From owner-bhaskar@jefferson.village.Virginia.EDU Wed Jun 25 05:16:18 1997 Received: (from domo@localhost) by jefferson.village.Virginia.EDU (8.8.5/8.6.6) id FAA47119 for bhaskar-outgoing; Wed, 25 Jun 1997 05:15:55 -0400 X-Authentication-Warning: jefferson.village.Virginia.EDU: domo set sender to owner-bhaskar@lists.village.virginia.edu using -f Received: from keynes.econ.utah.edu (keynes.econ.utah.edu [128.110.114.5]) by jefferson.village.Virginia.EDU (8.8.5/8.6.6) with SMTP id FAA13065 for <bhaskar@jefferson.village.virginia.edu>; Wed, 25 Jun 1997 05:15:49 -0400 Received: from marx.econ.utah.edu.utah.edu by keynes.econ.utah.edu (5.x/SMI-SVR4) id AA09971; Wed, 25 Jun 1997 03:15:07 -0600 Received: by marx.econ.utah.edu.utah.edu (SMI-8.6/SMI-SVR4) id DAA05873; Wed, 25 Jun 1997 03:07:16 -0600 Date: Wed, 25 Jun 1997 03:07:16 -0600 Message-Id: <199706250907.DAA05873@marx.econ.utah.edu.utah.edu> From: Hans Ehrbar <ehrbar@marx.econ.utah.edu> To: bhaskar@jefferson.village.virginia.edu Subject: BHA: rts2, Appendix to Chapter 2 Sender: owner-bhaskar@jefferson.village.Virginia.EDU Precedence: bulk Reply-To: bhaskar@jefferson.village.Virginia.EDU Status: OR [127] Appendix ORTHODOX PHILOSOPHY OF SCIENCE AND THE IMPLICATIONS OF OPEN SYSTEMS It may be felt that I have dealt rather summarily in Chapter 2 with some of the most hallowed doctrines of received philosophy of science; so I want here to turn to a more detailed examination of them in the light of the phenomenon of open systems. The structure of orthodox philosophy of science is based squarely on the Humean theory of the actuality of causal laws. But it is convenient to give Nicod's criterion,1 which presupposes and implies it co-equal status. I shall formulate them as two principles: P_1, the principle of *empirical-invariance*, viz. that laws are or depend upon empirical regularities; and P_2, the principle of *instance-confirmation* (or falsification), viz. that laws are confirmed (or falsified) by their instances. Post-Humean philosophy of science has called into question only the sufficiency, not the necessity of these principles; i.e. it has left the ontology implicit in them intact. Thus in the most advanced recent positions theory is regarded as irreducible; and as supplying at least part of the grounds for laws. The significance of such modifications will be considered in the next chapter. Here, as in the body of Chapter 2, I will not distinguish between philosophers who regard P_1 and/or P_2 as necessary and sufficient and those who regard them as merely necessary. Once more no harm will be done by this conflation as all my objections here turn on the lack of necessity of these principles and ipso facto of the theories they sustain. It is for the sake of explanatory convenience and to avoid repetition 1 After J. Nicod, Foundations of Geometry and Induction, p. 219. See C. G. Hempel, Aspects of Scientific Explanation, p. 11. 128 A Realist Theory of Science that I formulate and discuss them in their `necessary and sufficient' form. Both P_1 and P_2 depend upon a closure, and hence upon the assumption of a simple undifferentiated reality. P_1 gives rise to the truth-functional concept of natural necessity. This is the idea that the logical status of laws can be explicated, at least in part, by the formula "for all x, fx inplies gx," where the predicates `f' and `g' are defined extensionally or are at least given some definite empirical interpretation e.g. by means of correspondence rules. The definition may be ostensive or operational;2 and if ostensive, either sensationalist3 or physicalist.4 P_1 is susceptible of descriptivist (Mach)5 and instrumentalist (Ryle)6 interpretations; and of classical empiricist and transcendental idealist ones. P_2 is susceptible of inductivist (Carnap)7 and falsificationist (Popper)8 interpretations; and of positivist and conventionalist ones. I am going to use the term `deductivism' to refer to the ensemble of theories erected on the basis of P_1 and P_2. My choice of the term `deductivism' may not seem an altogether happy one in view of the fact that it is meant to cover philosophers who have regarded themselves, as `inductivists' and `instrumentalists', as opposed to `deductivism'. However, I do not think that this is a serious difficulty. For there is no way in general of getting an inductive policy going without appealing to an antecedently formulated lawlike statement. (The exception is provided by the 2 P. W. Bridgman, The Logic of Modern Physics, esp. pp. 1-25. 3 Physics cannot be regarded as validly based upon empirical data until [light] waves have been expressed as functions of the colours and other sense-data', B. Russel, Mysticism and Logic, p. 109. 4 See e.g. O. Neurath, `Universal Jargon and Terminology', P.A.S. Vol. 41, pp. 127-48. 5 `The communication of scientific knowledge involves description: that is, the mimetic reproduction of facts in thought, the object of which is to replace and save the trouble of new experience. This is all that natural laws are', E. Mach, Popular Scientific Lectures, p. 192. 