Walter recently made a remark that was very complimentary to A.J. Ayer, to which I suggested a useful antidote in Popper’s reply to Ayer’s paper for P’s Schilpp volumes. [We might also mention as an antidote Ayer’s misrepresentation of P’s position elsewhere, beginning with Language, Truth and Logic]. Ayer’s paper and P’s reply might be taken as a useful example of the kind of divide that opens up between those who think on ‘justificationist’ lines and those who profess not to. The typical criticisms of P’s theory of knowledge – especially his account of scientific knowledge – are ‘justificationist’; and the typical Popperian rebuttals are non-justificationist’. These typical criticisms involve alleging that either (a) (despite protestations to the contrary) P’s position is only viable given certain ‘justificationist’ assumptions [which P is alleged to smuggle in, for example, in the guise of his account of how one theory might be rationally preferable to another because of its greater verisimilitude or its being better corroborated]; or (b) without any justificationist underpinning, P’s theory of knowledge collapses into an outright scepticism [where, for example, no theory can be preferred to another on rational grounds]. The rebuttals seek to show that P’s position of ‘critical rationalism’ is viable without justificationist assumptions or underpinnings, and so both (a) and (b) are wrong. In evaluating these disputes, it is interesting to consider that there may be a kind of ‘paradigm-shift’ involved in switching from a justificationist to a non-justificationist theory of knowledge:- the hardened ‘justificationist’ cannot accept the viability of a non-justificationist approach and seeks to show that any supposed ‘non-justificationism’ either is a ‘disguised justificationism’ [in which case only, it may be viable] or is not viable [because it lacks the necessary element of justification]. In other words, the ‘justificationist’ cannot accept the idea that all knowledge lacks justification and is merely guesswork and yet some guesses may critically and rationally be preferable to others. The justificationist cannot accept that all knowledge is fallible – i.e. that any ‘knowledge-claim’ may be mistaken – even if this fallibilism does not imply that all knowledge is mistaken. In fact, where knowledge is taken in the typically ‘justificationist’ sense to mean ‘justified true belief’, it would seem that by definition ‘knowledge’ cannot be either mistaken or without justification. Whereas the identification of knowledge with JTB is, for P, simply another subjectivist and justificationist blunder in the western tradition of thinking about knowledge. Popper’s reply, to “Sir Alfred Ayer’s contribution”, extends over pages 1100-1114 and is divided into five sections. The first section, on Verisimilitude, rebuts Ayer’s critical question – “In what way, then, does the concept of verisimilitude afford us a criterion for assessing our progress towards truth?” – by explaining that P never offered this concept “as a ‘criterion’ of anything.” Verisimilitude is not a criterion of progress towards truth for “to say that a theory has a greater verisimilitude than one of its competitors remains essentially a matter of guesswork.” Here Ayer’s treating verisimilitude as “a criterion” of “progress towards truth” is to give it a ‘justificationist’ role: a role so that we would be justified in preferring one theory over another because its greater verisimilitude is guaranteed by some defined measure and its greater verisimilitude thereby guarantees progress towards truth. In P’s philosophy ‘verisimilitude’ is not guaranteed or justified [being “essentially a matter of guesswork”] nor does it guarantee progress towards truth [whether a more ‘truth-like’-seeming theory takes us closer or further away from ‘the truth’ is also a matter of guesswork]. In order to appreciate these respective positions, we might argue out particular cases – to see, for example, whether verisimilitude might plausibly be deployed in a rational and critical way but without its being underpinned by any justificationist guarantees. Some forty years ago, P’s own proposed definition of verisimilitude was proved defective [by Tichy and, independently, Miller], though this did not mean the concept was proven irremediably flawed. Subsequent search for a tenable definition has led to developments that logically refine what is at stake but no entirely satisfactory definition has been found as yet. Nor perhaps is it settled what follows from this or what follows if one is not found (despite the Wittgensteinian author of the Stanford Entry on Popper wanting to suggest the defects of P's definition shipwreck P's theory of knowledge). The search for a satisfactory account of ‘verisimilitude’ has become a highly technical area beyond the reach of most non-specialists. The implications of this ongoing work is also beyond the reach of most. The second section discusses Tarski’s theory of truth. Here P mentions “that Tarski’s conception of a metatheory was for me important for a different reason. “In the discussions following some of my early lectures to the fringe of the Vienna Circle I had been heckled by some Wittgensteinians for speaking of “methodological rules”: they indicated that, according to Wittgenstein, this must be nonsense, since such rules could not be truth functions of “elementary” (or “atomic”) propositions. Obviously they were not Carnapian “syntactic” rules either. I had never accepted these prohibitions, which appeared to me arbitrary and even high-handed. Nevertheless I was not really happy about how to explain what I was doing until I learned from Tarski that we need even in logic a metatheory or metascience not confined to ‘logical syntax’.” (It is the Wittgenstein of the Tractatus and not Investigations that is the source of the