Popper, Induction and Falsification Author(s): Gary Jones and Clifton Perry Source: Erkenntnis (1975-), Vol. 18, No. 1 (Jul., 1982), pp. 97-104 Published by: Springer Stable URL: http://www.jstor.org/stable/20010796 Accessed: 25-07-2017 12:43 UTC JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://about.jstor.org/terms Springer is collaborating with JSTOR to digitize, preserve and extend access to Erkenntnis (1975-)
GARY JONES AND CLIFTON PERRY POPPER, INDUCTION AND FALSIFICATION The typical Popperian response to the Duhemian problem concerning the impossibility of ever falsifying, and thus isolating, a scientific hypothesis is usually framed in terms of what we take as the hypothesis under test as opposed to theories assumed to be true for testing purposes. As such, Popper's response does not attack the logical problem posed by Duhem at the level of logical possibilities but rather at the level of scientific practice. Popper's response stands in curious contrast to the type of remarks prom? ulgated by Popper anent induction. Rather than speak of inductive prob? lems in the arena of scientific practice, Popper usually reverts to talk of the a priori possibility that positive or confirming evidence notwithstanding, a universal hypothesis might be falsified. In what follows we shall suggest that this contrast of attitudes towards the impossibility of falsifying and the possibility of verifying actually makes it impossible for Popper to effect a complete rout of Duhem's problem or of induction, that any stratagem taken to resolve the Duhemian problem enhances induction. Finally, it will be argued that the successful application of Popper's hypothetico-deduc tive method of explanation presupposes that nature is uniform. Hence for this reason as well, falsification and induction are of the same epistemic status. I According to Duhem, there obtains an intimate logical connection be? tween scientific theories and scientific hypotheses. Scientific theories con? ceived in terms of the "economy of thought", i.e., as abstractions of the set of germane laws, are such that they are presupposed in the deduction of any statement to be compared with observation statements.1 Not only is the entire theoretical system presupposed in the deduction of a given state? ment but also the comparison of the observation statement with the de? rived statement involves the use of instruments the intelligibility of which entails reference to theories not intentionally tested. Not only might such instruments be faulty but the theories which justify their employment may in fact be incorrect. Consequently any derived statement follows not from Erkenntnis 18 (1982) 97-104. 0165-0106/82/0181-0097 $00.80 Copyright? 1982 by D. Reidel Publishing Co., Dordrecht, Holland, and Boston, U.S.A.
98 GARY JONES AND CLIFTON PERRY an isolated hypothesis but rather from the hypothesis conjoined with other general statements of the same nature. Any conflict between the derived statement and the observation statement only indicates that something is wrong, but not which hypothesis is at fault. To deny the consequent of a conditional is to deny the entire antecedent, even if the antecedent is a conjunction. It would, of course, be to no avail to argue that when conflict did arise, each hypothesis might be tested in turn so as to determine which one was faulty and thus responsible for the conflict. Such a proposal would, accord? ing to Duhem, be impossible to actualize as each hypothesis tested would be in exactly the same doubtful position suffered by the initial hypothesis. The test of each hypothesis would, that is, involve not only those other equally suspect hypotheses immediately associated with the initial hypoth? esis, but also those connected with the hypothesis in question by virtue of the instruments employed in the test. If, therefore, conflict between the derived statement and the observation statement indicates merely that some hypothesis is incorrect but not which hypothesis, then a given hypoth? esis can never, pace Popper, be falsified. If a given hypothesis can never be falsified, then falsificationism is for logical reasons inadequate as a methodology for getting us closer to the truth in science. Popper's defense admits that although it is logically possible to attribute responsibility anywhere within the system, it is not logically necessary to do so. Since logical possibility is only a necessary condition for attributing responsibility for the conflict between the description of the event and the predicted event, the sufficient condition must rest beyond mere considera? tion of logic. In other words it might well be argued that the formal possibility of not attributing responsibility for theoretical conflict to any given hypoth? esis does not necessitate theoretical indecision. It may very well be the case that we shall have to appeal to something beyond mere logical con? siderations in order to attribute responsibility, e.g., methodological con? siderations of the sort consonant with falsificationism. Thus, although there may obtain no logically conclusive reason for suspecting any given hypothesis responsible for a conflict between a derived statement and an observation statement because of the logical possibility that responsibility may reside elsewhere in the theoretical system, it does not follow that, on methodological grounds, good reasons as opposed to logically conclusive ones could not be proffered ascribing responsibility to a given hypothesis.