6 G. Ryle, The Concept of Mind, esp. pp. 116-22. Cf. also C. S. Pierce, Collected Papers, Vol. 2, p. 354; F. P. Ramsey, The Foundations of Mathematics, pp. 194-7; and S. Toulmin, The Philosophy of Science, Chap. 3. 7 R. Carnap, The Logical Foundations of Probability or `Testability and Meaning', Readings in the Philosophy of Science, eds. H. Feigl and M. Brodbeck, pp. 47-92. | 8 K. R. Popper, The Logic of Scientific Discovery, esp. Chap. 4 or Conjectures and Refutations, Chap. 10. Actualism and the Concept of a Closure 129 pure Humean case where the events are intuitively ascertainable atomistic instants, and each event is a member of a linear series.) For one sequence to give support to another the antecedent events must be alike in relevant respects. But to talk of the relevance of the `respects' already presupposes a tentative (conjectured) or confirmed law. Hence the inductivist in theory must be a deductivist in practice. Similarly a rule of inference can always be recast as the major premise of a syllogism. That is to say, an inference ticket (a `season ticket', as Ryle revealingly calls it)9 remains valid only as long as some empirical generalization is true. Thus the instrumentalist questions only the descriptivist interpretation of the use of laws, not their logical form. I will first set out the overall structure of deductivism before examining its components, individually and collectively, in the light of open systems. Underpinning deductivism is the actualist thesis that laws are relations between events or states of affairs. If the world consists only of atomistic events or states-of-affairs then for a general knowledge of it to be possible their relations must be constant. And so we have P_1 the principle of empirical-invariance (1) laws are or depend upon constant conjunctions of events or states of affairs (which constitute the objects of actual or possible experiences). As such it generates the familiar Humean theory of causality. This theory is susceptible of different interpretations, viz. (2)' as a theory of what we *mean* by saying `X causes Y', viz. that the events as specified under their descriptions X and Y are regularly conjoined; and (2)" as a theory of how we must be prepared to *justify* the claim that X causes Y, viz. by showing that the event as specified under these descriptions are regularly conjoined. In the case of both theories it is possible to substitute the weaker requirement that the events be specified under these or some other set of descriptions. The weaker variants could be indicated by the subscript_1. We then have a theory of explanation, that (3) events are explained by subsuming them under one or more universal laws; i.e. by deducing them from a set of one or 9 G. Ryle, op. cit., p. 117. 130 A Realist Theory of Science more universal laws, together with a statement of their initial conditions. This has become known as the Popper-Hempel theory of explanation. Popper was the first to restate it in modern times10 and Hempel has been its most systemic advocate and defender.11 It is convenient to divide the theory into two requirements: (3a) a *deducibility* requirement, viz. that the explanandum be deducible from the explanans; and (3b) a *covering-law* requirement, viz. that the explanans contain at least one universal law. Next, a theory of prediction to the effect that (4) events are predicted by deducing them from a set of universal laws together with a statement of their initial conditions. (3) and (4) together give rise to the theory that (5) explanation and prediction are symmetrical (in the sense of (iii) on page 63 above). On this view their difference lies merely in the fact that in `explanation' the explanandum event lies in the past and in `prediction' in the future. Then we have a theory of the explanation of laws, namely (6) laws are explained by subsuming them under or deducing them from more general, abstract or inclusive statements. Such statements may be regarded as theoretical principles or hypotheses. They may be interpreted descriptively, instrumentally or as fictions. I leave aside consideration of the various theories of theories until Chapter 3. Next a theory of the explanation of theories and even sciences, viz. (7) theories and sciences are explained by deductively subsuming them under more basic or general ones. The explained theory or science is then said to have been `reduced' to the explaining one.12 Thus we have a theory of the development of science, viz. (7)* science develops monistically or in a linear fashion so as to leave meaning and truth-value unchanged. Theories (7) and (7)* are rejected by many philosophers, most notably Popper, committed to other components of the 10 K. R. Popper, The Logic of Scientific Discovery, p. 59. 11 See esp. C. G. Hempel, op. cit., Chap. 12. 