection does relate to the form of rational eXplanations and not just to their content. It is a question of whether goals, reasons or principles can be invoked in such cases at all, whatever these reasons might be; not a question of distorting or mistaking the particular goals or principles, and whether or not there are empirical checks on possible mistakes. The point remains that our perception of such a logical connection between eXplanans and explanandum cannot eXplain why an agent committed the eXplanandum- action in cases when the agent took no account of such a connec- tion. Moreover, this objection is not adequately countered by E. Barker's supeestion that actions of "historical significance" are often undertaken by an agent with "an awareness that he may have to eXplain his conduct to some audience."118 For, in this h8 E. Barker, "Rational EXplanations in History," in 8. Hook, 22. cit., p. 183. 179 case, lkle agent may have acted with the goal of satisfying some audience, as when a Congressman votes for the federal allocation of funds for his hometown power plant, in which case this was his reason. That is, he had a reason, and hence the objection is no longer applicable. But if the historical aeent had no reason, "historical significance" of his act notwithstanding, no constructed reason can explain his action. In fact, this kind of a defense of Dray's model points to a related difficulty. For with historically significant actions the historian, no less than the historical agent, is likely to produce by his construction of the reasons mere rationalization. As Passmore observes, the explanations an individual hives of his own conduct to a public or private audience are often "hollow-sounding... as if they were constructed to satisfy our audiences rather than 119 as explanations of our action." There seems little doubt, g.§., that were Parere to explain his own action X to some public audience at the time, he would obviously explain it in terms of (1) rather than of (2), even if (2) were the correct explanation. In a sense, however, Dray's renly does successfully counter this difficulty, since this is a ouestion of content, of the hazards of any search for a particular reasons. o‘urely Dray's model does not imply that whenever we think we know an agent‘s reasons we actually do know them. To uncover his "real," as opposed to "good," reasons is indeed a difficult task. But it is not peculiar to Dray's proposal h9 J. Passmo e, "Review, n Australian Journal of Politics and History, (1958). pp. 269 ff. 180 indcha as he claims, concerns only the form not the content of rational explanations. Despite the seriousness of Nowell-Smith's objection, one qualification seemswwrthy of comment. For sore actions that are decided upon "in a flash" and without conscious calculation or ' So deliteration seem similar to actions that are deliberated upon. Consider for example the many complex maneuvers involved in per- forminc a tonsillectomy. Xo doubt at first a doctor learns to perform such an operation only by "painful" calculation and delib- eration, but eventually the appropriate procedure becomes automatic and he can perform it "in a flash," routinely, with no corscious thOUVht of the complex maneuvers involved. Perhaps this is what E. Barker intended in the example of an historical event being aware of the possibility of having to eXplain his action to an audience. Consider here the skilled politician who has learned how to "explain" his actions depend- ing on the audience he is addressing, or the one who has learned how (by careful calculation at first but later by "instinct") to defend his conflicting actions to bothersome newsmen or Committee members. In such cases as these the habit pattern thus acouired could be interpreted not as fictitious but as "consisting in a set of dispositions to react in certain appropriate ways in various situations.” However, following dempel‘s sureestion, "a parti- cular act of this kind mi ht then be explained, not by a recon- 50 H mpel, "Reasons and Coverinf Laws in Historical Explanations," in S. Hook, £2. cit., p. loO. lkl struchXi calculation or deliberation which the awent did not in fact perform, ... but by presentinv it as a manifestation of a general behavior pattern that the a ent had learned." But in this case the aopropriate action can be explained by appeal to dispositions, such as cowardice or moderation, and perhaps to covering laws in accord with the CL model. This, of course, indi- cates the direction of Hemoel's alternative version of "rational explanations" to be considered in the next chapter. 51 , Ibid., p. loo Further Objections to Dray's -etional hodel The third objection turns on the ambiquity mentioned earlier in this chapter, the ambiguity of “rational action," which makes it unclear vhether Dray opts for model (A1), (A9), (A3), or (Ah)’ [—1. i.e. to what extent urav s proposing an explanatory model in the form of an ethical solution or an evaluative argument. Dray offers an important contribution in his reaffirmation of the claim that understanding an historical action often depends upon our discover- ing the agent's reasons or rationale for acting, as in examples (1) and (2). In the sense of nein: done for some reason, an action can be considered "rational." but this is a long way from, and surely does not entail, appraisine the normative rationality of the act in the sense of endorsing the agent's reasons as l”good reasons” or the action as "the thing to do." Strawson, for instance, objects that a view "which soes as far as this makes 52- history impossible." durely no reputable historian would want, Spa_historian, to appraise abortion or joinini the opposition to llohespicrre as the rational or moral thin? to do as such. Now Dray does not want to so this far. He refuses to commit the historian to the truth or moral acceptability of a principle of action in the way that an aryument like (A2) with an ethical conclusion would do. He stresses the fact that the agent's reasons must be good ones only "from the agent's point of View,” and that Hthe appropriateness of his act is to be assessed only in relation StrawSon, "Review," lind (April, 1959), p. 268. 183 to th2 circumstances as he envisaged them."53 But even this is not sufficient, for it contains a third meaning of “rational action." It still commits the historian to the truth of a norma- tive principle of action of the form (b), "when in a situation of type C, the thing to do is X." Though now the principle is rela- tive to the afent's beliefs and goals. But, Strawson persists, this does not no far enourh since there are "differences of intellicence, temperament, ability and character"gh as well as of coals and beliefs, and human actions involve faulty judgment as well as faulty information. Dray's reply to this objection is, I must confess, baffling. He retreats by removing the stinn from his earlier attack on the CL model. He claimed in Laws and Explanation in History, you will recall, that the CL model was "peculiarly inept” in account~ ing for typical historical explanations and that the letter re- quired a peculiarly different kind of "lovic." In his latest account of the rational model, he concedes that his model with its accompanying criterion of understanding and appraisal of the thing to do "cannot be the only one, even in history."55 This reopens the possibility that the CL model miwht also be applicable in history, though perhaps in a different context. Dray, “The Historical Explanation of Actions deconsidered," p. 112; and Laws and Explanation i3 History, p. 126. SLv Strawson, op. cit., p. 268. C, r-1 7 t c "‘ . a .‘. I JSDray, ihe Historical LXplanation of Actions neconSidered," 0.,113. 18h TYNE concession occurs as part of his reply to Strawson's objection. He anrees, first, that human actions can fall short of an ideal of rationality because of faulty judnment. And in licht of his earlier persistent belief that non-deliberate actions could still have a rationale conStructed for them, one would expect him to claim in this case that actions falling short of an ideal of rationalitv in this sense could also be explained in accord with his model. For, it seems we can understand the reasons or rationale of a faulty inference or judgment, as weLl as of a false belief, without of course endorsing it as a sound in- ference, but only as the aopropriate one from the aqent's view of the circvmstances. However, instead of continuing this use of "rational action" which allows one to understand the rationale of an act without endorsing it as completely rational, Dray switches mean- ings. He thus limits the aoplicability of his rational model, concerning inferences, to onlv those which are valid. "For it is obvious," he asserts, “that we cannot claim ratirnal under- standing of the making of a logical error.... And one cannot re-think a nravtical argument one knows to be invalid." In other words, his claim for the criterion of rational appropriate- ness is limited to "actions not judged to be defective in various waYS-"Sé But such statements clearly suycest that "rational action" is now being used in such a wav that an historian or logician would have to endorse the agent's inference as ration- ally sound in order to understand its rationale. 56Ibid., p. 113. 185 '1 find Dray's retreat extremely baffling. But apparently no more than do some of his own defenders. Scriven, g.g., quite well recoenizes that the soundness of an inference is clearly not necessary in order to understand the agent's rationale in making the inference. Surely "if the historian can play the role of judging reasonableness with respect to beliefs he knows to be false, he can eoually well do it with respect to infer- 57 ences which he knows are logically unsound." Doubtless most logic professors can, Sometimes at least, "rethink" a student's loeic exercise which one knows to be invalid. Just as one can also "rethink” one's own bad arguments in order to improve them. Scriven consequently suggests, in a more consistent manner than Dray, that all reason-explanations are essentially similar. And the way in which they are similar reverts back to Dray‘s earlier analysis, since "we may understand why X did Z, namely because he thoucht it would achieve D, without at all thinking his judg- ment was defensible."58 In Sfite of Scriven's consistent defense of the rational model, howaver, his proposal fails to eliminate a fourth diffi— culty often raised against Dray's model. In fact, his inter- pretation of a rational action as one the agent did in order to achieve some goal, D, seems to vary considerablv from Dray's. Scriven, "New Issues in the Logic of Explanation," p. 3b9. SBIbida, pp. 3119-50. 186 It contains no appraisal or principle of action correspondins to (h) in (l) and (2) or (Ab) above. Instead, Scriven appeals to a descriptive premise of the form (bl), "when in a situation of type C, the aeent believes the thing to do is X," in order to avoid the kind of objection brought forth by Stawson, via that understanding an arent's reasons for acting is different from and not dependent upon appraising the rationality of the action. In so doino he wives up (Ah) in favor of (A3) as the most ade- quate defense of the rational model of explanation. But if this is the case, Scriven would aoparently agree, at least in part as shown in the next chapter, with Hempel's major objection to Dray‘s model. Let us consider this fourth line of criticism in some detail since it elicits the essentials of the issue and pre- pares the way for Hempel's dispositional analysis of reason-explana- tions. In chapter two we described Hempel's two ideal models of explanation and a number of requirements which he considers to be necessary though not sufficient conditions for the soundness or adequacy of any explanation of a given event. One of the most important conditions stipulated that "any rationally acceptable answer to a question of the type 'why did X occur?‘ must provide information which constitutes good grounds for the belief that X did in fact occur."59 And this, for CL theorists, means either deductively or inductively inferring the actual occurrence of the 59 p. 1&6. Hempel, Reasons and Coverinr Laws in Historical Explanation," 187 event fInnn a set of laws and antecedent conditions in accord with either model (D) or (P). Now Dray's rational model, read as either of the form (Al) or (Ah), differs from the CL account in two ways: by replacing the laws with a generalized normative principle of action, and conse- quently by loosening the connecting link between eXplanans and explanandum from deductive or inductive implication to something like "rational necessity or coherence." The main objection to Dray's model, then, is that to the extent it differs in these two features from the CL model, to that extent it fails to satisfy the condition of adequacy just mentioned. Hence it cannot eXplain its eXplanandum-event as described in (c), say, why "Barere actively joined the opposition." ioreover, in order to satisfy this condition, to provide "cood grounds" for the eXplanandum (c), a rational eXplanation would have to replace the principle of action (b) with both a descriptive generalization about what a rational agent might do in the circumstances, and a statement that the agent was rational at T. In this case it Would cease to be an appraisal and would become a CL exolanation. For example, in case (1) we would have to replace (b) with the following two descriptive statements: (b2) "Barere was a rational aeent at T," and (b3) " rational agent, in a situation like (Cm), will with high probability do X.“ Thus if rational exglanations are to explain adequately why some act occurred, they must be reinforced by a probabilistic covering law, the truth of which the historian is then committed to. And if 188 such descriptive generalizations are missing from an historian's account of an action, the account is best viewed as enthymematic, as an explanatory-sketch or perhaps a partial eXplanation. But in any event not as a complete eXplanation of its own kind. Of course, information of the kind sugsested by (a) and (b) of model (A) - that an arent was in a given situation and that in such a situation X is the appropriate or rational thing for the agent to do - does provide ample grounds for believing a con- clusion such as (c2), that it would have been aopropriate or rational for Barere to actively join the opposition forces. But this information clearly does not afford good grounds for con- clusion (c) of (A), that Barere actually did X. The historian needs to discover, in other words, the agent's actual reasons, not merely what good reasons there might be, for acting as he did. And this requires-- as Hempel, Passmore, Nagel, Scriven and Nowell-Smith all agree-- showing more than that the action made good sense from the agent's point of view. It requires showing "that it was because it made good sense that he did it: otherwise it isn't an eXplanation. It may have made good sense / from his point of Vie”, but he may have done it out of Spite...."OO ‘We might note here that even Gershoy interprets the "logic" of his inquiry to be "not fundamentally different from that of the scientist“, since his eXplanation follows the basic requirements of the CL model by relying on "singular statements" and "empiri- cal regularities" which "could be tested like any scientific f. u; 00 Scriven, "New Issues in the Logic of Explanation," p.3bO 189 61 eXplanation for accuracy, cogency and logic." However, he no- where offers any evidence to justify his own eXplanation, (l), or why he takes it to be better than the commonly accepted one,(?). And without such evidential support it would fail to satisfy Hempel's epistemic condition of adequacy (being true or highly con- firmed). Hence it would be at best a potential eXplanation, as discussed in chapter two. For Barere might well have had both diSpositions. He might have been a cowardly opportunist and a moderate mediator at the same time, without any incompatibility. Moreover, either disnosition is often manifested in behavior similar to Barere's actual action X, and either set of reasons or calculations, (Cm) or (Co), are rationally possible and "make sense" of his action. Thus, citing either set of reasons, say (Cm), by itself will not sufficie to eXplain adequately why Barere did X. Gershoy can only establish this claim, as Nagel convincing- 62 ly argues, by showing not merely that Barere could rationally have acted as a mediator, but that this disposition actually did operate as a causal determinant in his actual action X. It may have made good sense from his point of view, yet he might still have done it for opportunistic reasons. Gershoy must show that Barere did X because his mediator-calculation made sense. But to show this is to show that the agent was acting on reason at the time, and hence to appeal to some generalization 61 62 Na:el, "Relativism and Some Problems of Working Historians," in S. Hook, op. cit., p. 89. Gershoy, "Some Problems of a Working Historian," p. 67. 