POPPER, INDUCTION AND FALSIFICATION 99 What, of course, constitutes good reasons will depend upon the particular methodological approach involved. It is, according to Popper, the prior determination of a priori improbability which methodologically designates one hypothesis rather than another as responsible for the conflict.2 This response acknowledges the logical possibility that the cause of the conflict may reside anywhere within the system but then narrows the field of rational or real candidates in terms of the prior determination of falsifi ability. Indeed, according to Lakatos, the sophisticated falsificationist ac? knowledges that logically the falsifying instance reflects negatively upon no one part of the body of science.3 Nevertheless, the sophisticated falsifi? cationist takes the 'modus tollens' to focus upon the most highly falsifiable portion of the antecedent. The methodologically accepted rule for replace? ment, in this case, specifies only that the newly employed hypothesis should not cause a degeneration (a reduction in empirical content) in the 'research program.' II Irrespective of whether or not the Popperian appeal to methodology would satisfy those troubled by Duhem's problem, the important point is that Popper takes the appeal to methodological considerations in response to a priori problems arising in scientific testing as being at least appropriate, if not adequate. The issue then becomes whether or not this attitude is con? sistent with that manifested in regard to another a priori problem arising within the context of scientific testing. The problem of induction has often been treated by Popper as an issue completely divorced from consider? ations of methodology. He has maintained that because of the universal nature of a general hypothesis, no finite quantity of evidence can ever serve to conclusively demonstrate the truth or even probable truth of the hypoth? esis.4 Nonetheless, an inductivist could surely admit that there obtains a log? ical gulf between the truth of a general hypothesis and that of evidence statements, and yet note that this fails to show that we cannot have 'good' reasons for supposing a given hypothesis to be true or probably true.5 It might be contended that positive evidence, in some pre-determined quan? tity, does serve as 'good', although not logically conclusive, reasons for accepting a given hypothesis.
100 GARY JONES AND CLIFTON PERRY III What constitutes 'good' reasons with respect to both problems of scientif testing, i.e., the problems of verification and falsification, are the same i both cases. With regard to the problem of falsification, although it migh be admitted that no logical guarantee could be given that the responsibilit for conflict did not rest elsewhere in the theoretical system, there obtained no empirical evidence to show that it might. The mere logical possibility that the cause of conflict could reside with something other than the hypoth? esis under test is not grounds for supposing that in fact the cause of conflict rests with something other than the hypothesis under test. Give that we either have some reasons for holding the hypothesis under test more suspect than the remainder of the theoretical system or that we hav no sufficient grounds for doubting anything other than the hypothesis itself, we have good reasons for supposing the hypothesis responsible fo the conflict. Similarly, in the case of induction although there is no logical guarante that future testing of the hypothesis will not vitiate the hypothesis' goo standing, the mere logical possibility that such a state might obtain is insufficient to cast doubt upon the truth or possible truth of the hypothesis. Given that logical possibility is not sufficient for accepting or rejecting the hypothesis, we have good, though not conclusive, reasons for accepting th hypothesis. IV Regardless of whether or not anyone might be persuaded by either of above remonstrations, it seems clear Popper wishes to both embrace reject the appeal to empirical or methodological considerations when te ing an hypothesis. To be consistent it is not enough to allow a prior considerations to dominate when the truth of an hypothesis is at issue allow empirical considerations to have hegemony when the falsity of hypothesis is at issue. It would not suffice to counter that the logic of general statements allow for refutation but not verification. It is not just a general statement (hypot esis) we are concerned with in scientific testing. Rather our concern such matters is, if Duhem's 'weak' thesis, i.e., the inability of 'modus lens' to isolate a hypothesis for falsification, is correct, with entire theore