12 See E. Nagel, op. cit., Chap. 11. Actualism and the Concept of a Closure 131 deductivist structure. As Feyerabend has pointed out these theories generate their own restrictive methodology, embodied in the conditions that (7)** theoretical innovations should be consistent and meaning-invariant with respect to established, i.e. currently accepted, theory.13 These conditions tend inevitably to have a conservative effect. And they may be regarded as rationalizing the practice of what Kuhn has called `normal science'.14 I do not intend to discuss theories (7)* and (7)** in any detail here. That will be done in Chapter 3. But their connection through thesis (7) with the doctrine of actualism and their consequent dependence upon the presupposition of a closure should be clear. According to theses (3), (6), (7) and (7)*, the explanation of events, laws, theories and sciences all partake of essentially the same ideal `deductive-nomological' form. And they share this also (through theories (2), (4) and (5)) with the activities of prediction and the identification of causes. Scientific knowledge then must consist (in part or in whole) of a deductive structure. But which one? Which out of all possible deductive structures is the best (or in Popperian terms, `least worst') ? Here P_2 gets to work. Thus we have the theories that: (8) laws, theories and sciences are directly or indirectly confirmed or corroborated by their instances (which constitute the objects of actual or possible experiences); and (9) laws, theories and sciences are directly or indirectly falsified by their counter-instances (which constitute the objects of actual or possible experiences). Unless the meaning of theoretical terms is reduced to ostensively defined instances, as in Machian descriptivism, such criteria cannot be sufficient, but they are normally posited as at least necessary. (8) and (9) have sometimes been interpreted in a conventionalist way. But as what is regarded as `conventional' is only the decision to accept a report as being genuinely `observational' (and so capable of furnishing a genuine instance or counter-instance of a putative law), a conventionalist interpretation does not affect the status of the principle itself.15 13 p. K. Feyerabend, op. cit., p. 164. 14 T. S. Kuhn, op. cit. Chap. II. 16 Poincare is widely regarded as the founder of conventionalism in the philosophy of science. For a conventionalist interpretation of Popper see I. Lakatos, `Criticism and the Methodology of Scientific Research Programmes', Criticism and the Growth of Knowledge, eds. I. Lakatos and A. Musgrave, pp. 104ff. 132 A Realist Theory of Science Finally, following on from (9), we have a maxim of scientific practice to the effect that: (10) scientists should, in formulating their theories, state quite unambiguously the empirical conditions under which they are prepared to reject them (cf (iv) on pp. 63-4 above). This may also be taken as a criterion of what it is to be `scientific' and `unscientific', viz. (10)* to be `unscientific' is not to be prepared to state such conditions or having done so to revise them ex post facto.16 Most received philosophy of science is based on a core extracted from theories (1)-(10).* These theories all share one great weakness: they all presuppose a closure. If, as I have argued, the world is in fact open then they must all be more or less drastically revised and in some cases completely rejected. I have already argued against (1) in Chapter 2 (especially section 4) above so I will not discuss it separately here. It will be remembered that the criterion of open-ness is the non-invariance of empirical relationships. Now clearly if the law-like statements whose antecedents are instantiated in open systems are interpreted as invariant empirical regularities they must be regarded as false. But this means that there can be neither laws, because there are no invariant empirical regularities; nor theories, because they are continually being falsified; so that neither explanation nor prediction can be given a rational basis. I have examined two actualist responses to this predicament (viz. weak and strong actualism) and showed how neither can sustain the concept of laws applying transfactually, viz. in open and closed systems alike, that we need to render intelligible both the experimental establishment and the practical application of our knowledge. It is important to keep the deducibility and covering-law requirements, as expressed in (3a) and (3b) distinct. For either can be non-trivially satisfied without the other. I have already shown in 2.6 how one can have explanation in terms of a network of normic statements (which may be strictly universal in the 16 See e.g. K. R. Popper, Conjectures and Refutations, Chap. I and passim. Actualism and the Concept of a Closure 133 sense of space-time-invariant) without the event being deducible. Of course the sense in which the covering-law `covers' in