190 showin€;'that an a; nt with such actual and good reasons will usually act as this agent did. Only in this way can we provide good grounds for believing that the agent's good reasons in- fluenced his action or were causally Operative. Thus, the basic logical defect in Dray's rational model: while it helps to render an agent's action intellisible by showing how he could plausibly have done it and even how he would have done it if he were ration- al and acted accordingly, it nevertheless does not explain why he in fact did act as he did. The wood reasons still need to be linked to the explanandum-action by descriptive covering laws, not by normative principles of action. DeSpite this apparently decisive criticism, Dray and Nielsen persist in denying its force, thouph for different reasons. we will pursue each case in turn. It is clear, of course, that to rebut this criticism one must argue either that Hempel‘s condi- tion of adequacy is not a necessary condition of sound explana- tions of why some event or action did in fact occur; or that the rational model offered by Dray is not intended to codify this kind of explanation, and hence is more properly construed as, say, of the form (Ah) instead of (Al). In this case the conclusion 0f (1) and (2) WOUlfl be, ___. ‘ 237 situation, about his other beliefs and goals which are required by (C) but not included in it. However, Scriven refuses to follow Hempel to the conclusion that the connection and hence premise (C) is of the ordinary descriptive, statistical kind. He takes, rather, the intermediate position that action X is "not only an 30 example of, but a quasi-necessary consequence of rationality." No specific acts are necessary conditions or consequences of rationality, since the failure to so act on one occasion surely does not make one irrational. Yet Scriven maintains that the connection between rationality and its specific manifestations is more than a merely empirical or statistical one. Thus he refers to it as quasi-necessary since "it is logically impossible that someone be rational and exhibit BEBE of these manifestations." He further explicates it so that the ”consequence of being a rational agent is that he will do A in C at T, where A appears to him to maximize his expectations of his desired goals D. This is what he will normally do but special circumstances may lead him to do A' without destroying his claim to be rational."31 Precisely how this account differs from Hempel's statistical construal, however, still remains obscure. Scriven offers some help by suggesting that premise (C) looks tautological because, and here he follows Dray, it is an evaluative prOposition, a pru- dential maxim or rule "informative as to what should (rationally) 30 Ibid., p. 3&1. » 31 Ibid., pp. 3113-11. .L 233 be done." He also regards it as a "tautology-sketch" since supplementing it with the detailed facts makes it into a tautology. That is, if we were given all the relevant facts, then "the con- ditional with all the facts in the antecedent and 'the rational "32 In an earlier man does A' as the consequence is a tautology. essay he characterized these substitutes for general laws in historical explanations as "truisms" or "normic statements," as a kind of logical hybrid having some universal and some statistical features. They alone are capable of providing the grounds fOr acceptable explanations of individual events, because of their crucial role as ”norm-defining" giving them a "selective immunity to apparent counter—examples." And the study of such statements he regards as the "logic of guarded generalizations,"33 since though they are not definitionally true, they still cannot be falsified in the relatively direct way that simple empirical generalizations can. Unlike statistical statements also, they are not restricted to saying about the things to which they refer that some do and some do not fall into a given category. Rather, normic statements say that "everything falls into a certain 3b category except those to which certain special conditions apply." Hence they claim a preferred status for some particular kind of 32 Ibid., p. 3&2. 33 Scriven, "Truisms as the Grounds for Historical Explana- tions," in Gardiner (ed.), Theories 2f History, pp. hob-S. 3h Ibid., p. L66. 239 behavior, since on theoretical grounds deviations from this behavior can be attributed to the operation of interference factors. Thus, in his example of eXplaining that William the Conqueror did not invade Scotland because he had no desire for the addi- tional lands of the Scottish nobles, the implicit generality and ground of the explanation would be some such normic truism as, "Rulers who are satisfied with what they have do not normally or usually invade neighboring territories.”35 By replacing Hempel's general laws, universal or statistical, by such normic generali- zations, Scriven claims to have captured the explanatory force of a law, yet without sacrificing the hold on the particular case. While not ruling out all cases to the contrary, unlike universal laws, they still inform us of more than that rulers sometimes or seldom act in this way, unlike statistical generalizations. By thus saying something weaker than one kind of law and stronger than another, because of representing the normal, proper or standard case, normic truisms purportedly overcome the basic weakness of statistical statements. They neither abandon the hold on the particular case, nor allow the particular case to rattle around inside a network of statistical laws, since it can now be located in the normic network. Such statements locate the particular action or event by telling us what had to happen in this specific case, unless of course certain exceptional 35 Ibid., pp. nun and b67. ?h0 circumstances obtained. In this way we can be sure that C, §.g., explains X even though we lack exact laws. The problem, however, as Scriven quickly recognizes, is just how to interpret the notion of "exceptional conditions," so that we can cogently distinguish between counterexamples which actually do falsify these normic truisms and those cases which are merely apparent counterexamples. Unless Scriven can clarify this point, his hybrid modification of the CL theory along with his criticism of Brandt will go the way of most empathy theories by being un- testable in principle. But here Dray's perceptive remark seems most appropriate: when pressed on this point, Scriven's normic generalizations are in danger of being assimilated either to Hempel's ordinary descrip- tive statistical laws or to Dray's normative principles of action.36 Scriven, of course, claims to have captured the advantages of each without their accompanying weaknesses. His support for this claim, nevertheless, remains unconvincing. And largely because he fails, along with the others we have considered, to locate prOperly the role of normative judgments and decisions in explanatory arguments. This failure leads him to support one version of SU by denying the cogency of the CL theory, and to consider history "the mother subject for explanations" on the grounds that historical explanations are logically no different from common-sense ones.37 36 Dray, "The Historical Explanation of Actions Reconsid- ered," p. 121. 37 Scriven, "Truisms as the Grounds for Historical Explanations," p. h62. L But to elicit the nature and import of this failure, we must show how his proposal can be assimilated either to that of Dray or of Hempel. Scriven's comments about the "norm—defining" role of normic truisms and their consequent immunity to any direct or straight- forward falsification lead one to suspect that they are, after all, merely what Dray called "principles of action." This sus- picion seems further supported by the fact that, following Dray's account, this normative element purportedly provides the explana- tory force of reason explanations. In fact, I think Dray correct in remarking that such truisms as, "Rulers who are satisfied with what they have do not normally invade neighboring territories," can be easily interpreted as norms reminding us of what is apprOpriate or reasonable to do in the given circumstances.38 If we simply substitute "appropriately" for "normally," we are reminded of what rulers will do, unless of course they are acting in unreasonable ways: arbitrarily, emotionally or ignorantly. In this way, the explanatory function of his normic statement would indeed be to provide the rationale of William's noninvasion of Scotland. Thus would Scriven's modification of the CL model be assimilated to Dray's rational model. Moreover, by stressing the difference between the way normic and statistical statements handle exceptional conditions, Scriven develops Gardiner's emphasis on the historian's need to make 38 Dray, "The Historical Explanation of Actions Recon- sidered," p. 122. ( \ i “ E . ‘ I. F - ‘ ‘8 _ ‘ l1 :7 .. . 1 ‘ ._I ‘ 1 \ ‘ - l, _ -. . -~ a;;.;-:~. 1h. 2&2 appraising judgments in applying his generalizations. Unlike Gardiner and Dray, however, Scriven recognizes the same use of judgment in applying physical laws, even though it plays a weightier role in history. Hence he refuses to concur with their rash claims about the disunity of science and history. He argues, §.g., that the physicist, as well as the historian, judges that his causal explanations are the right ones, and that this judgment is "unformalizable," aided by "empathy" and dependent on contex- tual considerations.39 Since the importance of normic truisms, for Scriven, lies in their knowability and utility even when all the special conditions on which their truth depends are impossible to state, they function as tactical rules in bridge or chess. "Second hand plays low," and "Deploy your pawns early," for instance, parallel "Power corrupts." And since such rules are clearly not exceptionless, the historian, as the skillful and experienced player, develops his trained judgment and capacity to handle these normic generalizations. He recognizes the network of exceptions, realizes the limits of their applicability and the degree of their acceptability, even when unable to formally state or articulate these matters. Yet, even granting Scriven the complex relationship between dispositions like rationality and their quasi-necessary manifesta- tions or consequences, the selective immunity to apparent counter- examples of explanatory generalizations, and the important role 39 Scriven, "Truisms as the Grounds for Historical Explanations," p. ho2. 2h3 of judgment and contextual considerations in applying these generalizations-~even so, it would still be feasible to assimi- late his normic truisms to statistical generalizations as required by Hempel's model (P). That is, it is not at all clear what logical novelty Scriven's truisms have which make their explana- tory force significantly different from that of statistical statements. Hempel's premise (C) will most often, of course, be such a nonuniversal statement as, "In a situation of type C, a rational agent will probably (more than likely, usually) do X," thus meeting the complaint that the original universal statement was too strong. In addition, it also counters Scriven's objection that the particular case X "rattles around" inside a network of statistical laws which loses hold of this case, by providing as equally direct relevance to the particular case as normic generalizations. But to do so involves invoking the "requirement of total evidence" as part of the statistical model (P). By thus requiring that the total evidence available be used as a basis for determining the acceptability or degree of confirmation of an explanatory hypo- thesis, Hempel provides a sound basis for assimilating normic statements to statistical generalizations. Not only are they comparable in explanatory force, but the latter, when supported by the requirement of total evidence, provide the basis for a more extensive understanding of the particular case. This satisfies Scriven's further requirement of "informative relevance," which he no longer has to misconstrue as a replacement for inferrability, 2hh since statistical generalizations also warrant the inference of the particular case.ho Further, if our argument is cogent, the residue of Scriven's objection to model (R) lies in its statistical provisions. In particular, his case rests finally on the claim that the needed methodological requirement of total evidence has "the drawback of being virtually unattainable,"lll and hence that Hempel's model r needs modification to include the element of judgment in order to be informatively relevant. This brings us, then, to the question: what role to assign to what Scriven correctly takes to be "the great truth in the Verstehen theories." That is, to their "recognition of the indispensability and efficiency of the his- torian's capacity to respond to the cues in a well-described situation, so that he may with justifiable confidence accept or propose a particular reasons-explanation as correct." As the next chapter will show, even Hempel concedes the inadequacy of the requirement of total evidence, and for a number of reasons. Thus it needs be replaced by some more manageable substitute, as well as to be reinforced by further requirements which explanations must satisfy if they are to qualify as ration- ally acceptable. But if so, then we must ask whether reinforcing the requirement of total evidence entails any serious logical or epistemological modification of the CL theory of explanation. hO Scriven, "New Issues in the Logic of Explanation," p. 357. "1 Ibid., p. 3u5. "2 Ibid., pp. 358-9. Scriven advises of course that it does. He modifies it by replacing the covering statistical or probabilistic laws with normic truisms, because only the latter cogently provide the informative relevance of explanations to the particular case. Ny contention, on the other hand, is that Hempel's theory can accom- plish this task without such logical modification. But it can do so only by conceding one of the major epistemological theses advocated by many recent Verstehen theorists, including Scriven, Dray, Gardiner and Lavine. It must surrender the value-neutrality thesis. In the next chapter we will argue not only that this modification would improve the CL theory, but that an adequate reconstruction of probabilistic explanations, as explicated by Hempel according to model (P), actually entails this epistemic thesis. The explainer must logically consider contextual and purposive matters, and make cost or value judgments, in order for his explanation to be rationally acceptable. This requires, as we attempt to show, interpreting scientific knowledge along the prag— matic lines suggested by Dray and Scriven but rejected by Hempel as presystematic. Their analysis, however, locates the pragmatic element in the generalization providing an argument with explanatory force (either principles of action or normic truisms), and hence supports SU by modifying model (B) accordingly. Instead, we take Hempel's model to be free of the purported logical difficulties. The major modi- fication needed concerns not the laws providing explanatory force but the epistemic requirement that the explanans-hypotheses be at 2116 least well-confirmed. lhis, we will claim, suffices to locate and account for the important pragmatic element of the historian's capacity to judge exceptions to laws and, more generally, to judge the acceptability of these laws or generalizations. Moreover, by locating the pragmatic value judgments in the role of appraising and selecting explanatory generalizations rather than in the generalizations themselves, we can avoid another r4 of Scriven's charges, which otherwise might prove embarrassing to E Hempel. It occurs as a dilemma which would, if cogent, "mark the “ end of any 'model.'" Either, Scriven argues, Hempel would treat the informality or looseness of many explanatory generalizations "as a sign of imperfection and reject them," or he would not. Now to reject them, I agree, would be to make a mistake analogous to the one Hempel seems originally to have made in taking the deduc- tive model to be more scientifically adequate and complete than the probabilistic model. But not to reject them is to make the probabilistic model "largely a matter of good, nonformalizable judgment about the 'weight of the evidence.'" This is not, as many suppose, a sign of unscientific explanation, but rather "a sign that much of scientific argument conforms more closely to legal and historical argument than it does to mathematical infer- ence."113 In either case the CL theory appears to be in trouble. The usual counter to this kind of charge, already noted in the replies to the early idealists and to Schutz and Dray, grants h3 Scriven, "Explanations, Predictions and Laws," in Minnesota Studies, Vol. III, pp. 227—8. 2b? the first horn of the dilemma. But it denies the second on the grounds that while such looseness entails pragmatic judgments about the weight of evidence, it does so only in some presyste- matic or psychological sense. Hence the logic of explanation remains free of such loose matters. In other words, CL theorists reply to such charges by the claim that the choice of premises or l their acceptability as explanans "is not in the strict sense a logical matter and does not affect the validity of a deduction."hh Here Hempel and Nagel concur with Brodbeck's counter to Scriven. Yet this seems an obvious but irrelevant response. Is it. not, in fact, a retreat to the first horn of the dilemma? For Brodbeck, yes; but not for either Hempel or Nagel. Both resist the temptation to assimilate all probabilistic explanations to the deductive model, but instead consider model (P) to be an indepen- dent ideal with peculiar epistemic requirements and difficulties of its own not encountered by the deductive model. One of these epistemic requirements of an adequate explanation concerns the very point at issue, which Brodbeck consequently ignores: the confirmation and acceptability of the premises in an adequate probabilistic explanation. It seems, then, that the force of the 'empathy' position, eSpecially as put by Scriven, has not been appreciated by CL theorists because of the overemphasis placed upon the logical requirements of the deductive model, and because of the little hh Brodbeck, 2p. cit., p. 2h2. 