Bartley on the Limits of Criticism

Section XVII to the end.

XVII. The Check of Logic. Lejewski

By far the most controversial critical check is the fourth: logic.

Important in any discussion of the dispensability, revisability, criticisability of logic are the views of W.V. Quine, as expressed in his treatment of the traditional distinction between analytic and synthetic. According to Quine, when a body of belief is brought to the test of criticism, any part of it may be revised. No segment of it, such as the set of analytically true statements, including logic, is so insulated that we could say in advance that "the mistake could not be here."

Quine's position is influential and has important antecedents. Quine's predecessor as Edgar Pierce Professor at Harvard, C.I. Lewis, for instance, wrote as foliows:73

"The difference between . . . the decision of relativity versus absolute space and time, and those more permanent attitudes such as are vested in the laws of logic ,.. is only a difference of degree ... Conceptions, such as those of logic, which are least likely to be affected by the opening of new ranges of experience, represent the most stable of our categories; but, none of them is beyond the possibility of alteration."

Such a conclusion - that everything, including logic, is open to revision - may appear to be identical to pancritical rationalism. Hence I want to indicate how importantly my views actually differ from those of Quine.

There are of course a number of senses in which logic not only is open to revision but has indeed been revised. To take an example: the traditional Aristotelian logic of categorical propositions has been largely abandoned because it is too clumsy and restricted to enable one to formulate in it and defend the validity of many rules of inference which are valid in ordinary discourse, not to mention the inferences of physics and mathematics. In addition, various artificialities are introduced into logical systems to avoid the logical paradoxes of Grelling, Richard, Russell, and others."74

No doubt our logical theories may be similarly repaired and revised far more than anyone supposes at present, and it is impossible to foresee when such repairs will be necessary. Nonetheless, however much alternative systems of logical rules of inference may differ among themselves, they have an important feature in common: whenever these rules are observed and, starting from true premisses, the argument is conducted in accordance with them, we arrive at a true conclusion. The question arises whether logic can be revised in the sense of denying that true premisses always need lead, in any valid inference, to true conclusions.

Certain of Quine's comments suggest that he does regard such radical revision as in principle possible. In "Two Dogmas of Empiricism," Quine mentioned only the possibility of revising the law of excluded middle. But in Word and Object he explicitly adds the law of contradiction.75

Quine seems to envision matters something like this. Let the circle below represent the "man-made fabric" of knowledge of which he writes:

[Editor's note - at this point Bartley inserts a graphic which can be described as follows: imagine a circle, and the bottom of the circle a triangle is pushing its way into the circle and is marked A. From the letter A an arrow is drawn that points towards the center of the circle where the following words are printed: logic, definitions scientific laws, empirical observations. ]

At "A," imagine what Quine calls a "recalcitrant experience," something that does not fit into the system and forces one to make some revision within it. To bring the system into equilibrium, that reconciling arrow may light anywhere: on a scientific theory, on a definition - or on logic.

But what about that arrow?

I contend that the workings of the arrow are outside of, and are presupposed by, Quine's "argument situation." The idea of revising in the light of tests presupposes the notion of deducibility, i.e., the idea of the retransmission of falsity from conclusions to premises and, ipso facto, of the transmission of truth from premises to conclusion. Deducibility appears to presuppose a minimal logic, in the sense of "logic" that refers to the general theory of derivation, demonstration, and refutation.

If so, the idea that a system might abandon even this minimal logic must be mistaken: so much logic would be needed to argue and learn - to bring the rest of our theories into closer correspondence with reality. Hence not all of logic can be part of the totality that is under test. There is an absolute difference in principle between, what is under revision and what is presupposed by the revision or argument situation. Contrary then to what Lewis says, this marks a difference in kind, not only in degree; contrary to what both Lewis and Quine maintain, there is a difference in kind between a fundamental alteration in logic and the replacement of other views such as (to take Quine's examples) Ptolemy's with Kepler's, or Newton's with Einstein's.

But how much of logic is presupposed by deduciblity?

In 1959 when I first thought this problem through, I supposed the notion of deducibility to be far richer in this connection than I do now. I was influenced, at that time by G. Gentzen, and also, especially, by the series of articles on logic which Popper published in 1947 and 1948, and which is no longer as well known as it deserves to be.76

Popper's work on logic was in part inspired by Tarski's "On the Concept of Logical Consequence" (1936), where Tarski showed that the concept of logical consequence could be readily elucidated once one had a list of the formative signs.77 Popper turned this around. He began not with the formative signs but with the concept of logical consequence. Taking that - understood as the transitive and reflexive relation of deducibility as primitive (together with some general primitive rules of inference not referring to formative signs), Popper tried to show that all the formative signs could be derived in terms of it. He wished to show this even for quantification and identity, neither of which had been definable by the truth-table method. From these "inferential definitions," and with "trivial inferences," the whole of mathematical logic would be derived (i.e., classical Aristotelian logic, Boolean algebra, propositional logic, the lower functional logic).

By this means Popper hoped to do something Tarski had doubted was possible: to provide a firm demarcation between formative and descriptive signs. More important, he would thereby attain the old goal of "trivializing" mathematical logic - in the same sense in which it had earlier been hoped to trivialize logic through the truth-table method: logic would be obtained entirely from definitions of formative signs - without employing axioms or primitive rules of inference for the formative signs.

Popper's work in logic was indifferently received, partly due to several blunders that he made; yet even one of the most unfavourable reviewers remarked that "Presumably [Popper's] ideas can be carried through, at least in principle."78 Certainly a number of logicians were much influenced by Popper's work - notably W.C. Kneale and Bruce Brooke-Wavell. 79 Evert Beth said that Popper's work had partly inspired him to build his own well-known method of "semantic tableaux." 80

Popper's work obviously relates importantly to my disagreement with Quine. If the situation Quine describes does not presuppose the idea of deducability (as I maintain), and if, from the notion of deducibility, a significant part of mathematical logic can be obtained, then much more than a minimal logic is isolated from revision in Quine's argument situation.

Hence I read Czeslaw Lejewski's brilliant and important contribution to the Shilpp volume - "Popper's Theory of Formal or Deductive Inference" - with a feeling of regret. Lejewski reconstructs Popper's theory of deductive inference and Tarski's theory of logical consequence in the language of S. Lesniewski, showing that Popper's and Tarski's theories are inferentially equivalent. In the course of doing this, he elicits and makes explicit the unstated presuppositions of Popper's theory. These turn out to amount to an axiom system; and, as Lejewski shows, cannot be reduced to a set of inferential definitions. Popper's system assumes certain postulates which are independent and prior to definitions. Thus Popper has failed to construct a logic without assumptions or axioms, and cannot by his method obtain logic through inferential definitions alone.

In a most impressive abandoning of a position, Popper accepts Lejewski's criticism, acknowledging that his papers were, as an attempt to build up a simple system of natural deduction, "just a failure". Popper also indicates that he has come to accept Tarski's scepticism about the possibility of clearly demarcating formative and descriptive signs.