heckling.) The third concerns The Verification of Theories. What P says here resists paraphrase so that a considerable part of that section is set out below: “Almost all philosophers since Kant agree with him that there can be no criterion of truth…and with the help of Tarski’s theory of truth it is even possible to prove that, for any but the most trivial languages, there can be no general criterion of truth, even of logical truth, to say nothing of empirical truth. “By a criterion of truth is meant a kind of decision method: a method that leads either generally, or at least in a certain class of cases, through a finite sequence of steps (for example, of tests) to the decision whether or not the statement in question is true. Thus in the absence of a general criterion of truth it may easily happen that we possess true theories, and yet are unable to show, to our satisfaction, that they are true. What can also happen is that we are able to establish some statements as true, by a sort of lucky coincidence rather than an application of a criterion of truth (which may not exist in the case in question). “Thus to assert that we have a general criterion of truth is to assert very much more than to assert that certain statements are true. “Now [to] the two crucial paragraphs of Ayer’s paper, as far as the verification of theories is concerned… “ ‘First, however,’ Ayer writes…, ‘I want to examine his [Popper’s] claim that, at any rate so far as empirical propositions are concerned, there is no general criteria by which we can recognize truth.’ Thus it is my claim which is being examined here, and the term ‘criterion of truth’ is clearly used in the sense explained here – a general criterion by which we can recognize truth. “Yet…Ayer reaches…a conclusion which is the precise opposite of mine: ‘Accordingly, if we can lay down a general criterion for recognizing the truth of basic statements [test statements], there is a sense in which we shall after all have a general criterion for recognizing empirical truth.’ “By ‘empirical truth’ Ayer means, especially, the truth of scientific theories; and the statement just quoted implies that, given an empirical method to decide on the truth or falsity of what I in Logik der Forschung called ‘basic statements’ (and now prefer to call ‘test statements’), we can decide the truth or falsity of scientific theories. “It may be said that Ayer does not and cannot mean this, because…he himself repeats most of the arguments by which I support my thesis of the one-sided refutability of universal theories. However, the last three sentences of the second paragraph are clearly designed to arrive at the conclusion they do arrive at, by an apparently smooth-running but quite invalid argument, leading off with a word – the word ‘Nevertheless’ – that succeeds in invalidating all the preceding admissions. “Ayer’s argument is, in brief: Step (1) Admission: ‘Finding a counterexample proves the statement false, but failing to find one does not prove it true….’ Step (2) ‘Nevertheless…the absence of any counterexample…is….a necessary and sufficient condition of truth….’ [This step is invalid, as I shall show, and it invalidates the argument; but even if we grant it, the argument does not become valid.] Step (3) ‘…the only way in which any empirical statement can meet with a counterexample is by its coming into conflict with a basic statement [test statement]…[; thus] the truth or falsity of any empirical statement…is entirely determined by the truth or falsity of some set of basic statements.’ (Italics mine.) Step (4) consists of the last statement of Ayer’s second paragraph, ending triumphantly, as quoted above, ‘…we shall after all have a general criterion for recognizing empirical truth’. “Is the argument valid, provided we admit step (2), as for the moment I am prepared to do? If it sounds so, it is because Ayer is a trifle indistinct about what set of basic statements is ‘some set of basic statements’. For the set of basic statements actually necessary to ‘determine’ the truth of a theory A would be the infinite set of all test statements which could be relevant to A, reporting on all possible tests undertaken anywhere in the universe, in the past, present, or future. “This is obvious; for by Ayer’s own admission (see step (1) above), “failing to find [a counterexample]…does not prove… [the theory] true’, because there may be unrecorded counterexamples. Thus only if that questionable set of basic statements includes complete reports about all possible counterexamples could the set ‘determine’ the truth of the theory A (always provided we grant step (2), which will be discussed later), while one counterexample, one test statement, could determine the falsity of A. “But if this set of basic statements is infinite – one might even say ‘indefinite’ – it is clear that we would need more than ‘a general criterion for recognizing the truth of basic statements’ in order to obtain a ‘general criterion for recognizing empirical truth’; in just the same way, and for just the same reason, that we need more than a ‘criterion’ for recognizing the whiteness of swans in order to determine whether all swans are white. (This error is perhaps the gravest defect of Ayer’s whole argument.) To ‘recognize’ the truth of all the basic statements belonging to the set is clearly not an empirical process: it would involve a kind of omniscience – an omniscience with respect to basic statements. “Thus Ayer’s argument establishes at best merely the thesis: ‘basic omniscience’ (as I will call it) is involved in any ‘general criterion for recognizing empirical truth’: or empirical omniscience involves basic omniscience. “This thesis sums up, I suggest, all there is in this part of Ayer’s criticism. It hardly needs saying that even if it were validly argued, it would not reveal any weakness in my views. “But even