2h8 attention paid to the intricacies and difficulties of model (P). The important question remains: must the pragmatic elements of judgment and acceptability of explanatory hypotheses be included in an adequate reconstruction of scientific probabilistic explana- tions? Part of the point of Scriven's dilemma is to force an affirmative answer, with which I concur. However, I take issue with the conclusion he draws from this th sis, 31%., that to allow these pragmatic ingredients into statistical explanations is "to mark the end of any 'model.‘")‘45 For this would follow only if the pragmatic elements were unforma- lizable in principle and were part of the logical explanatory force of the model, §.g., if they required general laws to be re- placed by normic truisms or principles of action. Since, instead,. the judgments are more properly taken as epistemic weights for appraising the acceptability of such explanatory generalizations, the model remains logically intact, unscathed by the challenge. But to show why this is so requires us to turn, in the next chapter, to a more detailed consideration of model (P) than has heretofore been offered. In particular, we want to see what specific difficulties give rise to the requirement of total evi- dence, why it is not by itself sufficient as a criterion of evidential adequacy or rational credibility for probabilistic hypotheses and, most importantly, why and how pragmatic utilities are needed to supplement purely epistemic utilities as criteria b5 Scriven, "Explanations, Predictions and Laws," p. 228. 2L9 of such rational credibility. In sum, our thesis is that Hempel's CL theory survives the varied logical criticisms of 'empathy' theorists, but only on condition that the value-neutrality thesis be rejected, that pragmatic elements be included essentially in a logical reconstruction of probabilistic eXplanations, and that the structural emphasis on explanations be extended to include a purposive ingredient. If this case can be made successfully, the insistence of Dray and Scriven on the "irreducible pragmatic dimension to explanation" will have been supported against Hempel's claims to the contrary. Supported, of course, by redirecting or relocating the pragmatic dimension in such a way as to preserve the cogency of Hempel's logical model of explanation against their other charges. But supported nonetheless. Hence, Hempel's reply--that the pragmatic concept of explanation, however important, can claim only psychological or genetic priority over the theoretical CL ideal; and that the latter "is objective in the sense that its implications and its evidential support do not depend essentially on the individuals who happen to apply or to test them"§émust be viewed with skepticism. ‘Whether a given set of explanatory premises, containing nonvacuously occurring statistical hypo- theses, adequately explains a certain action or event to a given person will indeed depend partially on just such "subjective," hé Hempel, "Explanation and Prediction by Covering Laws," 8. Haumrin (ed.), Delaware Seminar i2 Philosophy pf Science, Vol. I (New fork: Wiley and Sons, 1963), p. 130. 250 p '1_:.|_ rposive and contextual features as interests, attitudes, judg- m ants and values. But it will depend on these features in a S :Irstematically epistemological, not just psychological, way. Still, whether the introduction of these pragmatic features m 31 Res the logic of explanation essentially subjective or non- -t.‘ O rmalizable in a vicious sense, as Hempel and Scriven seem to “£3 Olieve, is another and quite independent question. My suspicion, however, is that it does not. Yet this belief, that value con- Esti_chept of scientific explanation. They force a fundamental re- EBJ<1mnination of the very meaning of this concept along pragmatic and F>Ilgrposive lines. For this reason we turn, finally, to a more de- tCiiled analysis of Hempel's inductive systematization of r1<>rmological explanations. \ . h? N. Rescher, "Fundamental Problems in the Theory of SClJEIitlfiC Explanation," ibid., Vol. II, p. bl, CHAPTER VI THE PRACEIAI‘IC DIEEEI‘JSION OF EXPW'IATIJN MID THE VN THESIS The Probabilistic Model and VN In the preceding chapter we suggested that an adequate a nelysis of Hempel's extended version of the CL model, covering teleological as well as causal explanations, requires investigat- ing the use of the probabilistic model (P) as the more apprOpriate f‘orm for rational eXplanations. Only thus can we determine to What extent the CL theory can be defended against the SU thesis, and in what way it entails a denial of the VN thesis. That is, Our twofold contention--that empirical generalizations are neces- S ary to provide the inferential explanatory force of rational eXplanations, and that the epistemic condition (Rh') needs be eXpanded to include inductive criteria other than confirmatory Strengthurequires a defense provided only by a detailed analysis Of Hempel's recent statistical systematization of eXplanatory arguments. To this task we now turn. Since the CL theory interprets explanation as an inferential relationship, deductive or inductive, between a set of laws con- joined with antecedent conditions and the description of some event or action, the main difficulties for the historical use of model (P) turn on requirement (311'), that an acceptable eXplana- tion must contain hypotheses or theories with a high degree of confirmation. This requirement in turn raises most of the 251 p roblems of induction generally, especially the problem of provid- i 1’1 g adequate criteria of good evidence and reasonable belief. P‘or, unlike deductive inference, statistical or inductive infer- e nee contains an essential gap between explanans and explanandum. -[I 't; contains premises which evidence or support but do not prove th e conclusion. The latter, by outstripping or asserting more tT1an the premises, is not "contained in" them. Thus the ensuing (ii fficulties concern "filling in” or bridging the gap, finding g—Pood reasons to support the conclusion, recognizing the degrees o f reasonableness, and determining what degree of evidence suffices to warrant the rational acceptance of our explanations. It is generally agreed, I think, that the amount of evidence required to warrant any particular belief is not fixed but a function of our varied purposes. That sufficient evidence also Ci€epends on such pragmatic factors as the cost concerning the sig- 1"Eificance of making a mistake when acting on the belief in question, however, seems less generally recognized. This latter thesis bears heavily on historical explanation and the invidious comparisons Often made between the kinds of explanations employed in various disciplines. For it appears that these disciplines differ not in the logical kinds of explanations offered, as indicated by the SU thesis, but rather in the different emphasis placed on just this Pragmatic factor. . The amount and kind of evidence currently a‘I'I=~3...‘7.la‘ole in the given field, the number and quality of competing beliefs, and the importance of accepting a false belief or I‘ejecting a true one all seem central. I” .I .l d . . . \v ‘, . / .3" "“‘“"fi_fi'... “ aw- 253 The emphasis of most empathy theorists on subjective judgment p r operly concerns not the explanatory force of an argument but the a C ceptability of the explanans hypotheses. When these hypotheses a re statistical in nature, their acceptability requires a consid- e ration of pragmatic criteria and hence a denial of the VN thesis. Further, when they occur in fields such as history where the a-mount of data is often negligible and where plausible competing hypotheses abound, the requirements of acceptable evidence must be loosened considerably. In other words, the degree of rigidity of our requirements for confirming evidence depends on these very Cost factors, and hence tends to distinguish the various sciences. But this contention makes one important assumption that will have to be defended later: that the goal of science is not merely truth for its own sake but truth as modified by other criteria. {“163 will pursue this general contention in the context of certain a~Spects of model (P) alluded to in the last chapter, _v_i_E., the peculiar ambiguity of statistical inference, the requirement of tC>tal evidence and the criteria of rational credibility of S ta ti stical generalizations. As already noted, much criticism of the CL model as an expli- Cation of the logical structure of historical explanations turns on Hempel's early emphasis upon the deductive pattern and his Stress on covering laws as universal generalizations of the form "All x is y. " Most actual historical generalizations, notoriously, describe not invariant relations, as the deductive model requires, but, at best only correlations of varying degrees of frequency. 25h T hus historians usually feel unaffected by the presentation of n egative evidence or counterexamples to their generalizations, m aking refutation of their hypotheses extremely difficult. But 813.011 a situation suggests to some philosophers of history a p eculiar paradox for the CL deductive model. Scriven, for :3_ nstance, notes the apparent inconsistency between the existence 0 f good explanations in historical practice and the nonexistence O 33‘ comparably good laws and predictions. For, "some historical eXplanations appear to be so well supported by the evidence that we cannot reasonably doubt them." Yet the CL theory postulates, a S a necessary condition for an explanation to be beyond reason- able doubt, that we have general laws which are also beyond doubt Eind license the deduction. But this condition is not met since " historians are not in possession of such general laws." The paradox of the CL model, then, lies in the fact that good explana- tions are explicated as deriving logically from dubious general laws, from laws in which we have less confidence than the GX‘planation they are used to support. No doubt such a situation is largely responsible for the Various moderate criticisms of the CL model, some of which were reviewed in the last chapter. But it is important to note again that, not all CL theorists cling to the deductive model and its I'equirement of universal laws as the explanatory force of histori- cal explanations. Hempel and Nagel, in particular, recognize in l _ M. Scriven, "Truisms as the Grounds for Historical EXP—Lanations," pp. th-h. 255 EE;.,ZZL_CL of their writings on this subject the need for loosening the -7—r71 (:pciel to allow generalizations of a statistical nature and an :fi__ tfixiuctive relationship between explanans and explanandum. And IE4}? GZBUEKfld in recent writings, contributes a logical analysis of ‘t:gjkvrese matters by his model (P) which, you will recall, has equal tESL‘tZJELUS with model (D) as an irreducibly complete idealization. The requirement of general laws is loosened to include those <:>;i? the form "Almost all instances of x are instances of y" or more };>:r‘ecisely, "The probability of an x being a y‘is r," i.e., 3;) Cry, x) = r. Such a view clearly dulls the edge of the apparent paradox. :3: 1; accounts for the historian's reluctance to reject his general- fi.aaations on the basis of a few negative instances, as well as for illne difficulties in refuting such hypotheses. Statistical laws Eizre much less dubious than universal laws, since not claiming "Llriexceptionable or invariant relations but only correlations E1:Llowing of many exceptions. Yet for this very reason they raise Siearious problems concerning the epistemological notion of reject- irigor accepting an hypothesis. Moreover, this extension of the CL theory to include model (P) ESEBevs too great a concession to many who were attracted to the 1c the particular case. The fundamental charge leveled against —t:;,1’113 probabilistic model rests on the fact that statistical laws, ‘1L;1lzr)LLike deterministic ones, are compatible with both the occurrence :EEL Irnri nonoccurrence of their eXplanandum-event. They fail to rule <::>1L25t the nonoccurrence of e, and hence fail to explain why e, in ‘1:>.z:1riicular, occurred rather than something else. For this reason EES.<::riven prOposed his novel normic truisms as a replacement for ‘t;}f1e laws of the CL model. As a kind of logical hybrid they pur- ‘3:><:yrtcdly possess the important advantages of both universal and =E:1batistical laws, along with those of Dray's rational principles 0 :5 action. Our immediate task is to review this aspect of the situation ’Ivzegarding statistical laws as explanatory hypotheses. We must ESeee how and why they lose their hold on the particular explanandum- Eaxient by their compatibility with both e and non-e, and also Ciéetermine what additional requirements need be imposed on model (P) 1Lc> eliminate this clearly objectionable feature. 'We will then be j~r1 a better position to examine the value-neutrality thesis. 257 Inductive Ambiguity and Total Evidence Unlike the nomological statements adduced in the explanans <::>ij? a deductive explanation, statistical nomological statements «532::cpress correlations between certain attributes or properties as £ZEL specified long-range frequency. Such statements assert that if <::.£ertain specified conditions are realized, say A, then an occur— ::“’<3nce of a given kind, say B, will come about with a certain long- f17~14n relative frequency, r. If, in other words, A and B are EEL‘ttributes, then such a statement will take the following form: “‘irhe probability for an instance of A to be an instance of B is r," <:>3r symbolically, "p(B, A) = r." For example, the probability of ‘tslie toss of a fair coin being heads is l/2. It is also to be 1i1<3ted that the probability r refers not to the class of all actual Zi.Iistances of B, 3.3., to a finite class, but instead to the class <3:f all potential instances of B. That is, the probability state- inieent ascribes a certain disposition to the coin, 323., that of 3ffiielding a head in one out of two tosses in the long run, not just jar) all the actual tosses of the coin. Now it might appear as if the CL models, (D) and (P), are €33<£1ctly parallel except for the different kinds of laws used in ‘tliee two cases, strictly universal and statistical laws. The fOlilowing two arguments, for instance, appear similar in this way. (D1) (a) All patients who suffer from a virus infection and are treated with penicillin are helped by penicillin. (b) Patient x suffers from a virus infection and was treated with penicillin. (0) Therefore, x will be helped. ."l r l .' . \/ .I \. K .7 - . V :l F I ‘r. " -~ _ ,_ “I! x' 258 (P1) (a') The probability of being helped by penicillin when suffering from a virus infection is .9. (b) Patient x suffers from a virus infection and was treated with penicillin. (c) Therefore, x will be helped. As Scriven and others point out, the truth of (a) and (b) is ::i_.1ricompatible with the nonoccurrence of what is described by (c), 53;:i'nce argument (D1) is deductive. And by so ruling out the non- <:>1c:currcnce of what is described by (c), (a) keeps its hold on this <:><3currence and hence explains why it in particular occurred rather ‘tilian something else. But the truth of (a'), on the other hand, ii :3 quite compatible with the falsity of (c), since there may be Eaanother reference class, say A, to which x belongs which makes it }1:ighly probable that x will not be helped by the penicillin. If, .f‘or'example, x was allergic to penicillin and if the probability <>;f being helped by penicillin in such circumstances was extremely :1_ow3 we would obviously not expect x to be helped. In such a czaase, the argument would have the following form. (P2) (a") The probability of a patient who is allergic to penicillin being helped by penicillin is .l. (b') Patient x was allergic to penicillin. (ac) Therefore, x will not be helped by penicillin. Pieance, since (a') is compatible with both (c) and Que), it clearly :L<>ses its hold on the particular case and fails to explain why x VVaAS in fact helped by the penicillin. This merely notes that the iJnIIrobable may be actual. The peculiar phenomenon just illustrated, whereby model (P) allows of interpretations (Pl) and (P2), both of which contain true premises yet the conclusions of which are inconsistent, Hempel labels the "inconsistency” or "ambiguity of statistical explanation."2 Generally, for any argument of the form (P) with true premises, there is a competing argument of the same form also with true premises but whose conclusion is incompatible with that of the first argument. The ambiguity in question lies, of course, in the different reference classes to which our patient was assigned, and relative to which he was assigned inconsistent properties. Furthermore, this difficulty seems to be absent from model (B) or deductive explanations. For incompatible conclusions, as (c) and Gvc), can be derived only from incompatible premise-sets, and true premise-sets containing strictly universal laws can never be incompatible. That is, for model (D) to allow of interpreta- tions parallel to (P1) and (P2), at least one of the universal laws, either (a) or its competitor, would have to be false. Hence, the choice of premise sets in this case is obvious: choose the true explanans and thereby eliminate the problem. But in the case of (P1) and (P2) the choice is not at all obvious, since both sets of premises are true. The problem and ambiguity thus remain. Little wonder then that Scriven and others find Hempel's model (P) suspect on the grounds that statistical laws lose their 2 C. Hempel, "Deductive-Nomological vs. Statistical Explana- tions," p. 127. Cf. also his "Inductive Inconsistencies," pp. 128-132; along with S. Barker, Induction and Hypothesis (Ithaca, N.Y.: Cornell University Press, 1957), pp. 75-8. 