Assuming that Lejewski's work stands, our insight about the presuppositions of the deducibility or argument situations may still be true, but it has much less support. We can no longer appeal to Popper's New Foundations to specify what part of logic is involved in the argument situation. But despite his disappointment, the basic insight remains correct: there is still a minimal logic presupposed in that situation. So we need to return to the question asked above: what is this minimal logic? How much logic does deducibility involve? What part of logic, if any, is unrevisable (within the argument situation) under any circumstances whatever?

XVIII The Presuppositions of Argument

I do not know what the full answer to this question may be but I think it to be rather different from what one might expect. 81

The key to the answer relates to the whole ecological purpose of arguing and using logic, as conceived here: to heighten criticism, to expose views to maximum review, to "make things difficult" for our positions.

Some components of ordinary or "standard" logic always contribute to this goal: and others do not. Ordinary logic includes both the law of noncontradiction and the law of excluded middle. Of the two, our minimal logic would have to retain the law of noncontradiction as a metalinguistic rule governing the argument situation: if contradictions were permitted, falsity could not be retransmitted and criticism in the sense intended would be impossible. The law of noncontradiction always works to heighten criticism. Any abandonment of it leads not simply to a weakening but to a minimization of criticism. 82

The situation with regard to the law of excluded middle is, however, interestingly different. For there are circumstances in critical argument where the law of excluded middle may be dispensed with precisely for the purpose of heightening criticism. That is, there are circumstances wherein the law of excluded middle can weaken criticism. An example is intuitionist logic, as developed and advocated by L.E.J. Brouwer, one well-known feature of which is to abandon the law of excluded middle (which, although a well-formed formula of the system, is not demonstrable within it). Thus the theorems of intuitionist logic must be proved by weaker means than are available in classical logic. In this case, the weaker logic serves to set up restrictions on the admissibility of certain kinds of proof. Indirect proof is ruled out. Not statement may be admitted unless it can be directly proved by producing an instance in a finite determinate number of steps. In introducing such requirements, and ruling out indirect proof, proof is made more difficult, and criticism is, in this context, for this limited purpose, consequently strengthened. 83

This does not mean that the law of excluded middle may always or even often be abandoned to the betterment of criticism. Quite the contrary. One must distinguish between the different critical aims of constructing or demonstrating or proving, on the one hand, and of nondemonstrative testing, on the other.84 Intuitionist logic is constructive, whereas in most situations in science we are faced not with a construction task, but with one of testing, in which not all legitimate statements are decidable.

This may be seen by recalling the controversy over indeterminacy in quantum physics. A number of philosophers, including Quine, Werner Heisenberg, Friedrich Waismann, and others, have suggested that indeterminacy may force abandonment of the law of the excluded middle. The argument can be summed up thus: it is impossible to determine both the position and the momentum of a particle at a particular time. Hence some statements about particles seem to be neither true nor false but undecidable. Only the law of excluded middle forces us to say that these must be true or false. Therefore abandon the law of excluded middle.

The argument is incorrect. Not only the law of excluded middle provokes the trouble here. Another assumption, unmentioned in the argument just sketched, is the old logical-positivist principle of verification: that there is something wrong with a statement in science that cannot be decided or verified -- that it is meaningless. Insisting upon this untenable principle leads to the rejection of the law of excluded middle, and to a consequent reduction of testability within quantum mechanics. By retaining the principle of verification, one distracts attention from the possibility of solving this problem by distinguishing ontological and epistemological levels, and noting that from the fact that a certain thing cannot be determined it does not follow that that thing does not exist or that anything said about it is meaningless. From the fact, given a particle's momentum, we cannot determine its position, it does not follow that a statement about its real position is not true, in fact, let alone that such a statement is meaningless -- unless one is assuming something like the verification principle. The verification principle, rather than the law of excluded middle, may be dropped here.

Thus Brouwerian "positivism" in constructive mathematics increase criticism, while logical positivism in quantum physics decreases criticism. In the latter, a stronger logic, retaining the law of excluded middle, serves to increase testability. For purposes of criticism, as opposed to proof, one will ordinarily wish a stronger logic.

Something can be done with these results.

First, we can put forward informally a sort of "revisability criterion."85 A statement would be unrevisable (in argument) if and only if there are no circumstances under which it can be abandoned without weakening the exposure of the system as a whole to criticism. Such unrevisable statements, if any, would mark the absolute presuppositions of critical argument.

This gives a fairly clear line which, although not corresponding to the traditional "analytic-synthetic" dichotomy, does mark off one portion of the class of truths traditionally known as analytic. This in turn rebuts Quine's claim that no sharp boundary line between analytic and synthetic truths may be drawn.

If we take Popper's falsifiability criterion (a statement is scientific if and only if it is incompatible with certain statements reporting possible results of observation) as a way of marking off scientific from nonscientific statements, and our revisability criterion as a way to demarcate what is revisable within the situation of critical argument from what is not, the spectrum of types of statement falls into at least three relatively clear categories. Various other categories are not so clearly demarcated.

What I have in mind is suggested by the following chart:

[Editor: The chart has been omitted, but a dedicated reader should be able to follow Bartley's arguments without it. ]

The chart is intended lightly, and should be discarded once it has served its purpose. Some of its details, which suggest some rezoning of the traditional philosophical map, I have discussed at greater length elsewhere. 86

In a small area on the left is located our minimal sense of logic: this is referred to as "Logic I ." In an area on the right are the theories of empirical science. Between, in a larger section, are some statements that have traditionally been called analytic and others that have traditionally been called synthetic, but which are neither empirically refutable (and hence scientific in Popper's sense) nor presupposed in all critical argument (as in Logic I). This middle area would presumably contain all of metaphysics, most of mathematics, and most of that class of statements like "All brothers are male siblings" around which much of the analytic-synthetic controversy has revolved. Accordingly, a number of classifications appear on the chart:

analytic/synthetic

nonscientific/scientific

nonexplanatory/explanatory

necessary/contingent

a priori/a posteriori.

These various categories have uses, despite their often vague and almost always overlapping character, most having been proposed to deal with some problem that none of the others quite fitted. Since the original problems overlap, the categories and distinctions do too, and remain a source of some confusion. If at least two ends of the spectrum are pinned down, it may be easier to salvage what is useful in the remaining categories, and also to see how varied is the spectrum that they cover.

The chart may help suggest, for instance, why some philosophers prefer to identify metaphysics and definitions, or lump them together vaguely even when they do not identify them. For these fit none of the restrictive logical or else empirical dichotomies that are so common. They have, however, under the influence of Aristotelianism, tended to be definitionalized rather than empiricized For Aristotle, the ultimate presuppositions of argument are definitions of essences, presupposed necessarily in order to terminate infinite regress.87 On such an account, metaphysics, definitions, and necessary propositions tend to become identified. Moreover, philosophies that stress verifiability rather than falsifiability as a criterion for science inevitably encourage attempts to reduce those large parts of science which cannot be verified to definitional or rule status, and thus tend to become essentialist or instrumentalist.