POPPER, INDUCTION AND FALSIFICATION 101 ical systems. If Duhem is indeed correct, then the logical problems with falsification are no less troublesome than those which effect verification. If a methodological appeal to 'good' reasons allows us to extricate ourselves from Duhem's problem, then such an appeal should be permitted to allow us to escape the problem of induction. Therefore, if the above is not mistaken, Popper must, if he is to me? thodologically allow for falsification in the face of obvious a priori problems, also methodologically allow for induction. If, on the other hand, Popper continues to impugn induction on a priori grounds, then he must give up hope of ever showing any hypothesis false. The ramifications of the above dilemma are twofold. If Popper con? tinues his invective against induction on a priori grounds, then he has to accept the logical inadequacy of 'modus tollens' reflecting unfavorably upon any isolated section of the antecedent within which the entire body of scientific knowledge is mentioned. But if no isolated hypothesis can, with? out an extralogical appeal, be falsified, then, at least at that level, no one hypothesis can be more falsifiable, i.e., have more a priori improbability or more empirical content, than another. Unfortunately, this would make Popper's notion of theory replacement self-negatory. If, logically speaking, all theories are on par with respect to falsifiability, then we cannot replace one theory with a more falsifiable one nor can any addition to our research program be seen as a progressive or degenerating move. This, pace La katos, renders the 'strong Duhem thesis,' i.e., there are no rational grounds for theory replacement, imminent. On the other hand, if Popper is to maintain that we can falsify a given hypothesis, then he will have to appeal to extra-logical considerations, e.g., considerations of methodology. But, as was argued above, to be consistent Popper will have to allow that methodological considerations are relevant to the logical problems circum? scribing the confirmation of an isolated hypothesis. This places falsifi? cation and induction on a methodologically equal footing. V In the remaining section, the claim that induction and falsificat the same epistemic status will be argued for in slightly differen Deduction, as utilized by Popper, is applied to empirical science following way. Scientific theories are proposed in the form '(x) (F
102 GARY JONES AND CLIFTON PERRY a statement which is universal as regards space and time.6 As such, these theories constitute the conjecture of individual regularities that may or may not accurately describe the world. Individual observation statements are then derived as predictions from the theories. Positive instances of the theory are not, following Hume, indicative of its truth value. Indeed, Popper would maintain, that in deductive terms any inference from posi? tive instances to the probable truth of a universal theory would at best constitute the fallacy of affirming the consequent. The only valid deductive inference that can be made on the basis of observation statements concern? ing the truth value of scientific theories is to their falsity. A necessary condition of truth of a universal theory is that every one of its instances be true. Hence, one false instance is sufficient for its falsity. Thus 'modus tollens' is the primary inference rule in the application of deduction to empirical science. Because of his rejection of induction, however, Popper has been accused of failing to provide an adequate basis for theory preference. It has been argued that unless Popper subscribes to an inductive principle, he cannot justify our choice of a theory that is well supported by the evidence over a theory that has no evidential support at all.7 There is reason to believe, however, that the difficulty lies at an even more fundamental level within Popper's methodology. As we have seen, for Popper a scientific theory or law is of the form '(x) (Fx D Gx)' where the antecedent expresses the initial conditions or cause, and the consequent expresses the effect. Such counter factual conditionals are, as noted above, universal. In the following sense it could be charged that Popper's method itself embodies the assumption that nature is regular. To qualify as a scientific theory, a statement must be universally quantified and thus express a fixed relationship between sorts of phenomena regardless of space or time. Individual theories then are 'complex' conjectures in the sense that they are statements as to in what particular ways nature is uniform; the presupposition, embodied in the methodology from which they are derived is that nature is in fact uniform. It is thus evident that a presupposition of the acceptability of theories such as '(x) (Fx D Gxy \x) (Hx D Ix)\ etc. is that theories of this form are proper objects of scrutiny. No amount of tests of individual theories could constitute a test of the above presupposition. In light of Hume's problem, it would seem necessary that Popper justify this feature of his method. The problem consequently manifests itself in Popper's notion of causal