260 hold on the particular explanandum-event, even though such laws are widely invoked for scientific explanatory and predictive purposes. The trouble might be thought to stem from the attempt to apply probability statements to individual events or persons, as with our patient x. For such an application is thought by some to have no meaning at all. Professor Brodbeck, following C. S. Peirce and more recently von mises, suggests that From a deterministic law, given the initial conditions, we can predict an individual event. From a statistical law and its initial conditions...we can predict only a so-called mass event, that is, the frequency with which an attribute will be distributed in the given class.... From a statistical law, then, nothing can 3e predicted [nor explained7 about an individual event. The statistical frequency of some property B in a reference class so restricted as to contain x as its only member would of course be of no use in judging the hypothetical probability that x had B. But it does not follow that the latter judgment cannot be made according to a more appropriate reference class. Horeover, as Hempel and others have argued, “there is only a difference in degree between a sample consisting of just one case and a sample consisting of many cases. And, indeed, the problem of ambiguity recurs when probability statements are used to account for the frequency with which a Specified kind G of result occurs in 3 M. Brodbeck, "EXplanation, Prediction and 'Imperfect' Knowledge," pp. 2b7-8. uhlu‘h 261 h finite samples, no matter how large." The troublesome ambiguity seems to arise then not from apply- ing probability statements to individual cases, but instead from two different sources. One such source consists in the misleading assimilation of such arguments as (P1) and (P2) to deductive ones as (01), on the basis of the analogous construal of strictly universal and statistical laws. For such an assimilation tends to make us overlook the important fact that both 'probability' and 'certainty' are relative terms. That is, the conclusion (0) is certain only relative to the premises of (D1) not in itself, and it is likewise highly probable (inductively) only relative to those of (P1) and improbable only to those of (P2), not in itself. In each case (0) is merely an elliptical formulation of a rela- tional statement. Once this is seen, the impression that statis- tical eXplanations or arguments warrant, on the basis of true premises, the acceptance of such incompatible conclusions as (c) and (we) vanishes. This impression trades on interpreting the incompatible conclusions as nonrelational. The incompatibility or ambiguity dissolves since (c) is warranted by a different set of premises, those of (P1), than those warranting Gvc), those of (P2). Nevertheless, even when clarifying this confusion, even when the apparent incompatibility is seen to result from relating the h Hempel, "Deductive-Nomological vs. Statistical EXplana- tion," p. 132. Cf. also W. Rozebloom, "Comments" in Current Issues lip_the Philosophy pf Scienqg, pp. 237-h1; and A. Pap, An Introduc- tion tg.thg Philosthy'gf Science, pp. 186-9. __ 262 conclusions to different premises, these premises remain true. Hence the problem now changes to choosing between them. 'He empha- sized in Chapter II that a rationally acceptable eXplanation, deductive or probabilistic, must be one whose explanans warrants the belief that the eXplanandum-event did occur. In the case of deductive eXplanations no such problem emerges, since if incom- patible conclusions are warranted by two different premise-sets, then at least one must be false and our decision obviously ought to be for the true set, if any. But, since in prdbabilistic eXplanations both sets of warranting premise—sets can be true, which of the alternative sets to choose as a rationally acceptable scientific eXplanation and prediction becomes a critical matter. In such a case it is clear, as Hempel and Oppenheim recognize, that the four conditions of adequacy laid down by them are surely not sufficient conditions for an eXplanation to be rationally acceptable. Indeed, two sets of premises might easily meet condi- tions (Rl'), (R2), (R3) and even (Rh) (i,g., contain empirically testable and true statistical laws which inductively imply their respective conclusions), but yet imply incompatible conclusions, in which case at least one such set must be ruled rationally unacceptable. Hence, Hempel's distinction between potential and true or well-confirmed eXplanations, discussed in Chapter II, clearly does not suffice to mark the difference between a merely genuine and a rationally acceptable eXplanation. How precisely to mark this distinction, however, remains one of the most baffling and perplexing of epistemological prdblems. Q63 Nevertheless, this problem of Specifying criteria for ration- ally acceptable eXplanations seems, contrary to advocates of the SU thesis, in no way to affect the first three conditions. A serious problem for scientific explanations as well as for histori- cal or social ones, it concerns basically (Rh')' Since we can never be assured of the truth of statistical laws, the question of importance is how, given two conflicting sets of empirically test- able premises, to decide what credence to give to each set. How much evidence is required or sufficient to warrant a set of premises as a rationally acceptable eXplanation? And can such a decision be made adequately within the confines of the value- neutrality thesis? More precisely, would an adequate explication of the notion of a rationally acceptable scientific or historical eXplanation, and hence of such a decision to accept or reject eXplanatory hypotheses, require the inquirer to make value judgments? To make a start in this direction we note first what is by now an obvious yet still important additional criterion adduced by numerous philOSOphers: the requirement of total evidence. Such a criterion helps to decide our residual problem concerning the ambiguity or inconsistency of statistical eXplanations. For, even when given the premises of (P1) and (P2) as true, neither would be adjudged acceptable without considering further relevant evidence. Otherwise, we run the risk of selecting only those true or well- confirmed statements as explanatory hypotheses which favor one's biases or desired conclusions. 26b To control such bias Carnap, among others, proposes that tin the application of inductive logic to a given knowledge situation, the total evidence available must be taken as basis for determining the degree of confirmation."5 Hempel adds the modification that a smaller part, e1, of the total evidence may be used if the remaining part, e2, of the total evidence is inductively irrelevant to the explanatory hypothesis h whose confirmation is to be deter- mined. Using Carnap's notion of inductive prdbability, "the irrelevance of e2 for h relative to e1 can be eXpressed by the condition that c(h, e1 . e2) = c(h, e1)."6 Hence, our residual problem dissolves since, g.g., (P1) and (P2) cannot both satisfy the requirement of total evidence. For the total evidence, as one body of consistent evidence, cannot confer high probabilities on two contradictory statements, as (c) and (cc), since the two probabilities add up to 1. But just this would be the consequence if both (P1) and (P2) satisfied the requirement.7 At this juncture it will be helpful to consider the nature of this requirement of total evidence, its epistemological status and also the kind of support that can be adduced for it. Especially instructive is the criticism of this principle recently leveled by A. J. Ayer as part of an attack against Carnap's notion of inductive S R. Carnap, Logical Foundations pf Probability (Chicago: University of Chicago Press, 1950), p. 211. Cf. also A. Pap, pp. cit., pp. 187-9; R. Chisholm, Perceiving; A Philosophical Study (Ithaca, N.Y.: Cornell University Press, 1957), pp. 2§;27; and S. Barker, pp, 313,, pp. 76-78. 6 Hempel, "Deductive—Nomological vs. Statistical Explana- tion," p. 138. 7 Hempel, "Inductive Inconsistencies," p. lhl. , 'i. .zAf.‘ 265 or logical probability. But for our purposes we will limit the discussion to Ayer's challenging question "Why have we to take as 8 evidence the total evidence available to us...?"’ Ayer intends not to deny the legitimacy of the principle but only to deny its justifiability on the basis of Carnap's principles of inductive logic, and hence to question the relevance of logical probability as the "guide of life" or as a basis for action. Now in Chapter II, we referred to Carnap's distinction between internal and external questions, between questions which occur within a linguistic framework and presuppose the framework or categorial principles and questions about the acceptability of this categorial framework. Put into this context, Ayer seems to imply that a categorial principle, gig}, the principle of total evidence, cannot be justified within the system. And indeed he is correct. But the import of his implication is negligible, since based on a confusion of an external question with an internal one. For clearly the principle of total evidence constitutes part of the framework of inductive probability, and as such, in Carnap's words, "is not a rule of inductive logic, but of the methodology of induction."9 Hempel further clarifies the external nature of the principle by taking it as "a partial explication of the condi- tions governing rational belief and rational choice." Such an 8 A. J. Ayer, "The Conception of Probability as a Logical Relation," in E. Madden (ed.), The Structure pf Scientific Thought (Boston: Houghton, Mifflin, 19505, p. 231. 9 Carnap, Logical Foundations 9}; Probabilitv, p. 211. explication "Specifies a necessary, though not sufficient, condi- tion for the rationality of inductive beliefs and decisions." Its acceptance is, as noted earlier, "pragmatic in character and cannot be defined in terms of the concepts of formal (deductive or inductive) logical theory."10 Notice in particular that if CL theorists were merely explicating the notion of formally sound potential or genuine explanations, the principle of total evidence would be unnecessary, along with the pragmatic dimension of eXplanation. But once they open the question to rationally acceptable eXplanations, to the acceptability of statistical explanatory hypotheses, then the principle of total evidence or some substitute must be invoked. With it comes a serious challenge to the value-neutrality thesis. For an adequate eXplication must now include an analysis of the acceptability of explanatory hypotheses, hence cannot exclude the pragmatic dimension as merely psychological or presystematic or merely extra-logical. Such a dimension is of course not purely formal. But this only concedes that the required eXplication is not of a purely formal notion. The explicandum, 'scientific eXplanation,‘ is essentially a pragmatic and methodological as well as a logical term. Hence any adequate explicatum must account for, not merely explain away, the fact that some extra- logical dictate of purpose or interest is required to direct our scientific acceptance or rejection of hypotheses. lO Hempel, "Inductive Inconsistencies," pp. lh2-3. 267 Moreover, it is but a step, though perhaps a long and diffi- cult one, from the inclusion of pragmatic elements in an adequate explication of rationally acceptable explanation to the denial of the value-neutrality thesis as an essential part of such an eXpli- cation. An indication of things to come can be gleaned in C. I. Lewis' version of the principle of total evidence. No inductive conclusion is well taken and justly credible unless the obligation to muster all the given and avail- able evidence which is relevant to this conclusion has been met....Indeed this principle of the required complete- ness of available and relevant evidence for the justified credibility of inductive conclusions, has a character which is plainly akin to the moral. It is unlike the textbook rules...and has instead the character of a maxim.11 11 C. I. Lewis, The Ground and Nature of the flight (New York: Columbia University Press, 19337, pp. 32—33. Rational Credibilitv and Utilities Lest we slide too quickly to our conclusion without the proper support, however. let me hark back to Scriven's objection to the requirement of total evidence, mentioned in the last chapter. His criticism rests with the claim that the principle has "the drawback of being virtually unattainable." From this claim he infers correctly that the CL theory needs modification to include the pragmatic element of judgment. With this claim and inference we can but concur. But we shall argue that such modifi- cation necessitates only the denial of the value-neutrality thesis, not, as he further concludes, a revamping of (R2) by replacing empirical covering laws with normic truisms or principles of action. Now, Scriven's objection to the principle of total evidence is conceded by Hempel, Carnap, Barker and others. It points to the unavailability, at least at present, of an apprOpriate general system of inductive logic whose rules would enable us to show that the part, eg, of our total evidence going beyond our premises is inductively irrelevant to our conclusion. Barker,12 in particular, shows that all attempts to utilize the principle in appraising the rational acceptability of statistical arguments like (P1) and (P2) meet with serious difficulties. More generally, even if we adopted any straightforward principle of induction by simple enumeration, we would still be plagued by inductive inconsistencies, since we 12 S. Barker, Induction and hypothesis, pp. 76-78. 269 possess no clear criterion for deciding upon the most appropriate reference class. Thus Scriven seems to be correct in charging that the principle is "virtually unattainable," and also in con- cluding that the CL theory needs a more manageable substitute for the principle, and reinforcement by additional conditions for rationally acceptable explanations. Any such substitute and reinforcement should of course be applicable to simple statistical arguments like (P1) and (P2). Hempel has recently proposed a rough substitute for the principle of total evidence as a criterion of evidential adequacy which, though it avoids many of the embarrassments created by the princi- ple, must be used with caution, qualification and discretion. His modified version requires two conditions to be met in order for statistical systematizations containing, say, the premises 'Fb' and 'p(G, F) a r' to meet the principle of total evidence. (i) the total evidence e contains (i.e., eXplicitly states or deductively implies) these two premises; (ii) e implies that F is a subclass of any class F* for which e contains the statement Fxb and in addi- tion a statistical law (which must not be simply a theorem of formal probability theory) stating the value of the probability p(GF*) However, even if this admittedly rough criterion were an adequate substitute for the requirement of total evidence, other conditions would still have to be satisfied by a rationally acceptable explanation. Such reinforcement also accords with Scriven's criticism of the principle. But the central question 13 Hempel, "Deductive-Nomological vs. Statistical Explanation," pp. IDS-9. 270 here is whether such reinforcement entails any serious logical or epistemological modifications of the CL theory of explanation. Hempel argues for a negative answer; Scriven for an affirmative one. So, having reviewed Scriven's unsuccessful defense in the last chapter, we turn now to examine Hempel's case. We want to see in particular the grounds for the CL defense of the value- neutrality thesis. he also want to determine whether covering statistical laws can provide the informative relevance of explana- tions to the particular case, without conceding any major point of the SU thesis. What other conditions, then, besides the requirement of total evidence must be satisfied by a rationally acceptable explanation? And do such residual conditions merely determine the degree of confirmation of explanatory hypotheses? Or do they, instead, authorize the provisional acceptance of such hypotheses on a given body of evidence which supports but does not logically prove them, where the acceptance of an hypothesis is taken as a case of a purely epistemic or theoretical choice between competing scientific hypotheses? Or, finally, do they authorize acceptance of such hypotheses in a sense which amounts to the adOption of a certain course of action? The grounds for the latter alternative might be that accepting an hypothesis in an open-ended situation, where no course of action is Specified, makes no clear sense, and hence must be viewed as a pragmatic notion. If either of the first two alternatives can be successfully defended, the value-neutrality thesis might survive. But if the 271 third alternative is cogent, as we hOpe to show, it argues strongly for the view that an adequate explication of the CL theory entails a denial of the thesis. For on this alternative other conditions than evidential strength are necessary for an explanation to be rationally acceptable. Let us look then at Hempel's most recent analysis of this situation, and specifically at his account of criteria of rational credibility and rules for the inductive acceptance of explanatory hypotheses. Following Carnap, he takes our question to concern the method- ology of induction, "the application of inductive logic to the formation of rational belief."1h In so doing he interprets scien- tific knowledge along the lines of an "accepted-information model" or schematization, and hence establishes rules of acceptance or rejection regulating membership in the body of scientific knowledge at any given time. Taking K to represent the total body of scientific statements accepted as true by scientists at a given time (whether or not these statements are true), he proposes three general necessary conditions or rules of inferential acceptance regulating membership in K: (CR1) Any logical consequence of a set of accepted statements is likewise an accepted statement; or, K contains all logical consequences of any of its subclasses. (CR2) The set K of accepted statements is logically consistent. 