Only two reasonably definite criteria are then indicated on the chart. I can state no such criteria for demarcating the traditional analytic-synthetic divisions. Had Quine said only that there is a good bit of shifting back and forth, and that the classification of some statements will be uncertain, there would be no disagreeing with him. 88

I have made a separate space, Logic II, in the middle area, to contain elements of logic, such as the law of excluded middle, which are not presupposed in all critical argument. Another "catch-all" category, Logic Ill, is indicated, which will vary with the logical and descriptive terms of various languages. Distinguishing these three different levels of logic (in a way that is hardly intended to be exact or exhaustive) may help to suggest why logic has at some time seemed entirely a matter of convention, and at other times, to the same or different thinkers, seemed necessary.

Logic in the first sense, Logic I, is required in order to have a critical argument at all. Whatever "necessity" logic in these further senses may have arises only within a certain domain and for particular purposes -- not for the sake of critical argument itself. Thus in Brouwer's ideal world of mathematics, one in which all legitimate statements are decidable at any time, one drops the law of excluded middle in order to maximize criticism, in order to make as difficult as possible the discovery of the population of that world, and to insure as much as possible against error in specifying that population. The situation differs in the world of experience, where all legitimate statements are not in fact so decidable. Thus, the combination of a particular task dependency and a particular domain dependency produced the need in intuitionism to proscribe the law of excluded middle. Yet even here there is no relativization of logic in respect to the fundamental aim of maximizing criticism, of stacking the cards, as it were, against our conjecture.

In insisting that a minimal sense of logic is presupposed by the critical argument situation itself, I have thus identified an absolute presupposition of argument. Philosophers have long sought to specify what was absolutely presupposed in intellectual activity and what not. This particular answer is, of course, not the one anticipated, since it is neither a principle from which all else can be derived nor a category to which all else need conform. Most categories and principles, along with much else traditionally regarded as a priori, necessary, analytic, unrevisable, indispensable, I see as revisable and dispensable within the argument situation.

XIX. The Despair of Reason

Despite what has been said here about how some minimal logic is presupposed by the argument situation, a reader who noted the apparent similarity, alluded to above, between Quine's view that everything (including logic) is open to revision and my account of pancritical rationalism in which everything (including the practice of arguing and revising and using logic) is open, to criticism -- may wonder whether I am not backing out of pancritical rationalism -- or simply contradicting myself.89 It might seem as if I were now insisting that we are committed to logic.

Some support for such a response to my argument is to be found in some work by Jonathan Bennett concerning the analytic-synthetic controversy.90 A brief look at his views serves to bring out the relationship and the difference between the problem of the limits of rationality and the problem of the analytic and synthetic.

For Bennett, to call a sentence "analytic" is to register the conviction that it is highly indispensable for establishing or justifying, in providing answers to "how do you know?" questions. Bennett attempts to refute Quine's views about the possibility of revising anything in argument by citing the argument about the limits of rationality; by arguing that, in any argument, the infinite regress of justification has to be stopped by referring to something that is not open to revision; that in every argument situation at least one sentence is not up for revision, on which the rest of the argument depends; and that such is necessary if one is to be able to construct an argument at all. For such terminal sentences, Bennett writes, "the question of their possible falsification" is ruled out: such "terminal sentences are put off time cards for possible revision"; once one works back to them, "there is nothing more to say."

Claiming that such an unrevisable statement would be "traditionally analytic" (in the sense Quine was attacking), in that the difference between such a statement and other statements was a difference not in degree but in kind, Bennett concluded that Quine's Views should be rejected.91

Bennett's approach is justificationist, and makes use of the argument concerning the limits of rationality which was refuted above. Outside a justificationist context his argument has no force. Apart from this, he fails to demarcate analytic statements in one important traditional sense. One steady element in the changing philosophical conceptions of analyticity has been the hope thereby to specify something that is common to, presupposed in, all argument whatever - such as Logic I, as explained above. Bennett argues only that something is presupposed, regarded as unrevisable, as ultimate, in each argument. He does not claim that this something that is presupposed in each argument must be presupposed in all arguments: different arguments can have different unrevisable principles. With the traditional notion of analytic-as-unrevisable, it was hoped that disputes could be limited -- by showing that something necessarily had to be accepted in common as unrevisable by experience. Yet Bennett appeals, as a defence of the "traditional" notion, to the "fact" that the dilemma of ultimate commitment -- the main source of ultimate relativism -- cannot be avoided. His argument, therefore, even if it were correct, hardly speaks to that aspect of the problem.93

My claim that all critical argument depends on Logic I (and that Logic I is hence unrevisable in argument) -- in the sense that such logic defines the space in which argument takes place -- differs then from Bennett's contention that every argument depends on some particular unrevisable ultimate premise.

Even after such clarifications and distinctions are made, an explanation may still be demanded of how I could argue myself out of the practice of argument and using logic while at the same time, necessarily, in that argument, presupposing logic. I cannot mean simply that I could be forced (by brainwashing, nervous collapse, etc.) to abandon the practice of argument and using logic.

I see no difficulty in satisfying this demand. A large part of the philosophical tradition testifies to the possibility of being argued logically out of the practice of argument and using logic. "It may seem a very extravagant attempt of the sceptics to destroy reason by argument and ratiocination," as Hume puts it in An Enquiry Concerning Human Understanding, but this is, as he adds, "the grand scope of all their enquiries and disputes."93 What I have in mind is indeed the despair of reason and truth to which Nietzsche referred when he wrote of "gnawing and crumbling skepticism and relativism."94

Consider several examples.

First, there are the logical paradoxes (the liar, Richard, Grelling, etc.). The paradoxes are reached in the course of rigorously logical argument. Therein lies their telling power: using logic and presupposing logic, one reaches illogic. If the paradoxes could not be avoided, then one would have grounds for deeply distrusting logic and rational argumentation. Of course one might say that in rejecting logic because it led to illogic, one is presupposing logic and is hence inconsistent. But the irrationalist makes no claim to be consistent. And would it be more consistent to accept logic that led to illogic?

If the example of the paradoxes seems unconvincing -- perhaps from the perspective that ways have now been found to avoid them -- take as another example, or set of examples, a line of argumentation which is particularly pertinent here, since we are concerned, after all, to build an argument against relativism, fideism, and scepticism.

1) Relativism, fideism, and scepticism contend, on the basis of arguments about the limits of rationality widely thought to be rational, logical, and indestructible, that serious argumentation is futile in the sense that, from a rational or logical viewpoint, one position is as good as another. Their basic contention is that there is a rational excuse for irrationalism.