this somewhat unexciting thesis is invalid, owing to the invalidity of Ayer’s step (2). I will discuss this step briefly, although it is somewhat subtle, and although its invalidity need not be established in order to show that Ayer’s conclusion – his step (4) – is mistaken, and that there is no ‘general criterion for recognizing empirical truth’. “Ayer contends that ‘the absence of any counterexample is a necessary and sufficient condition of truth’. This view may be defended for theories like ‘All swans are white’: if there exists (existed, will exist) no counterexample, that is, no non-white swan, then indeed the theory is true. But the view is untenable for all more abstract theories, such as Newton’s. Non-white swans are observable; Newton’s forces varying inversely with the square of the distance are not. (This is why Berkeley said that Newton’s forces were ‘occult’.) The idea that two theories which agree with respect to all testable consequences must be equivalent, is mistaken. Einstein’s special theory of relativity and Lorentz’s interpretation of it are two theories which contradict each other (Lorentz suggested the existence of an inertial system that is absolutely at rest). It does not help here to say that Lorentz’s interpretation contains a metaphysical element that has to be omitted. Einstein’s denial is just as metaphysical, or almost as metaphysical, because nothing observable follows from it. (It is not, in general, possible to split a theory into an empirical and a non-empirical part so that the ‘empirical part’ constitutes a system which can be characterized by a finite number of empirical hypotheses; on the contrary, Craig’s theorem can be used to show that the empirical part of a theoretical system will not in general by finitely axiomatizable. This, obviously, holds even for a comparatively simple theory such as Newton’s theory of gravity.) “Thus there may be two theories which are incompatible, but have identical observable consequences; and one of them may even be empirically better than the other as it may suggest a further generalization (such as General Relativity) which has new and interesting empirical consequences. “But if A and B are incompatible, they cannot both be true, even if there is no counterexample to either of them; and this means that Ayer’s suggestion (2), that the absence of counterexamples is a sufficient condition for the truth of a theory is mistaken.” In other words absence of a counterexample is a necessary but not (necessarily) sufficient condition of the truth of a theory. “So Ayer’s deceptively smooth-running argument….contains at least two steps each of which alone invalidates it.” The position may be somewhat more involved than might be gleaned from P’s account here – which strives to avoid needless complication. For example, ‘All swans are white’ only “may” be a case where the absence of any counterexample is a sufficient condition of truth. Here it may depend on whether the universal generalization [‘UG’] ‘All swans are white’ is taken to express a merely contingently true UG or one that is true because there is a law-like connection between swan-ness and whiteness such that the UG holds as a universal law of nature [‘UL’]. Consider a merely contingently true UG like ‘All dodos die within x years’ where there is no counterexample yet there is no law of nature that would forbid a dodo ever living beyond x years – here the absence of a counterexample would not be a sufficient condition of the truth of the UG as a UL, even if as a merely contingently true UG it would be true by virtue of the contingent fact that there is no counterexample. Where ‘All swans are white’ denotes a UL [and not merely a UG that might be merely contingently true] it denotes there being an abstract or unobservable property of the universe such that it is a structural property of the universe that a non-white swan is an impossible structure (not logically impossible but perhaps physically or chemically or biologically impossible). Taken as a UL, the absence of a counterexample is not a sufficient but only a necessary condition of the truth of ‘All swans are white’. But none of this rescues Ayer’s views from P’s logical demolition. The next section concerns The Verification of Basic Statements. P’s has defended the view that the acceptance of a basic statement is never logically forced on us [as they cannot be verified] and so their acceptance involves a decision, but that decision may be critically arrived at and need not be arbitrary [acceptance may, for example, be due to the basic statement being part of a reproducible and thus testable effect]. Yet for a justificationist this seems unacceptable: if there is no logical justification for accepting a basic statement, by way of empirical verification, then the decision must be arbitrary – and this indeed turns out to be Ayer’s criticism (the same criticism was proclaimed in Stephen Thornton’s Stanford Entry on Popper). For justificationists there is no viable way to critically arrive at a decision to accept certain ‘basic statements’ absent justification of their acceptance by way of their verification. Yet there is scant reason to accept this justificationist stance: and P points to the example of a jury which may arrive at a decision after careful critical examination of all kinds of evidence – their decision can never be derived as a matter of logic from the evidence, yet it may be far from arbitrary given the evidence. P concludes, “And the acceptance or rejection of [basic statements] is a matter for something like the scientific jury – the scientific community (which may or may not come to an agreement).” There is a final section on Subjective Experience and Linguistic Formulation, where P addresses “What is the fundamental difference between Ayer and myself?”; but we may be spared this story. (For now.) Donal