this case is different. In the same way but from another aspect, the deducibility requirement may be violated in the development of science although all the statements involved are universal. An example of this is given by the way in which Newton's theory both explained and corrected Kepler's and Galileo's laws.17 In these cases (3b) is satisfied but not (3a). On the other hand our atavistic causal explanation `Tania pushed the door open' is deductive, though no laws are involved. It is also possible for the deducibility requirement to be satisfied by statements mentioning named individuals or specific space-time regions.18 It should be noted that one can have deductive `explanations' of events under transient empirical regularities but not deductive explanations of their explanations: for the space-time restriction cannot itself be derived from a strictly universal law. Thesis (9) implies that all law-like statements whose antecedents are instantiated in open systems are false. It is therefore as stated quite useless as a decision rule for choosing between different law-like statements (or as an `organon of criticism').19 One way of dealing with this would be to allow theory a role in grounding laws. But with an unchanged ontology this is bound to be ultimately unsatisfactory. Of course if we possess a good theory it is irrational to relinquish it in the face of recalcitrant facts - without a better one. But our justification for holding on to the theory must be that it might eventually be able to explain them (by suitable modifications, refinements or developments) or show that they are not after all facts (i.e. that the statements used to state them are untrue). If it could never explain and/or correct them this justification would collapse. Of course the sufficiency of Nicod's criterion must be disputed: the grounds for a law or theory cannot be exclusively empirical. Theory must supply some idea of a `connection', without which it would be impossible to tell necessary from accidental sequences. But in an open world Nicod's criterion cannot 17 See e.g. P. Duhem, The Aim and Structure of Physical Theory, Chaps. 9-10. 18 Cf. A. Donagan, `The Popper-Hempel Theory Reconsidered', Philosophical Analysis and History, ed. W. H. Dray, pp. 136-8. 19 K. R. Popper, Objective Knowledge, p. 21 and passim. 134 A Realist Theory of Science be necessary either: the grounds for a law or a theory cannot be undifferentiatedly empirical. For the conditions must normally be carefully controlled so that a hypothesis about the connecting mechanism can be put to a fair test. If it is wrong to regard law-like statements and theories as being falsified by the non-occurrence of their consequents in open systems, it is equally wrong to regard them as being confirmed or corroborated by their occurrence in an undifferentiated way, i.e. independently of the context in which the putatively falsifying/confirming instance occurs. Theses (9) and (8) must therefore be restated so as to place a restriction on the system in which a genuinely falsifying/confirming instance occurs, viz. that it be closed. Not all evidence is equal; or rather not all evidence is evidence for or against a law. In general it takes a closed system to furnish evidence capable of falsifying or confirming a law. And within the class of closed systems, experimentally closed ones are preferable. For experimentally we can test and re-test a greater number and variety of subjunctive conditionals of the form `if x were to take on a certain value, then y_j would occur' by instantiating their antecedents. Whereas outside the laboratory we are restricted to observing whatever sequential performances nature is obliging enough to put on. Similar considerations apply to thesis (10). One cannot lay down hard and fast criteria spelling out beforehand which observable situations `if actually observed mean the theory is refuted' (see (iv) on pp. 63-4 above). For one can never know beforehand whether the system will be actually closed. On the other hand the closure of the system is not a part of the observable situation; so that it cannot be incorporated into the criterion of scientificity without destroying it. The judgement that the system is closed can only be made ex post after we have observed (and theoretically assessed) the observable situation. Theses (3) and (9) postulate a syntactical identity between explantion, prediction and falsification in that, taken together, they imply a correct prediction explains and an incorrect prediction falsifies. They depend upon the assumptions that it is possible to give a purely syntactical account of scientific Actualism and the Concept of a Closure 135 activities and that these activities always occur in the context of an antecedently given closure. Recent philosophy of science has clearly demonstrated the poverty of the former; it is with the incorrectness of the latter assumption that I am here