2) The upshot of determinism -- another position reached by many on the basis of reason and logic -- is that all argument is illusory. If the doctrine of determinism, often championed in the name of reason and science, is true, then reason and science are illusory. It would not just be some things that are fixed from all eternity: everything is so fixed, including the results of any particular scientific argument. Even the result of any argument about determinism is determined: that one person shall become a determinist is fixed -- without any regard to the weight of any argument; and that another shall become an indeterminist is similarly fixed. If one's views are fully determined by natural laws and boundary conditions, then they do not depend on the force of argument or on the weighing of evidence.95 If scientific determinism is true, we may believe or reject it; but we do so not because we judge the argument in its favour to be sound or unsound, but because facts and laws determine that we shall so believe or fail to believe. If determinism is true, then the distinction between being forced to reject logic and being argued out of logic loses its meaning.

If one were, by argument and logic, to reason oneself into the relativist, fideist, and sceptical positions -- or into the determinist position -- and at the same time were to notice these implications -- one might well conclude, through apparent reason and logic, that reason and logic are to be rejected as guides to life. Using the argument, and presupposing argumentation, we destroy argumentation.

I am not a determinist, and I believe Popper has refuted determinism (See Part II). Moreover, I believe that the argument of this study defeats scepticism, fideism, and relativism. Yet I want to hold myself open to the despair of reason, in case the argument should lead somewhere different tomorrow. Thus the pancritical rationalist may hold his practice of reasoning and obeying logic -- just like everything else -- open to comprehensive criticism and rejection.

The fact that argument presupposes a minimal logic as unrevisable in no way identifies commitment on the part of a pancritical rationalist. To be sure, the practice of critical argument and logic are bound up together. One can no more argue without a minimal logic than one can live without breathing or speak without language. None of these three -- living, speaking, breathing -- require irrational commitment to a dogma. Nor for that matter are logic and arguing even peculiar to or identificatory of a pancritical rationalist-- any more than breathing or speaking is. To the extent to which he wishes to employ arguments, any irrationalist must use a minimal logic too. The strength of the irrationalist's tu quoque was indeed that it defeated the rationalist logically and rationally, on the rationalist's own terms: it defeated the rationalist with an internal criticism. If one wanted to argue that any one who uses logic is committed to logic, then one would have to claim that the irrationalist too is committed to logic! In which case the tu quoque would, once again, vanish.

XX.

The Ecology of Rationality

The problem of the limits of rationality provides a powerful argument on behalf of attachment and commitment and, as long as it is unsolved, serves the interests of those who identify with, cling on to and defend, their positions and contexts, their "belief systems."

Assuming that this problem has now been solved, what is the next step? If criticism is, logically, unlimited, how is a more radical criticism to be instituted? What are the cultural ramifications of a change of metacontext? What would happen -- intellectually, psychologically, socially, politically? What would a culture lethal to positionality and attachment really be like? And would that really be desirable?

Although such questions are obviously both highly speculative and very difficult, Popper has, in his political philosophy, in his charting of the "open society," been attempting a preliminary answer to them. When his political philosophy is considered in this light, it takes on a different complexion. Although he has often been described as a conservative political thinker, his programme has its goal the personal and institutional implementation of a transformed metacontext -- one that in this case involves the transformation of western man away from the positionality and attachment that have marked his career. To this aspect of Popper's philosophy -- to his social and political thought, and his philosophy of history -- I shall turn in my next instalment, Part IV of this series.

There are, however, some related ecological issues that should be mentioned in this instalment. The first has to do with those philosophical presuppositions that work against the creation of such a metacontext. Earlier (section VIII) I spoke of these as "pollutants," and identified three which blocked a solution to the problem of rationality.

But there are many others. One contribution of Popper's work, taken as a whole, is to identify and combat a nest of epistemological presuppositions that work against criticism and keep individuals imprisoned within the justificationist metacontext. Since Popper has not been concerned to treat these in a systematic way, but has examined them piecemeal in scattered parts of his work, some readers may miss their total impact. So let me simply list, with little explanation, some of the more important of these. Taken together, they amount to a kind of moral code for intellectual endeavour -- a moral code to which Popper is vehemently opposed. A knowledge of these, it seems to me, is invaluable. To heighten criticism, one must understand thoroughly how criticism may be reduced, evaded, avoided. Until one has brought to light and exposed to criticism various hidden, often unconscious, strategies that work against criticism, one can hardly expect to be in a position to recommend effective ways of eliminating or circumventing their effects.

First, there is a series of prohibitions, such as "Do not speculate," and "Do not err,"96 and injunctions such as "Define all your terms and prove all your propositions."97

Then there are numerous justificationist dogmas, such as that which maintains the value of precision for its own sake, and which Popper combats with a call for appropriate precision and with a startling warning about the limits of precision: that we never do nor ever can know what we are talking about.98 There is also the doctrine that truth is manifest to the pure in heart and mind, and that it will, as Milton promised, prevail in any free encounter. Closely related to this there are the dogmas of the veracitas dei (as in Descartes) and of the veracitas naturae (as in Bacon), as well as the conspiracy theory of error. These Popper combats with his doctrines that truth is hidden, not manifest, and that errors -- such as the conspiracy theory of error -- persist naturally without any need for conspiracy.100 There is also the doctrine that the unknown must be explained by the known, the less familiar by the more familiar, which Popper combats in his account of explanation.101 Then there is the doctrine of the autonomy of disciplines (which erects a protective shield around each discipline). Very important, there is the dogma that "like causes like" (discussed by Watkins in Schilpp, p. 394); there is also the fundamental doctrine (discussed in Part II, this series) that "nothing comes from nothing."

Popper also uncovers many assumptions, such as the fusion in logical theory of derivation and demonstration; the transmissibility assumption; the authoritarian presumption of epistemologies wherein even the sceptic assumes that if there were any knowledge it would have to come from an authority. There is also the inquisitorial (as contrasted with the accusatorial) conception of scientific inquiry; and the assumption that without certainty learning is impossible.102 And there is "theoretical monism," the assumption that a plurality of theories on a given problem is undesirable. There is also the assumption that knowledge is a specially secure kind of belief, whereas, for Popper, knowledge has nothing to do with belief.

Among other pollutants identified by Popper are protective or evasive strategems, which have the effect of deflecting criticism and of making whatever they are turned on into a reinforced dogmatism. These include ad hoc auxiliary hypotheses, and ad hoc redefinitions.103

Taken together, these work to reinforce positionality and to shore up the justificationist metacontext, and to work against an ecology of rationality. An important step in combating the justificationist metacontext is to bring to light this nest of assumptions which undergirds it.

A second ecological issue that I would like to raise here pursues a different line of thought.