1h Hempel, "Inductive Inconsistencies," p. 151. 272 (CR3) The inferential acceptance of any statement L into K is decided on by reference to the total system K (or by reference to a subset K' of it whose comple- ment is irrelevant to L relative to K'). More specific rules are then established by rendering the problem as a Special case of formulating rules for rational deci- sion between several alternatives. It becomes a case of theoreti- cal choice between accepting a new hypothesis h into K, rejecting it in the sense of accepting -h, or suspending judgment by accepting neither h nor -h. The problem of rational choice, of specifying rules of decision, is posed in the following schematic fashion: An a5 ;ent X has to choose one out of n courses of action, Al, A9,..., An, which, on his total evidence e, lorically exclude each other and jointly exhaust all the possibili- ties open to him. The agent contemplates a set 01, 02,..., Om of different possible 'outcomes' which, on e, are mutually exclusive and jointly exhaustive....Then for any one of those actions, say A-, and any one of those outcomes, say Ok, the given system of inductive logic demands a probability for the hypothesis that, given e, Aj will lead to the outcome Ok- Indeed, if aj and ok are statements describing A- and 0k, r: 1zpectively, that probability is given by C(Ok, e - aj).1 Nevertheless, Hempel and Carnap clearly recognize the inadequacy of the system of inductive logic to determine a rational course of action for X. Rationality, as a relative concept, depends on the agent's goals or objectives, on the value or utility he attaches to the outcomes which might result from his action. Carnap therefore assumes that these values can be 15 Hempel, "Deductive-Nomological vs. Statistical Explana- tion, “ pp. 150-1. 16 . . . Hempel, "Inductive InconSistenCies," p. 152. 273 represented by a quantitative notion of utility. The value of each outcome 01, 02,..., On for X is assigned a real number, say uk, as the utility of, say 0k: for X at the time in question. Carnap then proposes a general decision rule to determine which of the available courses of action is rational to choose in the circum- stances: the action maximizing or offering him the highest expectation value of the utility attached to action Aj for X: 17 "u' (Aj, e) - C(ol, e . aj) . ul +...+ C(om, e . aj) . um.” Applying this schema and maxim to the problem of establishing acceptance rules for scientific explanatory hypotheses, Hempel views the decision to accept, reject or suSpend judgment on an hypothesis as a special kind of scientific choice. Such choice has three possible outcomes: K enlarged by the contemplated hypothesis h; K enlarged by the contradictory of h; K unchanged. The problem, accordingly, is which scientific hypothesis to accept and thus add to the body of scientific knowledge. In particular, the issue concerns what utilities to assign to these outcomes and, more importantly, what kinds of values the assigned utilities are to represent. Are they to concern merely the quest of truth for its own sake, of accepting as new information into K only true statements; or are such other aspects of science as simplicity and explanatory power also to be considered? Or are they to concern, in addition to these purely scientific or epistemic utilities, other cost factors and pragmatic gains and losses, i.e., 17 Ibid., p. 153; cf. also Carnap, Logical Foundations 9: Probability, p. 269. 27h utilities resulting from acting on the basis of the hypotheses in question? We will pursue these questions in detail in the remaining sections of this chapter. Regardless of the outcome of this discussion, one fact of importance emerges already. Once the notion of explanation is expanded to include the distinction between genuine and rationally acceptable explanations, as CL theorists do, the entire theory of explanatory inference is left in a highly unsatisfactory and incomplete state. That is, unless we can formulate conditions under which an explanation satisfies certain purposes and does so with maximum efficiency. Such conditions seem, generally, to cast a peculiar light on the degree of evidential strength or confirmation required to make an explanatory hypothesis acceptable. For to make such an appraisal, we would have to know both how efficiently the hypothesis satis- fied our goals or purposes and also how important our varied goals or objectives are. But in this last consideration lies what I take to be the essential defect of the value-neutrality thesis, as advocated by deber and CL theorists. In order to elicit the issue more clearly, let me consider in detail the kind of argument I find most compelling as an attack on the thesis in question. And in so doing, it will be necessary to contrast it with some weaker argu- ments used to derive the same conclusion, the denial of VN. But Haber and CL theorists successfully counter these weaker arguments. For they usually locate the value element in the context of dis- covering or imaginatively constructing plausible explanatory hypotheses. Since the VN thesis concerns only the justification, corroboration, confirmation or rational acceptability of explana- tory hypotheses, these usual arguments miss their mark. Accordingly, any argument successfully opposing the VN thesis must locate the value element in the very logic of explanation. Let us look then at one such recent argument. . A “.5~“\.I. , 276 Explanation and Value-Neutrality The value-neutrality thesis, you will recall from the first chapter, maintains that the scientist's value scheme be logically divorced from scientific standards of explanatory validity and reasonable or warranted belief. It requires scientific inquiry to be objective in the sense that the scientist remain evaluatively neutral when appraising the acceptability or rational correctness of his explanations, and also in the sense of yielding the same conclusion for all competent inquirers, independently of variable personal interests, attitudes or values. Now, to oppose this thesis, it must be shown that two scientific inquirers with the same evidence and the same probability assignments might nonethe- less disagree about the acceptability of a given explanatory hypothesis, and do so on rational grounds. Only in this way will one's values emerge as a basis for logically appraising the evi- dence, rather than merely for choosing problems to investigate or for even having a science at all. Value decisions in the latter roles can and have been, by CL theorists, relegated to the psychology or sociology of science, hence keeping science proper untainted, i.e., objective and value-free. In other words, it must be shown that the scientist 933 scientist makes value decisions and judgments, if we are to challenge the tenability of the VN thesis. Consequently, the case seems best argued on the basis of recent work in statistical analysis concerned with the problem of rational decision-making in the face of uncertainty. This type of R) \1 rd argument has been advocated by such experimentalist philosophers 18 It builds on the 19 as Fraithwaite, Churchman, Frank and Rudner. earlier statistical work of Neyman, Savage and Wald, especially in application to problems of quality control and more generally to problems of nondeductive inferences. These innovations in the logic of statistical inference-~based upon experimentally control- lable teleological concepts and hence capable of supplying a model broad enoush to confront the issue of a theory of experimental action-~seem a much more profitable point of departure for method- ology than the largely speculative position of Hume and J. S. in dll. Applications of statistical procedures are to be found, of course, in physics, genetics and the social sciences where a major goal is to decide what to believe. But they are also used in determining insurance rates and in market research where the goal is to determine how to act. Hence, the basic problem of statistics, rational decision-making in the face of uncertainty, suggests a structural similarity between these different kinds of problems: how to act and what to believe in the face of uncertainty. Accordingly, R. Rudner, summarizing much of this work, formulates 18 R. Braithwaite, Scientific Explanat.ion, Chapter VII- C. W. Churchman, Theory of Experimental Inference, Chapters lh-lS; C. W. Churchman, "Statistics, Pragmatics, Induction; “ P. Frank, Philosophy pf Science, Chapter 15; R. Rudner, nThe Scientist Qua Scientist Makes Value Judgments." 19 J. Neyman, A First Course in Probabilitv and Statistics (New York: Holt and_ Co., 1950 0;; .Savage, The Foundations of Statistics (New York: wilev and Sons, 19Sb); A. Wald, On the Principles of Statistical Inference (South Bend, Ind., Notre Dame University Press, 19h2f: 278 in.the following passage what I take to be the most explicit and compelling attack against the value—neutrality thesis. How I take it that no analysis of what constitutes the method of science would be satisfactory unless it com- prised some assertion to the effect that the scientist as scientist accepts or rejects hypotheses. But if this is so then clearlv the scientist as scientist does make value judgments. For, since no scientific hypothesis is ever completely verified, in accepting a hypothesis the scientist must make the decision that the evidence is sufficiently strong or that the probability is sufficiently high to warrant the acceptance of the hypothesis. -Obviouslv our dec:sion regarding the evidence and respecting how stronC is 'strong enough,‘ is going to be a function of the importance, in the typically ethical sense, of making a mistake in accepting or rejecting the _.____:_._.‘___..___.___e__ 20 hypothesis will depend on how seri.ous a mistake would be. But since Rudner‘s case is incomplete as it stands and has subsequently been subjected to sustained counterattack in the . . 21 a i , recent literature by I. LeV1, a Spokesman 01 the hempel-Carnap view, it will perhaps be instructive to unpack the argument in some detail. We will then be in a better position to assess Hempel's and Levi's defense of a value-free science by locating the issues more clearly. In the above-quoted passage four statements are explicitly noted as premises from which the conclusion, that the scientist nna scientist must make value judgments, is claimed to follow. Let us list these as follows: 20 Rudner, 9p. cit., p. 2. 21 I. Levi, "Eus+ c the Scientist Make Value Judgments?," Journal of Philosophy, Vol. LVII (1960); ”Decision Theory and Confirme.tion," Journal 9: Philosophy, Vol. VIII (1961); and ”On the Seriousness of Mistakes," Philosophy pf Science, Vol. XXIX (1962). 279 (l) "The scientist as scientist accepts or rejects hypotheses." (2) "No scientific hypothesis is ever completely verified." (3) Therefore, "the scientist must make the decision that the evidence is sufficientlv strong or that the proba— bility is sufficiently high to warrant the acceptance of the hypothesis." (h) The "decision regarding the evidence and respecting how strong is 'stron3 enough' is going to be a function of the importance, in the typically ethical sense, of mak- ing a mistake in accepting or rejecting the hypothesis." (5) Therefore, "the scientist as scientist does make value judgments." Now to some elaboration. Premise (2) states a central tenet of empiricism to which, as we saw in Chapter II, both Popper and Hempel, along with all other CL theorists, subscribe. In accord with Hempel's requirement (R3), it is but one version of the fallibilist claim that all scientific explanatory hypotheses are empirical and hence corrigible. No such hypothesis is ever without risk or ever completely Confinned by any amount of evidence, but can instead at best be rendered more or less probable. With (2) undisputed by CL theorists, the acceptability of (3) depends only upon that of (1). Moreover, if corrigibility or the chance of making an error were the only relevant consideration for the decision required by (3), the scientist would never reach any decision. He would merely keep increasing the amount of evidence indefinitely, i.e., until per impossible by (2) he attained certainty, before deciding to accept or reject the hypothesis. But if (1) is true, if the scientist must decide in the face of uncertainty, then some ECO additional factor must be considered in order for a decision to be made. And since it seems, at least pzipp fgpig, unreasonable to decide arbitrarily on the basis of sheer convention to accept only hypotheses with, say, more than 0.? degrees of probability, the statistician's suggestion that the additional factor be some measure of utility gains support. Since Hempel and Carnap concede this point, as indicated in the last section, we can for our purposes take it as established that the decision required by (3) is a function of some kind of utility. That is, the scientist must make a utility judgment or decision, granted of course that (1) holds. How then to defend (l)? The issue here turns on how to interpret the function of the scientist. For to defend the value- neutrality thesis, i.§., to deny (5), by rejecting (1), that the scientist accepts or rejects hypotheses, commits one to the "guidance-counselor" View of the scientist. This in fact is the view advocated by both Carnap and Hempel.22 Instead of accepting or rejecting hypotheses, the scientist, on this version, simply assigns degrees of confirmation to them relative to the total available evidence, and hence serves only as an advisor to policy makers who might want to apply such information to practical affairs. Carnap even defines logical or inductive inference in just this way, so that all probabilistic explanatory inferences 22 Carnap, Logical Foundations pf Probability, pp. 2hl-270; Hempel, "Review" of Churchman's Theory of Experimental Inference, in Journal 93 Philosophy, Vol. XLVI (1939‘), p. Soc. 281 in science consist not in an attempt to replace doubt by true or reasonable belief, but in assianina degrees of confirmation to hypotheses relatiVe to the given evidence. Hence, if this analysis of the scientific task is convincing, the objection to (l), and thus to (u), concedes that the scientist SEE policy maker must make the decision required by (3), yet insists that the scientist _gua scientist, i,§., SEE guidance-counselor, merely determines the degree of probability for an hypothesis. The value-neutrality thesis would thereby be defended at the price of giving up (1). Determining how costly this price is and whether we would be wise to pay such a price become the immediate problemS. No doubt the issue as to the aim and function of science is embarrassingly complex and difficult. At the present time we have no adequate answer. However, two considerations seem to argue against the Carnap-Hempel denial of (1). First, Rudner suggests that even the assignment of probabilities to an hypothesis is in effect the acceptance of an hypothesis, albeit of a different one. That is, "the determination that the degree of confirmation is say, p,...which is on this view being held to be the indispensable task of the scientist 313 scientist, is clearly nothing more than the acceptance by the scientist of the hypothesis that the degree of confidence is p...."23 30, even if the task of the scientist is to assign degrees of confirmation, this requires him to accept some hypothesis and hence to make the kind of decision involved in (3). 3 Rudner, 9p, cit., p. h. 292 But, secondly, it has already been shown that Hempel's analysis of eXplanation itself entails that the scientist do more than assign degrees of confirmation. For in our earlier discus- sion (Chapter II, Section 3) we had occasion to ask whether Hempel intended his theory of explanation as an explication of 'an eXplanation' or of a 'rationally acceptable explanation.‘ The result of that examination showed that Hempel clearly intends the latter, esnecially when he distinguishes between potential eXplan- ations which satisfy conditions (R1) - R3) on the one hand, and well-confirmed or true explanations which satisfy in addition either (Rh‘) or (9h) on the other. This is further evidenced when he invokes as a general and necessary condition of adequacy for any rationally acceptable explanation of a given event x that the explanation "must prOVide information which constitutes good . . . . . . 