concerned. Now explanation in open systems, failing the attainment of an antecedent closure, normally requires, as I have pointed out in 2.6 above, retrodiction; that is the inference from present effects to prior (perhaps hidden, perhaps just unrecorded) causes, via the application of normic statements. Now the significance of this activity is that it presupposes a non-conventional division of the class of law-like statements into those which are and those which are not capable of functioning in this way, i.e. into those which are accepted (for transfactual application) and those which are still under test. Now once we allow this the postulated symmetries between explanation and prediction and explanation and falsification break down. And it becomes important to distinguish between two kinds of prediction conflated in deductivism's syntactical account of science: practical predictions of categorical form which are rarely made in science but which are important in some of its practical applications in open systems and about which the applied scientist can never be deductively certain; and test predictions of hypothetical form made under effectively closed conditions in order to test a theoretical hypothesis or putative law. It is easy to see why the explanation/falsification symmetry collapses once we allow the legitimacy of retrodiction. For the activity of retrodiction presupposes the truth and applicability of the law used; the possibility that it is false is ruled out a priori. Now the intelligibility of falsification depends upon the idea that the would-be falsifier has independent grounds for the occurrence of the initial conditions. If the legitimacy of retrodiction is denied a vicious regress back to sense-experience ensues. Thus suppose we have a law-like statement of the form S_1 `whenever events of type E_0 occur events of type E_1 occur'. For S_1 to be used to explain the occurrence of E_1 in a way which is consistent with the idea of its being subject to falsification independent grounds for E_0 are required, say G_0. But the connection between E_0 and G_0 is itself a contingent causal one, 136 A Realist Theory of Science which may be represented by the hypothesis S, `whenever events of type G_0 occur events of type E_0 occur'. Hence we need independent grounds for the occurrence of G_0, say G'_0, if our use of S_2, is to be consistent with the idea of its falsifiability. But as G'_0 stands in a contingent causal relationship we need independent grounds for it too, and so on.... There is of course only one connection with E_0 which, being non-causal and non-contingent, does not require independent grounds, namely immediate sense-experience. Thus insistence on independent grounds for the initial conditions of an explanation, which is an inevitable consequence of the idea of its susceptibility to falsification, inevitably leads to the requirement that the initial conditions be apprehended in sense-experience; in which case the event could have been predicted. The root of the trouble here is that the causal relationship is taking too much strain: it is required both to be contingent (and as such to be subject to falsification) and to explain; functions that it cannot combine without vicious regress to sense-experience. Once we distinguish between open and closed systems, however, this regress can be avoided. For we may allow that events may be explained in open and closed systems alike, but that law-like statements may only be falsified under effectively closed conditions (where deductive test predictions are possible). Now the point of the explanation/prediction symmetry thesis is vitiated in open systems. For we can give excellent explanations, in virtue of the transfactual applicability of our independently validated knowledge, where we are incapable of any predictions (save perhaps of the most immediate or the most tentative sort). Moreover the kinds of statements involved in the two activities are radically different: explanation proceeding by way of normic, and prediction by way of empirical, statements. An empirical generalization typically merely generalizes the problem to be explained, whereas a normic statement locates it in the context of an explanatory theory. On the other hand, normic statements may be inferior predictors to the most crude generalizations or rules of thumb. Further, the occurrence of the event itself may be a practically necessary condition of our knowledge of the former state of the system, as in the case of the collapse of a bridge or an airplane crash, or even of the kind of system with which one is concerned, as in the case of the Actualism and the Concept of a Closure 137 sudden onset of uncontrollable hysteria.20 Again, the intelligibility of much practical science of an exploratory kind, such as prospecting for oil, depends upon the existence of a radical asymmetry between explanation and prediction. Because of such difficulties defenders