POPPER, INDUCTION AND FALSIFICATION 103 explanation. A causal explanation consists in deducing the description of a singular event from a conjunction of universal laws and initial conditions.8 Unless all events of a certain type can be explained in this manner, no explanation of a single instance is acceptable. Popper's schema thus im? plies that causal explanations constitute a statement as to what particular regularities exist. As such, Popper's scheme assumes that nature is in fact regular. This is indicated by the fact that although individual causal expla? nations can be refuted, the assumption that there can and should be causal explanations of the sort Popper requires could never be refuted. In other words, the claim that there is reason to believe that A causes B presupposes an assumption that events are related to one another in terms of causally sufficient conditions. The latter concept in turn implies a fixed relationship that is independent of space and time, i.e., a regularity. Given Popper's methodology, it is above presupposition that it can never be subject to test, in much the same way that an inductive principle would be presupposed in a scientific method. The upshot is that it is not justifiable to apply deductive logic - con? sidered as the organon of rational criticism, to empirical science unless it is assumed that empirical science is a proper subject of rational criticism. If one did not assume that nature was uniform, one would have no confidence in the conclusions of deductive arguments purporting to describe its structure as science does. If, on the other hand, one assumed that nature was irreg? ular or had no preconceptions at all, one would have no basis for any decisions regarding the structure of nature, no matter how tentative. Being committed to the thesis that deduction is the organon of rational criticism, that is, does not constitute grounds for rejecting scepticism. Many philos? ophers, e.g., Descartes, Hume, and Duhem have argued that although deduction is the only acceptable method of inference, it cannot be success? fully applied in scientific inquiry without a presupposition that nature is uniform. The difference between Descartes, Hume, and Duhem on the one hand, and Popper on the other is that Popper believes that in the absense of assuming the uniformity of nature, the notion of severe tests makes sense. As we have seen, however, it is just this belief which stands in need of justification.
104 GARY JONES AND CLIFTON PERRY VI The overall conclusion of this essay is the following. If induction does n stand in need of justification, then there is no need for Popper's metho which was initially proposed as a replacement for induction. If on the other hand, a justification of induction is nonetheless necessary, then a simil justification is needed for Popper's method. Thus Popper's method viable only if induction is viable as well. But if induction is viable, th Popper's method is again unnecessary inasmuch as it was initially pro posed as a replacement for induction. Therefore, Popper's method is un necessary. University of San Diego (G.J.) Auburn University (CP.) NOTES 1 Duhem, Pierre: 1974, The Aim and Structure of Physical Theory, Atheneum, New York, pp. 183-188, 204-205. 2 Popper, Karl: 1974, 'Replies,' in The Philosophy of Karl Popper, ed. P. A. Schilpp, Open Court, La Salle, p. 1035. 3 Lakatos, Imre: 1970, 'Falsification and the Methodology of Scientific Research Program? mes,' Criticism and the Growth of Knowledge, ed. I. Lakatos and A. Musgrave, Cambridge University Press, Cambridge, pp. 184-189. 4 Popper, Karl: 1972, Objective Knowledge, Clarendon Press, Oxford, pp. 1-32. Popper, Karl: 1963, Conjectures and Refutations, Harper Torchbooks, New York, pp. 33 59. Popper, Karl: 1959, The Logic of Scientific Discovery, Harper Torchbooks, New York, Section 89. 5 Salmon, Wesley: 1966, The Foundations of Scientific Inference, University of Pittsburgh Press, Pittsburgh. 6 The Logic of Scientific Discovery, Chapter III, op. cit. 7 Cf. Salmon, Wesley: 1967, 'The Justification of Inductive Rules of Inference', in The Problem of Inductive Logic, ed. I. Lakatos, North-Holland, Amsterdam. Ayer, A. J.: 1956, The Prob? lem of Knowledge, Penguin Books, Harmondsworth, pp. 73-74. Gr?nbaum, A. : 'Is the Meth? od of Bold Conjectures and Attempted Refutations Justifiably the Method of Science?' Br. Journal Philosophy of Science 27, 105-136. Jones, G. E.: 'Popper and Theory Appraisal' in Studies in History & Philosophy of Science 9, 239^49. 8 Cf. The Logic of Scientific Discovery, op. cit. Manuscript submitted 24 June 1981 Final version received 26 January 1982