I referred in Part II (pp. 702-6) to Popper's comparative neglect of what he calls World 2, subjective consciousness. This neglect, however benign, and however appropriate in combating the subjective orientation of the justificationist metacontext, becomes regrettable here. For it appears that human beings have a natural psychological tendency to identify with positions and contexts, and thereby become positional. To the existence of such a tendency, both western and oriental psychologies appear to agree. For a fallibilist metacontext to succeed, it would be necessary to combat this tendency. Such a tendency, and the behaviour that follows from it, work against the implementation of a fallibilist metacontext and against a favourable econiche for World 3 -- against those conditions that facilitate change of position, innovation, intellectual "sporting." By the same token, the critical examination of positions and innovation are threatening to this natural tendency. Here, there appears to be a conflict not simply between fallibilism and various dogmas, assumptions, practices and institutions. There is also a conflict between fallibilism and deep-seated psychological mechanisms.

So much is recognized in the orient by yoga and other practices for cultivating non-attachment through psychological and spiritual discipline. The radical cultivation of World 3, working for the growth of the products of rationality, calls for an equally radical psychological (World 2) transformation: a growth in subjective rationality. Thus the cultivation of World 3 might itself work as a kind of yoga. A healthy growing World 3 works for detachment and against the ordinary tendency towards positionality: although the objective development of knowledge, in injecting new forms into the world, contributes raw materials for positionality, it also, in providing contrasts, alternatives, and culture clash, works against positionality on the World 2 level.

Popper's approach is incomplete so long as greater attention and responsibility is not given to World 2.

XXI.

Theories of Demarcation Within the Justificationist Metacontext

We now reach the problem of demarcation. Although, like almost any attempt at classification, Popper's theory of demarcation between science and nonscience has disadvantages as well as advantages, it has an illuminating effect. If contains the best characterization of science that I know, and also lends itself to insights concerning the history of philosophy.

This is so if one thinks of the theory "in the large," bearing in mind Popper's most general intentions. In detail, his theory of demarcation is confusing: in his development, statement, and application of it, from 1932 to the present, Popper has been inconsistent.

In the following, I try to indicate some of the necessary context, so that the spirit of Popper's theory can be appreciated as well as some of the difficulties. We may begin with a few preliminary remarks.

1) Theories of demarcation are a longstanding feature of the justificationist metacontext, and play an important role in justificational criticism (see Section IX). They arise out of the basic problem of demarcating good ideas (fantasies, myths, scientific theories, whatever) from bad ideas, and thus are an important part of traditional approaches to rationality. Consider this check list:

good traits

bad traits

true

false

probable

improbable

clear and distinct

unclear and indistinct

demonstrable by reason

nondemonstrable by reason

empirical

nonempirical

verifiable

unverifiable

meaningful

meaningless

scientific

unscientific

Which indicators of goodness and badness are taken most seriously depends on in which part, and in which period, of the justificationist metacontext one finds oneself. Thus, for Descartes, good ideas are demarcated from bad ones by finding which can be reduced to clear and distinct ideas; for Hume, good ideas are demarcated from bad ideas by finding which are, empirical, i.e., which can be reduced to sense observations. And so on. As to bad ideas, on some demarcations they are simply undesirable in some respect: being confused, unclear, or poorly related to evidence, and so on. On other demarcations, they are much worse: e.g., straying beyond the bounds of human understanding or of human language.

The items on the check list have a staying power. Thus, even though clarity and distinctness are now commonly regarded as insufficient, they are, in and of themselves, still prized. As to truth, although no modern philosophy claims a criterion of truth, all agree that truth is a hallmark of goodness. The focus of attention in modern and contemporary philosophies has, however, been on probability and on the last four items on the list. Most forms of positivism and empiricism, for instance, agree that good theories will be of high probability, and will also be empirical, verifiable, meaningful, and scientific. Demarcations focusing on science have been of prime importance since Kant.

For most justificationist philosophies, such demarcational criteria serve to indicate acceptability. That is, if a particular theory satisfies the appropriate criteria, then it is acceptable, is acceptable as a belief, or is worthy of belief, deserves to be believed: belief in it is justified.

Falling within this general approach to demarcation, there is a subordinate problem which has been of particular interest during the past century, that of purging or cleansing science itself of occult or metaphysical elements, which is of course associated in recent times with positivism.

2) Another traditional demarcational problem has been directed at a very, different way not as a matter of demarcating, say, good scientific ideas from bad nonscientific ideas, but rather of circumscribing the aims of science so as to prevent them from conflicting critically with some other area of human thought. Religious writers often take such an approach. One may find in their writings demarcational criteria intended first to determine (a) what is endorsed by religious authority; (b) what is prohibited by religious authority; (c) issues on which religious authority is neutral. After having demarcated the scope of religion, they will proceed to demarcate science subordinately to the demarcation of the scope of religion, so as to insure that science -- so understood or demarcated or defined is able to penetrate only those areas wherein religion is neutral, or is otherwise of such a character as to prevent conflict between the two. Sometimes such writers will create their own conceptions of science; more often they will take over and exploit conceptions of science -- such as conventionalism or instrumentalism that are readily adaptable to their purposes. For even where religions have not claimed evalutional omnicompetence (and they have), they have wanted protection from the possibility of scientific challenge in those areas where they do claim competence. One beautiful example of the importance of demarcational criteria battles between scientific and religious world views may be found in the controversy between Abel Rey and Pierre Duhem over the latter's instrumentalist conception of scientific theories, which led to Duhem's essay, "Physics of a Believer."104 Duhem regarded metaphysics as explanatory and legitimate, but within the province of religious authority. Whereas physical theory was for him not explanatory, but classificatory, made no truth claims, and thus could not conflict with metaphysics or with religion. Another rather more complicated example is to be found in Robert Cardinal Bellarmino's attitude to the theories of Copernicus and Galileo, stated in his famous letter to Father Antonio Foscarini of 12 April 1615.105

Both these traditional demarcational efforts are important. The whole history of philosophy (and of its interrelationship and interaction with science and religion) takes on a clearer, look when focused, as Popper wishes it to be, in terms of demarcational problems. When one brings such demarcational efforts together with our ecological goal of heightening criticism amid the growth of knowledge, it also becomes apparent that both these demarcational efforts have often served to hinder rather than to heighten criticism. Justificationist and positivist attempts to use empirical reducibility as a sufficient strategy of criticism have, for example, as Popper puts it: "far from defeating the supposed enemy metaphysics, in effect presented the enemy with the keys of the beleaguered city."106 Such criteria of evaluation excluded intuitively acceptable views, such as Newton's and Einstein's theories, and included intuitively unacceptable views. And instrumentalist or conventionalist conceptions of science may use their power to circumscribe the domain of science precisely in order to circumscribe criticism as well: scientific instrumentalism allows theological and other doctrines to avoid any clash with scientific theories.

XXII.

A Popperian Shift in the Problem of Demarcation

Popper's theory of demarcation between science and nonscience does not, despite his own description of it, really fit into the traditional aims of theories of demarcation.

Most traditional theories were, as we have just seen, chiefly concerned with demarcating good ideas from bad ideas. Whereas Popper's theory of demarcation between science and nonscience does not directly serve this aim -- at least not consistently.