of the symmetry thesis have been forced to modify it so that it requires only that were we to be in possession of all the information available at the time of the explanation then we could have predicted it.21 Now I have argued in 2.6 that it is possible to give a complete explanation of an event without thereby being in a position to deduce it, namely if the different generative mechanisms at work are of radically different kinds; so that the reformulated symmetry thesis is either false or, if deducibility is built into the definition of `explanation', uninterestingly tautologous.22 One further point on prediction. We are only deductively justified in predicting an event if the system is closed. But there is no way of knowing in advance (at the only time when a prediction is relevant) whether the conditions for a closure will in fact be satisfied. Hence the probability of an event's occurrence can never be 1. 20 Cf. M. Scriven, `Explanation, Prediction and Laws', Minnesota Studies in the Philosophy of Science, Vol. III, eds. H. Feigl and G. Maxwell, esp. pp. 176-90. 21 See e.g. A. Grunbaum, `Temporally Asymmetric Principles, Parity between Explanation and Prediction and Mechanism versus Teleology', Philosophy of Science Delaware Seminar, Vol. I, ed. B. Baumrin, p. 73. 22 The interesting question is then of course shifted to that of whether complete explanations must be `deductive'. Most of the early objections to the deductive model turned on the non-availability of generalisations connecting events like the cracking of radiators or missile failures. Although this was no doubt encouraged by the way in which its advocates presented it, it was somewhat disingenuous of its critics not to realise the possibility of sophisticated reformulations of it. Mandelbaum, for example, has argued correctly that such events must be regarded as complex and analysed into components (M. Mandelbaum, `Historical Explanation: The Problem of Covering Laws', History of Theory, Vol. I, pp. 229-42). However he still sees explanation as depending upon a knowledge of laws (which he interprets in the Humean way) covering the component events (ibid., p. 241); and given this, the complex event itself still remains deductively predictable. I have argued, by contrast, that the laws covering the components are normic and that they may involve reference to radically different kinds (so that they cannot be incorporated within a single theory). Hence the complex event, even though completely explained, may not be deducible. 138 A Realist Theory of Science The undifferentiated ontology of received philosophy of science results in the very damaging view expressed in (v) on page 64 above, viz. that the acid test of a theory is its predictive power. On this view the more accurate a theory's predictions - no matter of what or where - the more worthwhile it is retaining. Coupled with permissiveness over the use of the CP clause, such a position provides a powerful rationale for scientific conservatism of any school. It has been used as such by Heisenberg in physics and Skinner in psychology, by Friedman in economics and Osiander in astronomy. But armed with our concept of a complex and differentiated reality we can see what is wrong with it. For, on the one hand, there will always be more than one hypothesis capable of saving any given set of facts, so independent tests for them will always be necessary; and, on the other, it is only under closed conditions that such tests can be decisive. Consistency with the facts is neither necessary nor sufficient for a theory. Popper does not seem to see the connection between the criteria of explanation and rationality he expouses and the historicist view of social science he condemns. That there is a connection is clear. For if we know that power corrupts and regard this as a true, if trivial, empirical generalization then given only a knowledge of the initial condition that N is powerful we can predict with deductive certainty that N will be corrupted. On the other hand if we cannot know when the initial condition is satisfied the law cannot be applied and so is quite useless for either explanatory or social engineering purposes. If Popper is committed to thesis (3) in social science he is committed to thesis (4) and so to the historicist view of science as the prediction of events (savoir, pour prevoir). Popper nowhere denies the applicability of thesis (8) to social science.23 But he equivocates between a conception of historicism as the view that the aim of the social sciences is the prediction of future events24 and the view that its aim is to make unconditional historical prophecies.26 23 On the contrary he repeatedly emphasises the essential similarity in the logical form of the natural and the social sciences. According to him what is peculiar to the latter is its subject matter, and to history its interest in the particular (see K. R. Popper, The Poverty of Historicism, p. 143). But the pattern of explanation is the same. 24 See e.g. ibid., p 3. 