2h grounds for the belief that x did in fact occur." Hempel has never been satisfied to analyse the notion of explanation merely in terms of a valid argument containing empirical laws among its explanans. Instead, he also insists that the eXplanans be accept- able in order to provide good grounds for the eXplanandum, and hence to be a "correct" eXplanation. Further, when attempting to eliminate_the inconsistency of inductive inference, Hempel again acknowledges the inadeouacy of (R1) - (Rh) as sufficient conditions of rationally acceptable explanations, and supplements them with the principle of total 2h Hempel, "Reasons and Covering Laws in Historical Explanation," p. lho. 2R3 evidence. This immediately goes beyond the scope of inductive logic proper to the methodology of induction and to general prag- matic considerations. Now, if his task were just to explicate the notion of a formally sound (deductively or inductively) potential explanation, then pragmatic and methodological matters would be irrelevant. The scientist might escape by simply assigning degrees of confirmation to eXplanatory hypotheses. But since he intends to eXplicate the notion of an acceptable explanation, which requires acceptable statistical hypotheses as premises, the principle of total evidence and other methodological principles must be invoked. To aeree therefore that these are not purely formal logical principles is only to concede that the explicandum in question is essentially a pracmatic as well as a syntactical and semantical term. And this amounts to Rudner's claim in premise (1), that an adequate explication or rational reconstruction of scientific procedures must include a statement to the effect that the scien- tist as scientist accepts or rejects hypotheses. The price of giving up (1), then, seems to be restricting the CL theory to genuine or even formally valid explanatory arguments, a price Hempel wisely refuses to pay. If these considerations are cogent, (3) follows as a conse- quence. That is, we seem committed to viewing the scientist not just as a counselor of policy makers, but as himself a decision maker. The scientist 333 scientist has then as one of his major goals to replace doubt by true or reasonable belief, and to do this by deciding in the face of uncertainty when the evidence for an eXplanatory hypothesis suffices to warrant his belief in that hypothesis. Further, since (5) obviously follows from (h) and with (3) already established, the tenability of Rudner's argument rests with the move from (3) to (h), from the fact that scientists must make decisions to the fact that these decisions are a function of the seriousness of error. Unfortunately, Rudner offers no argument in the article cited for this central connection. Instead, he fills the gap by appeal- ing to examples from quality control with some of the guiding theories of statistical inference. "If," to take one of his cases, the hypothesis under consideration were to the effect that a toxic ingredient of a drug was not present in lethal quantity, we would require a relatively high degree of confirmation or confidence before accepting the hypothesis-- for the consequences of making a mistake here are exceed- ingly grave by our moral standards. On the other hand, if say, our hypothesis stated that, on the basis of a sample, a certain lot of machine stamped belt buckles was not defective, the degree of confidence we should require would be relatively not so high.25 Thus, taking such an example as a paradigm case of scientific inquiry, he assimilates cases of deciding what to believe to cases of deciding how to act in the face of uncertainty, and concludes that the seriousness of making a mistake, the cost or ethical factor, must be considered in any scientific assessment of statistical explanatory hypotheses. What is unfortunate in the appeal to examples at this crucial stage in the argument, however, is not that there are no better 25 Rudner, pp, cit., p. 2. can: *- grounds for moving from (3) to (h). For indeed there are. Rather, the suppression of the needed assumptions has led defenders of the VN thesis to serious misunderstandings. As a result they recon- struct Rudner's argument by appealing to altogether unnecessary premises in order to fill the gap, and then proceed to challenge these very premises. Professor I. Levi, for instance, clearly sees the required connection between believing or accepting an hypothesis and acting on the basis of the hypothesis relative to some objective or goal. This stresses again the need to interpret explanatory hypotheses, at least in part, methodologically as means or instru- ments adopted in order to achieve some Specified objectives. Rudner's argument assumes, and Hempel concurs, that scientific questions regarding explanation are raised for some purpose, that all problem-solving is an aspect of purposive behavior, and that assessing the rational acceptability of explanatory hypotheses requires weighing cost or utility factors in addition to degrees of probability or confirmation. Hence the relevance of decision theory to the logic of inductive inference rests on its applica- bility to inductive behavior. But when Levi states explicitly the assumptions necessary to move from premise (3) to (h) and hence to the denial of the value- neutrality thesis (5), he commits Rudner and Churchman to two additional premises, one of which is unduly and unnecessarily strong, but which is nevertheless made the center of controversy. The two additional assumptions cited by Levi are as follows: ' ~rvm .h—A——a—— 9,86 (6) "To choose to accept a hypothesis H as true (or to believe that H is true) is equivalent to choosing to act on the basis of H relative to some specific objective P." (7) "The degree of confirmation that a hypothesis H must have before one is warranted in choosing to act on the basis of H relative to an objective P is a function of the seriousness of the error relative to P resglting from basing the action on the wrong hypothesis.“2 Now (7) he accepts on the authority of the statistical theories of such statisticians as Pearson, Neyman and Wald. But since (7) without (6) fails to yield (5), the tenability of Rudner's case for (5) is made to rest on (6), on an equivalence between believing an hypothesis and acting on the basis of it relative to some objective P. But whether or not Rudner or Churchman would accept (6) as an adequate formulation of their suppressed premise is simply not the question at issue. For a weaker statement will suffice. Hence Levi raises a false issue and thereby confuses what I think are the important issues. Let me clarify. Levi assumes that the only way to derive the desired conclusion (5) is on the basis of a "behavioralist" analysis of belief, whereby "accepting a proposi- tion H as true" is synonymous with, equivalent to, equated with, or reducible to "acting on the basis of H relative to a practical 7 objective P."2 26 Levi, "Must the Scientist Make Value Judgments?", p. 3h8. 27 Levi, “Decision Theory and Confirmation," p. 615; and "On the Seriousness of Mistakes," pp. h8 and SO. 28? Yet such an assumption is not essential in order to yield (5). It is unnecessary, in other words, to take a behavioralist view of belief. The only condition required for Rudner's argument is that belief be tied to action in such a way that beliefs can be justi— fied by practical considerations of action, and that different mistakes in action be taken with different degrees of seriousness, so that the acceptance of beliefs is not merely the result of a r quest for truth and nothing but the truth. In place of the equiva- E lence condition in (6) an implication suffices: (6') To choose to accept a hypothesis H as true (or to believe that H is true) im lies as a necessary (but not necessarily a sufficient condition the disposi- tion to act on the basis of H relative to some Specific objective P. Once the substitution of (6') for (6) is made, the issue no longer turns on a behavioralist reduction of belief to action, but clearly focuses on the question posed by Hempel's attempt to rein- force the principle of total evidence by formulating acceptance rules for rational belief, viz., what kinds of values the assigned utilities are to represent. Three different kinds of values seem relevant: truth for its own sake, truth tempered by such other theoretical concerns as simplicity and eXplanatory power, and truth tempered by these theoretical concerns in addition to such pragmatic and practical concerns as the seriousness for action in making mistakes. Once again, however, confusion threatens. For Rudner‘s premise (h), which invokes the notion of the seriousness or impor- tance of error, does not require pragmatic costs or utilities to 288 replace truth, high confirmation, simplicity or even explanatory power as appropriate criteria for appraising the rational accepta- bility of hypotheses. Nor does it entail defining any of these prOperties in terms of pragmatic utility. In fact, in another article Rudner explicitly refers to three independent weights necessary for an adequate theory of inductive inference and for the assessment of rational choice among explanatory hypotheses: evidential strength, simplicity and pragmatic utility.28 HaVing warded off the false issues concerning the move from (3) to (h)--a behavioralist reduction of belief to action and a pragmatic reduction of truth to utility-~let us return to the main issue at hand, to our incompleted discussion in the last section of Hempel's "accepted-information model" and his notion of "epistemic utility." Rudner's argument entails that scientists do not rationally accept hypotheses as an outcome of a quest for truth and nothing but the truth, untempered by such other factors as simplicity, cost of explanatory power. Since Hempel's "accepted-information model" assumes they can, this conflict serves as a firm basis from which to begin a defense of (h). Suppose, then, an inquirer to be seeking the truth and nothing but the truth, to be concerned only with adding true or false statements into the body of knowledge K, or, in Levi's phrase, to be "concerned to select from a given list the one and only one proposition that is true regardless of any other properties 28 . . . Rudner, "An Introduction to Simplicity," Philosophy of Science, Vol. XXVIII (1961), p. 110. “— 289 that it might have."29 To be rational in this case, the inquirer seems committed to weighing each possible mistake he might make as equally serious. But such a commitment, on Rudner's argument, cannot be satisfied since different mistakes have different degrees of seriousness or importance. It will be helpful at this juncture to see how Hempel develops his notion of epistemic utility, in order to exhibit the relative weakness of his position and the strength of Rudner‘s. In the last section we noted that the problem of establishing acceptance rules for scientific explanatory hypotheses, of decid- ing which scientific hypotheses to accept and thus add to the body of scientific knowledge, turns on what kinds of values the utili- ties assigned to outcomes are to represent. Hempel's answer is clear: the utilities should reflect the value or disvalue which the different outcomes have from the point of view of pure scientific research rather than the practical advantages or disadvantages that might result from the application of an accepted hypothesis, according as the latter is true or false. Let me refer to the kind of utilities thus vaguely characterized as purely scientific, 93 epistemic, utilities.3O The problem then concerns finding a measure of the epistemic utility of adding an hypothesis h to the previously established system of knowledge K. But Hempel immediately recognizes the need to make such utilities depend on how much of what h asserts is new information 29 Levi, "On the Seriousness of Mistakes," pp. b9 and 51. 3O Hempel, "Inductive Inconsistencies," p. 15h. 290 not already contained in K. Taking k as a sentence with the same informational content as K, the common content of h and K is given by hl’k. Thus, since h is equivalent to (h v k). (h v - k), the content of h which goes beyond that of K is eXpressed by (h v - k). Hempel then introduces the notion of a "content measure function for a [suitably formalized7 language L," 1.3,, a function m assigning to every sentence 5 of L a number m (s) such that "(i) m (s) is a number in the interval from O to l, inclusive of the endpoints; (ii) m (s) = 0 just in case s is a logical truth; (iii) if 31 and 32 have no common content--i.g., if the sentence 51 v s2, which expresses their common content, is a logical truth--then m(sl - $2) = m(Sl) + m(82)."31 If m is then used as a content measure for a language suited to the purposes of empirical science, the utility of adding h to K is given as the Tentative measure of epistemic utility: the epistemic utility of accepting a hypothesis h into the set K of previously accepted scientific statements is m(h v - k) if h is true, and -m(h v - k) if h is false; the utility pf legging h in suSpense, and thus leaving K unchanged, is 0. Finally, Carnap's rule of maximizing estimated utility, mentioned earlier, warrants the acceptance or rejection of h as epistemi- cally rational in accordance with the following "Tentative rule for inductive acceptance: Accept or reject h, given K, according 31 $2190: Po 15h; and also Hempel, "Deductive-Nomological VS. Statistical Explanation," p. 15h. 32 Ibid., p. 15h. IA‘O-1 ~ I as c(h, k)>l/2 or c(h, k)< 1/9; when c(h, k) = 1/2 h may be accepted. rejected, or left in suspense."33 Since on this account a scientist is interested only in accepting h when it is true, the possible correct answers of accepting h when true and rejecting h when false are to be considered equally desirable and the corres- ponding mistakes equally undesiraole. Hence, we are led to the unsettling recommendation of accepting, i.e., acting on the basis of, h when its degree of confirmation or probability is merely 0.51. Both Hempel and Levi readily concede the difficulty with this account: the rule is much too liberal or lenient to be suitable for even pure scientific procedure. Further, Levi points to the same problem when other methods are used instead of Carnap's maximizing utility, g.g., Bayes method and the method of signifi- cance testing proposed by Neyman and Pearson. He also acknowledges that accepting h "when its degree of confirmation is low does not seem reasonable unless this acceptance is reduced to action under- taken to realize some objective other than seeking the truth and nothing but the truth,"311 g.g., formal or structural simplicity and eXplanatory power. With this Hempel concurs. Levi, however, pursues the third option, suSpended judgment or remaining in doubt, as a way of avoiding the difficulty and hence of defending the View that scientists accept or reject hypotheses in quest of the 33 Ibid., p. 155. 3h Levi, "On the Seriousness of Mistakes,“ p. 55. 292 truth and nothing but the truth. In this sense, the scientist is to suSpend judgment on h when he feels the total available evidence warrants neither the acceptance nor rejection of h. Consider one of Levi's examples of the Bayes method applied to the problem of replacing doubt by true belief. Suppose an eXperi- mental psychologist wants to determine whether or not some person has extrasensory perception. By assuming that the subject has ESP if he guesses correctly the colors of cards drawn randomly with a frequency greater than .6, and otherwise not, the psychologist must decide on the basis of a sample of guesses whether the long- ranee frequency of correct guesses is .6 or less (H1), or is greater than .6 (H2), when A1 is the act of accepting hypothesis H1. Since the eXperimenter presumably is interested only in accepting that hypothesis as true which is true, he should consider the possible correct answers 011 (accepting H when it is true) and 022 (accepting H2 when it is true) as equally desirable and the correSponding 'mistakes' or 'errors' 012 and 021 as equally undesirable....Conseouently, the matrix for this problem will be as follows: H1 H2 A1 1 0 A2 0 l The Bayes method recommends adopting A if the probability of H1 is greater than 1/2 and A2 if the probability of H is less than l/2. Consequently if the probability of H2... is .3%, we would be warranted in accepting H2 (adopting A2). Ibid., p. Sh. 292 Now, to avoid this unsettling and unreasonable consequence and yet preserve application of the method to cases where the goal is to seek the truth and nothing but the truth, Levi revises the problem by adding a third option: the act S of suSpending judgment or remaining in doubt. In this case the matrix changes to the following: H1 H2 Al 1 O 3. A2 0 1 S k k where the utility k is the value of act 3 when the hypothesis H1 is true. But the price of thus avoiding the unsettling consequence, by assigning a sufficiently high value to k so that a high degree of confirmation is required to warrant the acceptance of any hypothesis, consists in selecting arbitrarily the value assigned to k. However, this arbitrary assignment brings us back to the essential point of Rudner's move from premise (3) to (h), from the fact that scientists must make decisions (which Levi concedes) to the fact that these decisions are a function of the seriousness of error. For Rudner's argument hinges on the assumption that an arbitrary selection of a value to be assigned to k in such cases as the above is unreasonable, eSpecially if there are grounds for making such assignments of values. This is why premise (6'), which Connects beliefs or the acceptance of hypotheses with acting on the basis of them relative to some goal, provides the crucial link a- - *J’.