It would of course be possible to create a check list for Popper on the general lines of that presented above. One might try the following:

good traits

bad traits

falsifiable

unfalsifiable

high content

low content

low probability

high probability

scientific

nonscientific

The list is not inaccurate: Popper does prize falsifiability, high content, low probability, amid scientific status in a theory. But to juxtapose the two lists is to suggest a continuity that is not there. For Popper is dealing with different problems, and his criteria are put forward not in a justificationist metacontext but in a fallibilist one. Popper's criteria are, for example, not criteria of justifiability, acceptability, or worthiness of belief. A falsififable theory of high content, low probability, and scientific status remains for him forever unjustified. And, very important, the bad traits in the right-hand column are not really bad for him -- at least not in the same way. For his methodology allows for, even esteems, some metaphysical theories which are unscientific and unfalsifiable, and of relatively low content. He emphatically denies that science is exhaustive of legitimate claims. Nor is he concerned with problems of the limits of human knowledge or human language, to show how some particular attempts to chart such limits fail.

Understandably, a complete and satisfactory fallibilist metacontext did not spring forth full grown, but has had to emerge through the encompassing fog of the justificationist metacontext. In his earlier work, Popper does not appear to have been aware that he had made a metacontextual shift; and this may explain some anomalies in his writing. Thus his own account differs somewhat from the one I have just given. He himself presents his falsifiability criterion more as if a solution to a continuing historic problem which he shares, most immediately, with the positivists. Writing of the problem of demarcation, he titled his first long book -- devoted to the twin problems of induction and demarcation -- Die beiden Grundprobleme der Erkenntnistheorie. He calls the problem of demarcating science from nonscience the central problem of the theory of knowledge; he states that he has tried to demonstrate that "the problems of both the classical and the modern theory of knowledge (from Hume via Kant to Russell and Whitehead) can be traced back to the problem of demarcation, that is, to the problem of finding the criterion of the empirical character of science."107 He describes the solution to this problem as the "key to most of the fundamental problems of the philosophy of science."108

I see matters differently. The key traditional problem which both Popper and the positivists were at work upon was the problem of charting accurately the critical relationship between theoretical statements and observation statements. This was the problem of induction, but within many justificationist approaches, and particularly within positivism, it coincided with the problem of demarcating between good and bad ideas. In Popper's nonjustificationist approach these problems no longer coincide, but he produces another demarcation the demarcation between science and nonscience which does coincide with the problem of charting accurately the critical relationship between theoretical and observational statements . He then frequently treats this second and distinct demarcation as if it were responding to the same problems as traditional theories of demarcation. Herein lies the mistake.

Among the anomalies produced by this mistake is a certain ambivalence about the whole issue of acceptability. In one place, in reflecting on his work on demarcation, Popper states categorically that, in demarcating science from nonscience, he was not interested in the question of the acceptability of theories or in "the human problem of reasonable belief." "My problem," he writes, "was different. I wished to distinguish between science and pseudoscience.'109 Elsewhere, however, the issue of acceptability creeps back in. He writes of "degree of corroboration" as "degree of acceptability,"110 and as "degree of rationality of our belief that a certain theory has achieved (a certain degree of) truthlikeness"; and he declares his interest, in "the problem of the reasons for accepting or rejecting scientific theories."111

Thus I suggest, and will argue in the following:

1) that there is a radical discontinuity between the traditional problem(s) of demarcation and the problem(s) of demarcation dealt with by Popper, due largely to the shift in metacontext;

that Popper, particularly in his early work, fails to take this discontinuity into account, and is not fully aware of the metacontextual shift that his work constitutes, or of all the implications of that. Thus his solution to his problems of demarcation is frequently, and wrongly, presented by him and others, as if a solution to the traditional demarcational problems.

XXIII Two Problems of Demarcation

Contained within Popper's theory of demarcation are two separate problems and two separate demarcations, both of which are discontinuous with the traditional, and which, nonetheless, are frequently treated as if one demarcation and as if responding to the traditional problem.

Popper's first demarcation, which figures prominently in The Logic of Scientific Discovery and in Die beiden Grundprobleme der Erkenntnistheorie, defines scientific theories as those which, when embedded in a context which maximizes critical scrutiny, are subject to, exposed to, empirical argument or refutation or falsification by the production of an empirical counter-example. An example might be: "All orbits of heavenly bodies are ellipses," which is specifiably falsifiable by an observation of a heavenly body which orbits non-elliptically. Since it is logically possible to protect such a statement from falsification by indulging in various protective strategems (as for example ad hoc redefinitions and ad hoc auxiliary hypotheses), Popper stresses (Logic of Scientific Discovery, sections 19 and 20) that his criterion is methodological, not logical, and that any such statement, to be falsifiable, must be placed in a setting, an econiche, which excludes criticism-reducing strategems.

To be demarcated from scientific theories are other theories which -- also when embedded in a setting which maximizes critical scrutiny -- nonetheless are not subject to, exposed to, empirical argument or refutation or falsification by the production of an empirical counterexample. There are a variety of such statements and types of statements; but the two chief types discussed in the Popper School are the purely (i.e., uncircumscribed) existential statement, such as "There exists a fountain of youth," or "There exists a perpetual motion machine," or "There exists a God"; and, on the other hand, the "all and some doctrines," such as "For every mental illness there exists an effective cure," or "For every event there is a cause," or "For every crime there exists an appropriate punishment."112 It is not their form alone which renders such statements unfalsifiable; it is, for instance, easy to construct some falsifiable statements in an all-or-some form; but historically important unfalsifiable statements are commonly found in these two forms. Such statements, even when embedded in a maximally critical context, are compatible with all possible observation reports. Popper generally refers to these as nonscientific and metaphysical theories.

Popper's second demarcation -- historically the older of the two (Unended Quest, p. 41) -- is different. He also wished to exclude from science those theories (often ones which claim or aspire to scientific status) which have built-in devices for avoiding or deflecting critical arguments -- empirical or otherwise. Whereas the first kind of theory discussed could, of course, be set in a context in which criticism was deflected, in the second kind of theory criticism deflectors are built into the theory itself. Examples which Popper adduces are Freud's psychoanalysis, Adler's individual psychology, and Marxism. Freudian and Marxist theories, for instance, often include the essential assumption that ideas may be causally reduced or explained. So apologists for such views may, in the first instance, point to the sexual disposition of the critic, or -- in the second instance -- to his economic origins and status, in order to explain away the criticism. A Freudian can invoke the theory of resistances or of the censor to discredit the objections of a critic on the grounds, say, that he had not been analyzed and was, because of his own repressions, resisting the "obvious truth" of Freudian theory."113 A similar strategy might be used against a Marxist critic, to explain his "vested interest" in criticizing Marxism. There is here, in effect, a deliberate; theoretically integrated, deployment of an ad hominem argument."114 Even when such theories are set in a critical context, the theories themselves deflect criticism. When such theories are set in an uncritical context, they become quite rigid. Popper tends to speak of this type of theory as pseudo-scientific, and as "reinforced dogmatism."