25 See e.g. K. R. Popper, Conjectures and Refutations, p. 339. Actualism and the Concept of a Closure 139 But the authors he attacks did not make unconditional historical prophecies: Hegel in fact made no predictions, and Marx only conditional ones. Popper's real argument is not against predictability as such in social science, but against the predictability of certain kinds of social events, viz. large-scale social changes and their consequences. It is an argument against certain theories of social becoming which he interprets historistically. Indeed one could almost say it is an argument for an a-historical form of historicism (in which the laws involved are regarded as strictly universal) against an historical one (in which they are regarded as spatio-temporally restricted). Historicism is in general invalid in all its forms. And it is invalid in both natural and social science. And for exactly the same reasons. Incidentally, this is true for the refutation of historicism that Popper derives from the logical impossibility of predicting the precise effects of future knowledge in as much as any such prediction depends upon a knowledge of that knowledge.26 For its effects on nature are no less predictable than its effects on men.27 I turn now to the central unifying theory of explanation, viz thesis (3) and the theories of causality, viz. (2)' and (2)", that underlie it. Consider once more the paradigm of the kind of context in which a causal claim is made. `Why is the door opening?' `Because Tania's pushing it'. Now it is certainly not the case that in saying `Tania's pushing it caused the door to open' we mean that every time Tania pushes it the door opens. For there are times when it is locked and times when she must turn the door knob too. (Nor equally is it the case that every time the door opens is it because Tania pushes it.) But neither could we produce any universal law which would show why the door opened in this particular case. Generally, theories (2)' and (2)" are only plausible if in the case of (2)' we mean to imply and in the case of (2)" we have grounds for supposing that a closure has been obtained of the system in which the events occur. Without this we cannot possibly mean nor can we reasonably be 26 K. R. Popper, Poverty of Historicism, pp. vi-vii. 27 Indeed one could even go so far as to say that the latter depends upon the former; that is, that it is only in so far as the effects of future knowledge are unpredictable on nature than they are unpredictable on men, in as much as all human action has a physical aspect but the converse is not the case. A social closure presupposes a natural one but not vice versa; there could be a nature without men, but not men without a nature. 140 A Realist Theory of Science committed to showing that the events which we claim are causally connected are so in virtue of being constantly conjoined. In short the Humean theory cannot be a general theory of causality but at best a theory of what is involved in the making of causal claims in closed systems and where the events are separately identifiable atomistic instants. Davidson has proposed a sophisticated reformulation of the Humean theory. `It is an error', he says, `to think that no explanation has been given until a law has been produced. Linked with [it] is the idea that singular causal statements necessarily indicate, by the concepts they employ, the concepts that will occur in the entailed law. Suppose a hurricane which is reported on page 5 of Tuesday's Times causes the event reported on page 13 of Wednesday's Tribune. Should we look for a law relating events of these kinds? It is only slightly less ridiculous to look for a law relating hurricanes and catastrophies. The laws needed to predict the catastrophy with precision would, of course, have no use for concepts like hurricanes and catastrophies'.28 Davidson's analysis of `singular causal statements' is as follows: ` "A caused B" is true if and only if there are descriptions of A and B such that the sentences obtained by putting these descriptions for "A" and "B" in "A caused B" follows from a true causal law'.29 The objection to this suggestion is that it places a requirement on the verification of causal claims which is (a) impossible, (b) useless, and (c) unnecessary. For, on it, `A' and `B' stand in, in the original causal claim, for the complete atomistic state-descriptions that would form the antecedents and consequents of such a causal law. But it is presumably only because (at least from a Humean viewpoint) we have not got and perhaps cannot get such state-descriptions that we make the original causal claim in the first case. Davidson holds that such claims are defended by `producing a relevant law or giving reasons for believing such exists'.80 How? then would one set about defending a claim about the causes of the French Revolution? By giving grounds for believing that a true 28 D. Davidson, `Actions, Reasons and Causes', Journal of Philosophy 1963, reprinted in ed. A. White, Philosophy of Action, p. 92 29 Ibid, p. 92. See also D. Davidson, `Causal Relations', Journal of Philosophy 1969, p. 701. 