“-‘ «as 79h between the decision to accept in (3) and the different degrees of seriousness of error in (h). Instead of assigning a value to k arbitrarily or on the basis of sheer convention, we take our cue from (6'), from the fact that belief implies a disposition to act. Our choice of a value for k then depends upon how serious a mis- take would be if we acted upon the hypothesis, i323, if we accepted a false hypothesis or rejected a true one. This in turn depends upon how useful the hypothesis in question is to us. So, while on purely losical grounds the value assigned to k is arbi- trary, still a rational choice of assignments can and must be made on extralogical grounds, 3223’ on the basis of pragmatic utilities: the purposes for which we intend to use the hypothesis and the relative advantages and disadvantages of acting on the basis of it relative to these objectives. It should be noted here that Levi is not unaware of this type of consideration. After noting the unhappy result of his revision of the ESP problem, i.e., the arbitrary assignment of a value for k, he offers further revision by introducing the notion of "degrees of caution."36 In the case of his truth-seeking experi- menter, who assigns a value to k high enough to assure that the Bayes method will recommend suspended judgment for some assignments of probability to H1 and H2, we have the case of a scientist who "takes mistakes more seriously in relation to eliminating doubt" than might some others. In contrast to a more tenacious scientist, 0 Ibid., p. 56. 2 95 he exercises a higher degree of caution, indicating ”that to seek the truth and nothing but the truth is to aim at a complex objec- tive. It is an attempt to eliminate doubt tempered by an interest 37 in finding truth and avoiding error." But this of course con- cedes the point at issue, i.e., the move from premise (3) to (b). It concedes as Levi acknowledges "that the choice of a roblem ’ 3 , whose goal is to replace doubt by true belief and, as a conse- ouence, the choice of a degree of caution in realizing such a goal 38 can be both influenced and justified by practical considerations." Horeover, the fact that this conclusion implies nothing about a behavioristic reduction of beliefs, acceptance of hypotheses or suspended judgments constitutes less solace for the value-neutrality thesis than Levi recognizes. For, as we argued earlier, Rudner's argument opposing the value-neutrality thesis depends not on (6) but on (6'), not on a reduction of belief to action but merely on an implication between them such that accepting h entails (but is not necessarily equivalent to) a disposition to act on the basis of h relative to some objective. Nevertheless, in concentrating his efforts against (6), Levi fails to see that by conceding the scientist's need to reckon with the seriousness of mistakes, even though not allowing a reduction of belief to action, he is forced to acknowledge that different mistakes be taken with different degrees of seriousness. In other words, he totally neglects (6') 37 Ibid., p. 55. as Ibic., p. 57. 296 which suffices to get dudner's argument to the conclusion (5), that scientists 323 scientists make value judgments regarding the seriousness of errors in acting upon accepted or rejected hypo- thesis. For, the different purposes for which one acts clearly require different cost factors and, as a result, different degrees of seriousness. Even when the use for which one intends to accept an hypothesis is purely intellectual or theoretical, "the measure of these losses will depend upon the difficulties of devising alternative theories, and the comp ications of accepting a wrong theory from the point 39 of view of other fields of investigation."” 30 merely by grant- ing (6'), that acceptance or belief entails a disposition to act, and that the seriousness of mistakes must be reckoned with in deciding to accept or reject hypotheses, Levi seems forced to con- cede, in addition, that different degrees of seriousness must be considered, since these degrees result from the different purposes upon which we act. Hence, Levi has in effect conceded (S), the denial of value-neutrality. Clearly, if the preceding considerations are cogent, the scientist must take into account in his decision to accept or reject hypotheses (3), not only epistemic utilities but also prag- matic utilities. Consecuently the decision will vary with the kind of action to be based upon the hypothesis. Consider again the example cited earlier regarding the toxic ingredient of a drug. 39 Churchman, Theorv 2: Experimental Inference, p. 250, 1297 Because of the different utilities involved, a decision rule might warrant, on the same evidence, that the hypothesis in question should be accepted if applied to eXperimental animals, but rejected if applied to humans where an error would obviously be more serious. Hence the decision to accept or reject hywotheses is in part instru- mental to action; it is partially a decision to adopt one of alternative courses of action. And this decision can be ration- ally defended only by considering the pragmatic gains and losses attached to the possible outcomes of the actions. This amounts to resting the acceptance of an hypothesis upon whether or not it leads to successful action. Having thus unpacked and defended the experimentalist argument opuosine the value-neutrality thesis of CL theorists, let me now present it in full array. (1) The scientist as scientist accents or rejects hypotheses. (2) No scientific hypothesis is ever completely verified, but all are corrigible. (3) Therefore, the scientist must make the decision that the evidence is sufficiently strong to warrant the acceptance of the hypothesis. (6') To choose to accept a hypothesis H as true (or to believe that H is true) implies as a necessary (but not necessarily a sufficienti condition the diSposition to act on the basis of H relative to some specific objective P. (7) The degree of confirmation that a hypothesis H must have before one is warranted in choosing to act on the basis of H relative to an objective P is a function of the seriousness of the error relative to P resulting from basing the action on the wrong hypothesis. (L1) 2 93 Therefore, the decisign in (3) regarding the evidenc and reSpecting how strong is "strong enough" to rationally accept H is going to be a function of the importance or seriousness, in the typically ethical sense, of making a mistake in accepting or rejecting the hypothesis. (D Hence, the scientist as scientist does make value judgments. 299 Value-Neutrality and Historical Explanation In order to focus attention once again on historical explana- tion, let me consider briefly two final but important and related objections to the foregoing argument, objections posed most recently by Hempel, Nagel and Rescher but the roots of which lie deeplv embedded in Aristotle's distinction between the theoretical and practical sciences. The import of the experimentalist argu- ment, it would seem, is to force a reconsideration not only of the meaning of scientific and historical objectivity but also of the relations between the theoretical, technological and policy-making aspects of rational inquiry. Each of these issues is obviously too complex to be considered here in any detail. we will accordingly limit discussion to two related points. First, though Hempel is unable to find any "satisfactory general way of resolving the issue between the two conceptions of science" (represented by the "accepted-information model" and the "pragmatist or instrumentalist model"), he still claims "that it would be pointless to formulate criteria of applicability by reference to pragmatic utilities; for we are concerned here wdth purely theoretical (in contrast to applied) explanatory and pre- dictive statistical arguments."h0 Nagel follows suit by objecting to the experimentalist argument on the grounds that it mislead- ingly "suggests that alternative decisions between statistical hypotheses must invariably lead to alternative actions having ho Hempel, "Deductive-Nomolcgical vs. Statistical Explanation," p. 162. 300 immediate practical consequences upon which different Special values are placed."hl Beyond the controversial issues raised by these passages which have already been considered in our defense of Rudner‘s argument, this kind of objection forces one additional clarifi- cation. For the argument against the value-neutrality thesis in no way depends on considering, in the weighting of pragmatic utilities, only the “alternative actions having immediate practical consequences." No doubt some scientific hypotheses will be closely linked to such actions. But others, of a higher degree of "theo- retical purity" will be more tenuously linked to such actions via the former ones. In fact, however, this linkage need be no more tenuous than the linkage of empirical sicnificance between hypotheses containing theoretical constructs and relevant observa- tional statements. In other words, it seems to make as much sense to sneak of degrees of practical consequences and pragmatic utilities, with a system or network of hypotheses linked at its edges to immediate practical consequences, as it does for Hempel to Speak of degrees of testability or empirical significance, with a network of hypotheses linked at its edges to immediate observations. Surely nothing in either linkage precludes the other. Moreover, it should come as no surprise that the goals or objectives used to weigh the consequences of wrong estimations in the "purely" kl Nagel, The Structure 2: Science, p. D97. 30.1 theoretical sciences will be of a more complex nature and require more general criteria than those used in the more practical sciences, such as qualitv control and engineering. Hence, even if there are cases where the choice of policy for deciding to accept or reject hypotheses does not depend directly upon immediate practical consequences or costs, still the costs or pragmatic utilities are relevant in a more indirect manner. In this sense, then, I think Hempel is correct in his charge that, on the basis of pragmatic utilities alone, "what is quali- fied as rational is, prOperly Speaking, not the decision to believe h to be true, but the decision to act in the given context .112 as if one believed h to be true....‘ Yet, as with Levi's earlier objections, this damages Rudner's argument only if (6) is substi- tuted for (6') and the behavioralist reduction fails. 'flithout this unnecessary assumption, the correct residue of Hempel's point is that pragmatic utilities are relevant to the decision to accept or reject hypotheses only to the degree to which such hypotheses or beliefs are linked to actions. The second point of interest brings us, finally, more directly to the application of the eXperimentalist argument to historical eXplanations. In a stimulating article under the joint authorship of Professors Rescher and Joynt, the close relationship between evidence in history and in the law is challenged. The grounds are that historical inquiry, unlike legal proceedings, involves no 11,2 Hempel, "Deductive-Nomological vs. Statistical Explanation," p. 162. "stake," since "history has but one single purpose: the establish- ment of truth regarding the past...and the truth alone," while the law tempers this goal with "a view to the protection of individual "143 rights. Rescher and Joynt note some important differences between historical investigation and legal adjudication. For example, heresay evidence is not wholly inadmissable in history as in the courts. Nor can the historian afford the luxury of the high standard of evidential acceptability prevalent in the law: of automatically excluding certain witnesses, of admitting testimony only from witnesses who can be cross-examined, of imposing severe limitations on admissable evidence, and of imposing certain minima of evidence combined with rigid standards of proof. DeSpite the importance of these differences, however, it is not at all clearly established that they stem from the same root, that the explanatory rationale accounting for such differences lies in the suggested different purposes of history and law. By assimilating legal procedures to scientific inquiry, with Bentham and J. H. Wigmore, yet divorcing both from historical investiga- tion since the latter seeks the truth and nothing but the truth, the authors in effect argue that even if the eXperimentalist argument is cogent for scientific and legal inquiry, it nonethe— less fails toaapply to history. Now, the main reason why we appealed to the denial of the value-neutrality thesis for scientific explanation in the first l. “3 N. Rescher and C. Joynt, "Evidence in History and in the Law," Journal pf_Philosonhy, Vol. LVI (1959), pp. 567 and S77. 303 place was to show that the emphasis of empathy theorists on the subjective element of judgment was correct, but that it concerns the rational acceptability of the explanans hypotheses, not the logical eXplanatory force of an argument. He thus defended the CL theory against the SU thesis at the price of giving up value- neutrality. That is, we have tried to develop what seems to be the important insight of Verstehen theorists, and to show that the attempts to reconstruct the significant residue of Weber's SU thesis are well-intended though ill-enacted. The residual element is indeed overlooked by CL theorists in their attack on Kerstehen. But the development of this element, along the lines suggested by Lavine, Natanson, Schutz, Dray, Donagan and Scriven, proves singularly ineffective against the general logical provisions of the CL theory. For the notion of Verstehen in no acceptable way argues against (R1) - (Rh), inclusive of (Rl') and (Rh')' Instead, as we have shown in taking our cue from suggestions of Lavine, Dray and Scriven, the significant insight of Verstehen theorists on the need for the historian to make decisions and judgments in order to accept explanations as correct entails denying the value- neutrality thesis, not affirming the SU thesis. Nevertheless, we also tried to support Dray's claim that this residual element, though clearly in the pragmatic dimension of explanation, is not merely a presystematic or psychological matter of discovering hypotheses as CL theorists suggest. Rather, it plays an essential role in any adequate philosophic explication or theory of rationally acceptable explanation. However, both Dray 301; and Scriven overplay their hand. They substitute this pragmatic utility or value element, in the form of principles of action and normic truisms, for the eXplanatory role of empirical laws. Our argument, in other words, rests with the twofold claim that the denial of value-neutrality can be sustained, but that this in no way iuplies support for the SU thesis in any of the reconstructed ve~sions considered herein. Consequently, we take the important residue of Weber's view as an additional pragmatic condition to be tacked on to the other provisions of the CL theory, rather than, as Dray suogests, a condition making the CL theory "peculiarly inept" in accounting for human actions. We have shown, accordingly, that while Dray's argument does not support the conclusion he wants, it still suggests the criticism of the CL theory pursued in this chapter: that the historian and scientist must, in his explanatory practice, make value assessments or decisions, not in place of but over and above, and eSpecially about, his covering laws or eXplanatory hypotheses. Moreover, Dray's attempt to enter a wedge between scientific and historical explanations has not been successful largely because, following other defenders of the "empathy" position, he uncritically accepts the CL theorists' claim of value-neutrality in scientific inquir- ies. This error, I think, prevents him from sustaining a success- ful attack against the CL theory of explanation, and also hinders him from developing the important residue of heber's position. Further, in considering why Dray so persistently resists the CL proposals in the face of their cogent opposition to his , —'r r' / ""2. ; /;" /1 //1 ,/ ' ,1, " / .r r/ 4. ‘ . a -fi‘- . a..