The preceding discussion assumes a distinction between a position and its critical context. In particular cases it may not be easy to distinguish these, in which case the two demarcational problems may tend to merge. Thus the component parts perhaps need to be delineated and explained. First, in talking of falsification and demarcation, Popper always emphasizes that he is not speaking of individual statements; that individual statements are parts of a theory -- a system of statements -- which contextualize them; and that the question of falsifiability has to do with the theory as a whole, not with its individual components. Thus he notes that taken out of context the existential statement that there exists an element with the atomic number 72 (i.e., Hafnium) might appear to be unfalsifiable. This statement is nonetheless part of a testable scientific theory which indicates how this element is to be found. Thus in context it is a circumscribed and falsifiable existential statement (See Logic of Scientific Discovery, section 15).

In addition to this theoretical context for particular position statements there is also a methodological context for the theoretical system. This latter is what I have been referring to in speaking of critical context. One, uncritical, methodological context would allow for ad hoc redefinitions, conventionalist strategems, and such like; another, critical, methodological context would prohibit these.

To apply this to our discussion above: the theories of resistance and of the censor in Freud are part of the theoretical context of the individual statements of Freudian theory, and serve to deflect criticism from the theory. This theory, which is internally, theoretically, armed against criticism, can in turn be set in a methodological context which is either critical or uncritical.

As noted, these two types of context may blur into one another in particular cases, particularly when the theory in question is undeveloped and not yet sufficiently systematized to allow for clear distinctions. Yet it will remain important to continue to make such distinctions: knowing, for instance, that Freudian theory contains a built-in theoretical deflector against criticism, we shall -- from a fallibilist metacontext -- be particularly concerned to couch it in a critical methodological context in order to expose it, despite its theoretical defenses, to maximum criticism.

Yet this second demarcation should not; I think -- and here I differ from Popper -- be considered as one between science and nonscience. Nor is there any point to calling a method which avoids strategems for deflecting criticism "the empirical method," as Popper does (Logic of Scientific Discovery, pp. 41-2). There is of course a sense in which the presence of such strategems identifies a theory as nonscientific. Any such theory containing such strategems is noncritical; and if it is noncritical it is nonscientific. There is no objection to this -- except that it fails to demarcate science from nonscience. For a theory in which such strategems are absent is not necessarily scientific.

When defensive strategems are built into theories, such as versions of those of Freud, Adler, and Marx, which purport to be especially scientific, it may be appropriate to label them pseudo-scientific (since pseudo-critical). Yet such strategems for weakening criticism can arise in or out of science -- in ethics, metaphysics, political thought, and elsewhere. They may be present -- or absent -- in a theory which is explicitly metaphysical and makes no pretence to scientific status.115 If present, they may enable that theory to deflect a nonempirical criticism just as deftly as a similar strategem may, in other circumstances, deflect an empirical criticism. An ethical system, to take another example, is not scientific or empirical if such strategies are absent. Nor is it pseudo-scientific if they are present (it may be pseudo-critical). It may well contain devices for deflecting criticism which have nothing to do with empirical argument. For example, it may contain a theory about the nature of moral theories which entails that attempting to reform them is immoral, impossible, or rationally arbitrary.116

Thus, although the "manner of exposing to falsification, in every conceivable way, the system to be tested" (Logic of Scientific Discovery, p. 42) may be characteristic of the empirical method, it is not characteristic of the empirical method alone."The distinguishing characteristic of empirical statements" is not, contrary to Popper, "their susceptibility to revision." Precisely within Popper's methodology one gains a new opportunity to root out criticism reducing strategems, wherever and however they may appear, and to arrange one's econiche so as to expose all one's intellectual structures, scientific or otherwise, to searching criticism.

In sum, the evaluatory problem is less to demarcate scientific from nonscientific theories as to demarcate critical from uncritical theories or from theories that are protected from criticism -- particularly pseudo-critical theories. One might, if useful, call a pseudo-critical theory which purported to be scientific a pseudo-scientific theory. But in this broader context the demarcation of science and nonscience is, per se, unimportant.

Let us return now to Popper's first demarcation, which has to do strictly with the question of whether a theory, in critical context, has a potential empirical falsifier. The first demarcation does demarcate science from nonscience in an intuitively and practically satisfactory way; but it is of limited evaluational importance, and is hence noncontinuous with traditional demarcational attempts. This can be seen as follows.

Some of Popper's most fundamental work relates to how empirical observation statements bear on theories (See section XIV). With the aid of this, he is able to chart differing degrees of corroborability for competing theories based on their empirical content, and this provides a way to express preferability of one empirical theory to another (See Conjectures and Refutations, Chapter 10). All this is, however, a ranking in terms of empirical content within science.

What of those theories which are ranked outside of science by this demarcation? Two different strains regarding this question arise within Popper's work. On the one hand, one finds, especially in his earlier work, words more or less condemnatory of such theories. For instance, he writes: "Those [theories] which are nontestable are of no interest to empirical scientists. They may be described as metaphysical."117 Or he writes that "Irrefutability is not a virtue but a vice." Elsewhere he writes that the closer study of metaphysical statements is "not . . . the concern of empirical science."118 This strain in Popper's early work is most in keeping with traditional demarcational efforts, directed as they were to the problem of discriminating between good and bad theories. This strain is also closest to the positivist animus against metaphysics, and fits Popper's correct contention that the efforts of the positivists were directed to the traditional demarcational problem.

At odds with this theme in Popper is another, already present in his early work but emerging more clearly in his later work, in which the positive value of non-testable metaphysical theories is not only not denied but is strongly emphasized. He has also, in his later work, considered various ways whereby such nontestable theories may nonetheless be rationally assessed: the check of the problem is one such way (See section XVI). In such discussions, Popper emphasizes that such metaphysical theories are indeed the concern of science, and may exert an important regulative, interpretive, and programmatic influence on the development of science.119 He has also argued and dramatically illustrated the interaction between such metaphysical theories and empirical theories in his interpretation of quantum mechanics (See Part II, this series), and elsewhere.120

I would like to see all of this discussion made more consistent; and this would require that Popper's earlier antimetaphysical remarks be qualified or discarded. As is apparent from his later work, the question whether a theory is nonscientific in the first sense is often irrelevant to the question of its desirability, seriousness, interest. There is good metaphysics; and this includes Popper's own indeterminism and realism.

Once it's agreed that not all metaphysical views are illegitimate, that some may be valuable rather than their all sharing the ultimate disreputability of "meaninglessness" -- and that a deductive logical relationship can obtain between some synthetic metaphysical and scientific statements, the old attempt to demarcate between metaphysics in this sense and nonmetaphysical views loses much of its evaluative value. Rather, what now are needed are more general ecological considerations applicable to the whole range of logically interrelated claims -- metaphysical, scientific, ethical, theological which help sift theories of dubious interest from those that deserve further discussion. Empirical content, as defined by Popper, is only one among such considerations.