30 D. Davidson, `Actions, Reasons and Causes', p. 92. Actualism and the Concept of a Closure 141 `neurological, chemical or physical law'31 exists ? But such laws, were they to be known, would cover any set of historical causes of the French Revolution; in short, they would not enable us to discriminate between true and false causal claims at the level we are concerned with; that is, at the level at which the initial causal claim is made. Neurological laws are consistent with any social event and so cannot be possibly used to defend specific causal claims involving people. Worse still, in the end the only way of defending the belief that a Humean causal law exists covering some particular case will be by appealing to the truth of regularity determinism. Hence the defense of the most specific causal claim becomes an expression of faith in the philosophical dogma that the world is so constituted that the simple formula `same cause, same effect' everywhere applies. On the other hand, once we deny the premise that a causal claim entails a Humean causal law, and hence a closed system, these absurdities can be avoided; and we can then allow that causal claims are defended, where they require justification in terms of general statements at all, by an appeal to the normic statements of the level at which the original causal claim was made. Hempel makes a similar mistake in arguing that singular causal statements of the `q because p' (man bites dog) type `claim by implication', `tacitly presuppose', or `assert by implication' the existence of covering laws.32 Now it is not clear from his account whether he regards this as part of the meaning of the original causal claim (cf. thesis (2)' above) or merely as indicating the way in which it is to be justified (cf. thesis (2)' above). If the former we have the absurdity that `man' and `dog' change their meaning when they are used in the making of a causal claim. If the latter a confusion between what it is necessary to know to ascribe causes and what is necessary for our knowledge of the ascription of causes to be possible, viz. that certain underlying laws hold. The latter is neither necessary nor sufficient for the former. The non-availability of Humean causal laws is undoubtedly an embarrassment to the modern Humean. Inevitably he falls back on the idea that our explanations are sketches to be filled out in the fullness of time (cf. strong actualism) or that they are 31 D. Davidson, ibid., p. 93. 32 C. G. Hempel, op. cit., pp. 360, 362 and 363. 142 A Realist Theory of Science subject to an implicit ceteris paribus clause (cf. weak actualism). In addition he may relax the stringencies of the deductive model by allowing other categories of explanation: such as statistical, elliptical and partial formulations.33 Statistical explanations, however, spoil the point of the deductive model: for, as Scriven has put it, `they abandon the hold on the individual case'.34 Moreover there is no a priori reason to suppose that the world is not statistically open.35 As for the other categories: all explanations, contextually speaking, are elliptical; and the deductivist must show how partial explanations can be universally applicable. At the methodological level one of the most unfortunate consequences of the spell of actualism is the blurring of the real differences that exist between the various sciences: both in their subject matters and the degrees to which they have achieved knowledge of them. The experimental sciences have been able as a result of theoretical endeavour and technical ingenuity, to carve out a chunk of the uncontrolled world and use it as an object of inquiry. The non-experimental scientist has no such easy access. Now the view of the world as closed sets him off in the wrong direction - for it sets him looking for a complete description of a given field. A view of the world as open can, on the other hand, concentrate his endeavours enormously. For it means that all he has to do is to identify and describe (in ways to be considered later) some interesting and significant object of inquiry, without supposing that this will enable him to make deductively successful predictions. 33 C. G. Hempel, op. cit., pp. 376ff and pp. 415ff. 34 M. Scriven, `Truisms as the Grounds for Historical Explanation', p. 465. 35 The question of the closure of systems is distinct from that of their statistical or non-statistical properties. The latter turns on the deducibility of single as distinct from mass events; whereas the former turns on the stability of empirical relationships. Clearly statistical systems can be open or closed. --- from list bhaskar@lists.village.virginia.edu --- .