- .. ~,,j_:~ ‘90 Q \ J1 rational model, the answer again seems to turn on his uncritical acceptance of the CL theorists' value-free claim. He seems to believe, as a result, that if appraisals or value decisions do not constitute the explanatory force of historical explanations, then Weber's positivist insistence on the value-neutrality thesis will go unchallenged, except in some presystematic sense. Believing, additionally, that the value-neutrality thesis is unacceptable, he is led to View history as a branch of the humanities, and thereby to a twofold criticism of the CL theory: as a kind of conceptual barrier to a humanistically oriented historiography, and as a purely visionary attempt to reform history along scientific lines. But since, as we have argued in this chapter, even scientific eXplanation requires the making of value judgments, and hence at least to this extent a humanistic orientation, Dray's resistance to the CL theory is both unnecessary and poor strategy. ‘While a humanistic orientation of history and of science is indeed impor- tant, the price of abandoning history as a branch of the science of society seems much too high, eSpecially when we might well have both. And just this constitutes the point of our thesis that Hempel's CL theory survives the varied logical criticisms of 'empathy' theorists, but only on condition that pragmatic elements be included essentially in a logical reconstruction of probabilis- tic eXplanations, that the emphasis on the structural aspects of eXplanation be extended to include a purposive ingredient, and hence that the value-neutrality thesis be denied. But denied, of course, so as not to be committed prematurely, if at all, to 306 either a behavioralist reduction of belief to action or a crudely pragmatic reduction of truth or confirmation to utility. This being in sum our case, its completion requires a defense against the charge of Rescher and Joynt. It needs to be shown, in particular, that history does not differ from scientific or legal inquiry by having, unlike them, but one goal or purpose, the pursuit of truth and nothing but the truth. But, since the other differ- ences between history and the law seem cogent, we must then account for these differences in standards of evidential acceptability without appealing to the value-neutrality thesis as the explanatory rationale. For if they are correct in thinking that this is the rationale, then our Opposition to value-neutrality in scientific eXplanation mijht not support our general case regarding the extension of the CL theory to historical eXplanations. Let me offer then, as a finale, a brief alternative account of the weaker standards of evidential acceptability in history. Rescher and Joynt correctly acknowledge that good or suffi- cient evidence for reasonable belief is not a matter of any abso- lute standards but a function of both pragmatic and expedient considerations. It concerns the purposes for which we recuire the evidence and also the kinds of relevant data available in the hb, particular discipline. But they stress only the different pur- poses or functions of history and law, a difference between the investigative quest for truth and the adjudicative quest for hh ibid,, p. 56h. 307 adjustment of conflicting claims regarding things of value. This to the exclusion of the matter of what they call exnediency. In fact, their major conclusion, that history seeks the truth and only the truth, rests completely on this exclusion. One further related puzzlement seems noteworthy. In the context of the experimentalist argument presented in the last section, one would eXpect this conclusion about the truth-seeking purpose of history to be defended by assimilating history to the pure theoretical sciences rather than to the applied sciences. If there are in historical inquiry no other goals than truth-seeking, it would seem unreasonable to accept historical explanations that provide a mere comparative likelihood, i.e., that are merely more 'kely to be true than their competitors. ‘Ne should instead be inclined to accept only explanations providing a high likelihood pg: is. In other words, if we are to accept hypotheses in a reasonable manner on the basis of a comparative likelihood, which might mean that the degree of confirmation or evidence for the hypotheses EE£.§§ is extremely low, we should eXpect this accept- ance to be based on action undertaken to realize some objective or goal other than just truth-seeking. Yet Rescher and Joynt clearly acknowledge the evidential criteria operative in history to be "akin to those of practice, i.e., of applied science, rather than to those of theory or 2332 science." Thus can the historian rest satisfied when his explana- tions are shown to be merely "significantly more likely than any 108 of the comparable alternatives.”hg The unsettlin: and unreasonable conseouence of this position can be avoided, however, only by grounding the acceptance on other goals than truth—seeking. And this, in effect, they do by makine the notion of sufficient evi- dence a function of the expedient matter of the kinds of evidence available, i.e., of "the paucity of the data and the difficulties in their interpretation, which are typical and insuperable features hé throughout his [the historian'g7 domain." Hence, their case for the conclusion that history seeks the truth and nothing but the truth, and thereby for their account of the other differences between history and the law, rests on excluding just this expedient or cost factor. For when this factor is considered in the analysis of these differences in the strength of evidential standards, the differences between the strong criteria applied in law or science and the weaker criteria utilized in history turn out to be not clear-cut but differences of degree. Further, the explanatory rationale of these differences is not that history, unlike law, seeks the truth and only the truth, but rather that the cost factor or pragmatic utilities referred to in the last section differ. Surely there can be no doubt, for instance, that the seriousness or importance of making a mistake in legal adjudication is much greater by our moral standards than com- parable mistakes in historical inquiry. Hence, Rescher and Joynt hS Ibid., p. 563—h. hé Ibid., p. 56h. I ,/ / ../' » 7/ Ft 309 correctly point to the protection of individual rights as the reason for requiring stricter evidential standards in legal pro- ceedinbs. In this sense, legal hypotheses are more closely linked to "alternative actions having immediate practical consequences." Nevertheless, the historian also has pragmatic utilities to consider in accepting or rejecting explanatory hypotheses. Surely the reason he often accepts hypotheses with little likelihood Nun's-wanna- r '- per se but high comparative likelihood is not because he seeks {:5 I truth alone, but because his quest for truth is tempered by the paucity of data and difficulties of interpretation. It is not because there is no "stake" involved but because the "stake" is less critical, i.e., because the losses involved are only indi- rectly related to immediate practical consequences. Of more direct concern for the historian are the costs involved in the difficulties of devisine alternative eXplanatory theories and the complications of accepting a wrong theory from the point of View of such allied fields as psychology and socioloey. So, since historical explanations have a higher degree of "theoretical purity" than legal hypotheses, they are more tenuously linked to immediate practical consequences in action, and hence the degree of seriousness in making a mistake in history will be much less than in level proceedincs. Put the crucial point of our argument, to repeat, is that the only reasonable arounds for accepting hypotheses on a comparative likelihood rather than on a high likelihood pe£_§e is that the acceptance be based on action undertaken to achieve some goal 310 other than just truth-seekins. This very point was recognized by Bishop Butler even in the following passage quoted by descher and Joynt: "In matters of practice it will lay us under an absolute and formal obliaation, in point of prudence and of interest, to act under that presumption or low probability, though it be so low . ‘ . . _ , . . . h? as to leave the mind in very great doupt which 18 the truth." It seems unsettlinf and unreasonable indeed to accept an hypothesis on evidence which leaves the mind in very great doubt as to its truth, if our only goal is truth seeking. If this point be granted, it follows that the acceptance must be tied to action, however tenuous the link might be. Hence, according to the experimentalist argument, the istorian as well as the scientist must make value judements, in which case his goal cannot be just truth-seeking alone. And, finally, as a result of these considerations we rest our case for the twofold thesis that hemnel's CL theory 0? explanation survives the varied attacks of 'empathy' theorists, but only on condition that the value-neutrality thesis be surrendered. “' Ibid., p. s63. BIBLIOGRAPHY Hooks Aristotle. The Basic Works of Aristotle, R. McKean (ed.), New York: Random House, l9hl. Ayer, A. J. Language, Truth and Logic, New York: Dover, l9h6. Barker, 8. Induction and Hypothesis, Ithaca, New York: Cornell University, 1957. Becker, H. Through Values to Social Interpretation, North Carolina: Duke University, 1950. Braithwaite, R. Scientific Explanation, New York: Harpers, 1953. Bunge, N. Causality, New York: World Publishers, 1962. Butterfield, H. History and Human Relations, London: Oxford Uni- versity, 1951. Carnap, R. Logical Foundations of Probability, Chicago: Univer- sity of Chicago, 1950. , Meaning and Necessity, Chicago: University of Chicago, l9b7. Churchman, C. w. Prediction and Optimal Decision, Englewood Cliffs, New Jersey: Prentice-Hall, 1961. . Theory of Experimental Inference, New York: Macmillan, 1 £8. Collingwood, R. G. The Idea of History, New York: Oxford University, 19h6. Dray, W; Laws and Explanation in History, London: Oxford University, 19270 . PhilOSOphy of History, Englewood Cliffs, New Jersey: Prentice-Hall, l96h. Feigl, H. and Brodbeck, M. (eds.). Readings in the Philosophy of Science, New York: Appleton-Century, I953. Feigl, H. and Maxwell, G. (eds.). Current Issues in the EhiIOSOphy of Science, New York: Holt, 1961. . Minnesota Studies in the PhilOSOphy of Science, Vols. I, II, III, Minneapolis: University of Minnesota, 1962. 311 \/ I.” 312 Frank, P. PhilosOphy of Science, Englewood Cliffs, New Jersey: Prentice-Hall, 1957. Freud, S. Basic Writings of Sigmund Freud, New York: Random Gardiner, P. (ed.), Theories of History, Glencoe, Illinois: Free Press, 1962. . The Nature of Historical Explanation, London: Oxford University, 1952. Gross, L. (ed.), Symposium on Sociological Theory, Evanston, Illinois: Row and Peterson, 1959. Hook, S. (ed.), PhiIOSOphy and History, New York: New York Univer- Hospers, J. Human Conduct, New York: Harcourt, Brace and werld, 1961. Lewis, C. I. Ground and Nature of the Right, New York: Columbia University, 1955. Madden, E. (ed.), Structure of Scientific Thought, Boston: Houghton Nifflin, 1960. Nises, L. von. Epistemological Problems of Economics, New York: Van Nostrand, I960. Nagel, E. The Structure of Science, New York: Harcourt, Brace and World, 1961. Natanson, M. (ed.), Philosophy of the Social Sciences, New York: Random House, 1963. Neyman, J. A First Course in Probability and Statistics, New York: Holt, 1950. POpper, K. Logic of Scientific Discovery, New York: Basic Books, 1959. . Open Society and Its Enemies, New York: Harper and Row, 1952. . Poverty of Historicism, London: Routledge, Kegan Paul, 1957. Quine, N.V.O. From A Logical Point of View, Cambridge, Massachusetts: Harvard University, 1953. . Word and Object, New York: Wiley, 1960. Ryle, G. Concept of Mind, New York: Barnes and Noble, l9h9. 313 Savage, L. Foundations of Statistics, New York: Wiley, 195h. Scheffler, I. Anatomy of Inquiry, New York: Knopf, 1963. Schilpp, P. (ed.), Philosophy of Rudolph Carnap, La Salle, Illinois: Open Court, 1963. Singer, E.A. EXperience and Reflection, Philadelphia: Pennsylvania University, 1959. Singer, N. Generalization In Ethics, New York: Knopf, 1961. Wald, A. On the Principles of Statistical Inference, South Bend, Indiana: Notre Dame University, l9h2. Nalsh, N.H. PhiIOSOphy of History, New York: Harper and Row, 1951. Haber, N. Methodology of the Social Sciences, Glencoe, Illinois: Free Press, l9h9. . Theory of Social and Economic Organization, New York: Oxford University, l9h7. Yolton, Thinking and Perceiving, La Salle, Illinois: Open Court, 1962. Articles Abel, T. "Operation Called 'Verstehen'", in Madden, E. (ed.), Struc- ture of Scientific Thought, Boston: Houghton Nifflin, 1960. Ayer, A. J. "The Conception of Probability as a Logical Relation", madden, E. Structure of Scientific Thought, Boston: Houghton Nifflin, 1960. Barker, E. "Rational Explanations in History", Hook, S. (ed.), PhiICSOphy and History, New York: New York University, 1963. Barker, S. "The Role of Simplicity in Explanation", Feigl, H. and Eaxwell, G. (eds.), Current issues in the Philosophy of Science, New York: Holt, 1961. Brandt, R. "Personality Traits as Causal Explanations in Biography", Hook, 5. (ed.), Philosophy and History, New York: New York University, 1963. Brodbeck, M. "EXplanation, Prediction and 'Imperfect} Knowledge", Feigl, H. and Maxwell, G. (eds.), Minnesota Studies, Vol. III, Minneapolis: University of Minnesota, 1962. Churchman, C.N. "Statistics, Pragmatics, Induction", Philosonhy of Science, xv, July, l9h8. ‘ AA. “A regs? . .. “3"? ’. IAII) 31h Cohen, J. "Review", Philosophical Quarterly, April, 1960. Donagan, A. "Explanation in History", Gardiner, P. (ed.), Theories of History, Glencoe, Illinois: Free Press, 1962. Dray, N. "The Historical Explanation of Actions Reconsidered", Hook, S. (ed.), Philosophy and History, New York: New York University, 1963. Feyerabend, P. "Comments", Feigl, H. and Maxwell, G. (eds.), Current Issues in the Philosophy of Science, New York: Holt, 1961. Gershoy, L. "Some Problems of a Working Historian", Hook, 3. (ed.), Philosophy and History, New York: New York University, 1963. Goodman, N. "The Significance of Der Logische Aufbau Der Welt", Schilpp, P. (ed.), The Philosophy of Rudolph Carnap, La Salle, Illinois: Open Court, 1963. Grunbaum, A. "Temporally—Asymmetric Principles, Parity Between Explanation and Prediction, and Mechanism Versus Teleology", Philosophy of Science, XXIX, April, 1962. Hempel, C. "Concept of Cognitive Significance", Proceedings of the American Academy of Arts and Sciences, LXXX, l951-l95h, 61. . "Deductive-Nomological vs. Statistical Explanation", Feigl, H. and Maxwell, G. (eds.), Minnesota Studies, Minneapolis: University of Minnesota, 1963. . "Explanation in Science and History", Colady, R. (ed.), Frontiers of Science and Philosophy, Pittsburgh: University of Pittsburgh, 1962. "Function of General Laws in History", Gardiner, P. (ed.), Theories of History, Glencoe, Illinois: Free Press, 1962. "Inductive Inconsistencies", Logic and Language, Holland: Reidel, 1962. "Logic of Functional Analysis", Gross, L. (ed.), Symposium on Sociological Theory, Evanston, Illinois: Row, Peterson, 1959. . "Problems and Changes in the Empiricist Criterion of Meaning", Linsky, L. (ed.), Semantics and the PhiIOSOphy of Language, Illinois: University of Illinois, 1952. . "Rational Action", Proceedings and Addresses of the Ameri- can Philosophical Association, XXXV, 1962. . "Reasons and Covering Laws in Historical Explanation", Hook, S. (ed.), Philosophy and History, New York: New York University, 1963. 315 "Theoretician's Dilemma", Feigl, H. and Laxwell, G. (eds.), Minnesota Studies, Vol. II, Minneapolis: University of Minnesota, 1958. Hempel C. and Oppenheim, P. "Studies in the Logic of Scientific Explanation", Feigl, H. and Brodbeck, M. (eds.), Readings in the Philosophy of Science, New York: Appleton—Century, 1953. Kim, J. "Inference, Explanation and Prediction", Journal of Philoso- phy, LXI, June, l96h. Krieger, L. "Comments on Historical Explanation", Hook, S. (ed.), Philosophy and History, New York: New York University, 1963. Lavine, T. "Note to Naturalists on the Human Spirit", Natanson, M. (ed.), Philosophy of the Social Sciences, New York: Random House, 1963. Levi, I. "Decision Theory and Conlirmation", Journal of PhilOSOphy, VIII, 196l. "Must the Scientist Make Value Judgments?", Journal of Philosophy, LVII, 1960. "On the Seriousness of Mistakes", PhiIOSOphy of Science, XXIX, 1962. Martindale, D. "Sociological Theory and the Ideal Type", Gross, L. (ed.), Symposium on Sociological Theory, New York: Row, Peter- son, 1959. Maxwell, G. and Feigl, H. "Why Ordinary Language Needs Reforming", Journal of Philosophy, LVIII, 1961. Mazlish, B. "On Rational Explanation in History", Hook, S. (ed.), Philosophy and History, New York: New York University, 1963. Natanson, M. "A Study in PhilOSOphy and the Social Sciences", Natanson, M. (ed.), PhiIOSOphy of the Social Sciences, New York: Random House, 1963. Nielsen, K. "Rational Explanations in History", Hook, 5. (ed.), Philosoohy and History, New York: New York University, 1963. Nowell-Smith, "Review", Philosophy, April, 1959. Passmore, J. "Review", Australian Journal of Politics and History, 1958. Popper, K. "The Demarcation Between Science and Metaphysics", Schilpp, P. (ed.), The PhiIOSOphy of Hudolph Carnap, La Salle, Illinois: Open Court, 1963. 316 . Potter, D. "Explicit Data and Implicit Assumptions in Historical Study", Gottschalk, L. (ed.), Generalization in the Writing of History, Chicago: University of Chicago, 1963. Randall, J. "Nature of Naturalism", Krikorian, Y. (ed.), Naturalism and the Human Spirit, New York: Columbia University, l9bb. Rescher, N. "Fundamental Problems in the Theory of Scientific Ex- planation", Baumrin, B. (ed.), Delaware Seminar in Philosophy of Science, Vol. II, New York: Wiley, 1963. Rescher, N. and Joynt, C. "Evidence in History and in the Law", Journal of PhiloSOphy, LVI, 1959. Rudner, R. "Comments", Feigl, H. and Maxwell, G. (eds.), Current Issues in the Philosophy of Science, New York: Holt, 1961. ________ "The SCientht Qua Scientist Makes Value Judgments", PhiIOSOphy of Science, XX, 1953. Scheffler, I. "Explanation, Prediction and Abstraction", British Journal for the Philosophy of Science, 7, 1957. Schutz, A. "Common-Sense and Scientific Interpretation of Human Action", Natanson, N. (ed.), Philosophy of Social Sciences, New York: Random House, 1963. "Concept and Theory Formation in the Social Sciences", Natanson, M. (ed.), Philosophy of the Social Sciences, New York: Random House, 1963. Scriven, M. "Explanations, Predictions and Laws", Feigl, H. and Maxwell, G. (eds.), Minnesota Studies, Vol. III, Minneapolis: University of Minnesota, 1962. . "New Issues in the Logic of Explanation", Hook, S. (ed.), Philosophy and History, New York: New York University, 1963. "Truisms as the Grounds for Historical Explanations", Gardiner, P. Theories of History, Glencoe, Illinois: Free Press, 1962. Simmell, G. "How Is Society Possible?", Natanson, M. (ed.), Philos- Ophy of the Social Sciences, New York: Random House, 1963. Strawson, P. "Review", Hind, April, 1959. Strout, C. "Causation and the American Civil War", History and Theory, Vol. I, 1961. Watkins, J. "Ideal Types and Historical Explanation", Feigl, H. and Brodbeck, M. (eds.), Readings in the Philosophy of Science, New York: Appleton-Century, 1953. ”'TITI'I‘NINfillflffliifliulfflill“