XXIV.

Popper's Contribution to the Problem of Demarcation

What then, we may well ask, has Popper contributed to the problem of demarcation? In this context his achievement is twofold:

He destroyed the evaluational programme of the early logical positivists, by showing that their answer to their demarcational problem (the traditional demarcational problem) was radically inadequate. But to do this is not tantamount to answering the main problem of the logical positivists. Both Popper's arguments and the positivists' views were intended to heighten criticism. In both cases the result of this intended heightening of criticism was expected to be a sorting out of competing views, some sort of sifting of "good" views from "bad" ones. The device the positivists chose had the unhappy consequence of in fact working against the criticism of theories rather than heightening it. But their chief aim was clear: to wipe out nonsense, and to put a final end to metaphysical controversy.121 Popper's criterion did not work against, i.e., diminish, criticism. On the other hand, it did not achieve the radical sort of sifting which the positivists -- and traditional justificationist- - wanted.122 I doubt that such a radical, and mechanical, sifting ever will be achieved.

Thus, Popper and the positivists -- operating from different metacontexts -- were unwittingly concerned with rather different problems and quite different general aims. The positivists wanted some way to spot nonsense and to eliminate bad theories; but Popper's falsifiability criterion was, in and of itself, not sufficient to sort out the rational from the irrational meaningful theories -- let alone sort out the meaningful from the nonsensical.

In the course of battling with the positivists, Popper did present an excellent, and l think correct, resolution of the problem of induction -- which surely has been a fundamental philosophical problem. Curiously, however, Popper's clarification of the nature of the relationship between theory and observation somewhat diminishes the importance of that relationship. For the critical relationship between empirical fact and theory has been most important historically because of the assumption that experience is the source and justification of all knowledge. Where this assumption is dropped, as it is in Popper's theory, the problem of stating the critical relationship between empirical fact and theory diminishes in critical importance, and becomes a part of the larger ecological problem of rationality.

The development of Popper's thought, and the generalization and application of his ideas outside science have then rendered his discussion of demarcation obsolete. Popper suggested to the positivists that the problem lies not in the demarcation of the scientific from the nonscientific. In fact, the problem lies not simply in the demarcation of the scientific from the nonscientific, but in the demarcation -- contextually and metacontextually -- of the critical from the uncritical or pseudo-critical.

XXV.

Research Programmes. Kneale, Lakatos, Maxwell

Three of the contributions to the Schilpp volume those of Kneale, Lakatos, and Maxwell -- are concerned importantly with demarcation.

Of these, the essay by Professor William C. Kneale, of Oxford, on "The Demarcation of Science" can be dealt with rapidly, since most of it depends on a trivial misunderstanding of Popper: namely, that only universal statements fall within science as conceived by Popper, and hence that all historical statements are for Popper metaphysical. The discussion that Kneale weaves in terms of this misunderstanding contains some interesting points and examples, but has nothing to do with Popper.

The late Imre Lakatos's contribution, "Popper on Demarcation and Induction," is a poor, even perverse, paper. I indicated above some of the trivial errors in scholarship and reporting that mar it. And there are others: for example, Lakatos utterly misdescribes what Popper means by "crucial experiments," giving as an example an ordinary noncrucial experiment, and interpreting the word "crucial" in a psychological sense (p. 249). Lakatos also misdescribes the history of Popper's account of truth, and of his relationship to Tarski. But I do not want simply to catalogue Lakatos's errors here, but rather want to talk briefly about the main themes of his paper, which have to do with demarcation and "scientific research programmes.''

Lakatos came to the United Kingdom from Hungary in 1956, met Popper in 1958, and became his colleague and an assistant lecturer in the London School of Economics in 1959. By the 'forties, Popper had begun to develop a cosmological or metaphysical outlook -- the theory of propensities -- which he described as a "metaphysical research programme." As he developed his theory, and traced its historical background, he began to notice such metaphysical research programmes everywhere in the history of philosophy and of science; and the important role of such programmes formed a dominant theme of his writing and teaching in the 'fifties; the important final chapter of his unpublished Postscript to the Logic of Scientific Discovery, written in 1954-5, deals with these programmes. At this time, Popper's assistant, later colleague, Joseph Agassi, wrote a distinguished dissertation on a phase of these programmes, particularly in their relationship to Faraday - "The Function of Interpretations in Physics" (London, 1956); and Popper's colleague J.W.N. Watkins published an important series of articles in the late 'fifties devoted to metaphysical research programmes.123 This was one of the most important topics of discussion at the London School of Economics when I arrived there in l958.

In his work Popper stressed the importance of these programmes for science, emphasizing their regulative, programmatic, interpretive effects. He called them "metaphysical" precisely because they were not empirically refutable, and developed certain other means (e.g., the check of the problem: section XVI) to criticize them. He also emphasized their dangers, giving as his own most dramatic and important example, his analysis of the effect of such interpretations and programmes within quantum physics. This work of Popper's is not widely known outside the Popper School, since the most important part of it was presented in The Postscript and in his lectures inside the school; however, it was common knowledge.

Lakatos took this material, and first incorporated it into a brilliant historical study in mathematics.124 Later he changed the name "metaphysical research programmes" to "scientific research programmes," and began to argue that it provided the basis for a decisive rejection of Popper's methodology: for instance, pointing out how these programmes can deflect falsifications (anomalies), he claimed that refutation and falsification played a minor role in science, and that Popper's demarcation criterion, which emphasized falsifiability, was consequently defective. All of this lies at the root of Lakatos's misrepresentation of Popper's position. Popper left these research programmes out of science because they were not falsifiable; Lakatos put them into science, and then discovered that "science" was not falsifiable!

Lakatos could not have developed, his "original" approach in this way were it not for the considerable sloppiness and inconsistency in Popper's own presentation of theory of demarcation, as pointed out above. Taking advantage of this, Lakatos systematically distorted and garbled Popper's approach.

The way this works can be illustrated simply. Take the problem of basic statements discussed above (Section XIV). I have tried to argue that the "decisions" and "agreements" that Popper introduces there (and in certain other parts of his work, as in his fideistic statement of "critical rationalism") are superfluous remnants of justificationism, are out of line with the main thrust and intent of his methodology, and may be dropped without loss, and with considerable improvement in consistency, clarity, and generality, in the position as a whole.

Lakatos proceeds differently. In his contribution to the Schilpp volume, without any argument, he pounces on these agreements and decisions, treats them (without argument) as essential to Popper's outlook, and systematically reinterprets Popper's approach accordingly. Thus he emends Popper's demarcation criterion to require that not only basic statements but also universal statements must be accepted by agreement as conventions. From there Lakatos is embarked on a whole-scale conventionalism, which makes scientific hypotheses subject not to the veto of nature but rather to the votes, and the "value judgements" of scientific "elites." Thence Lakatos pushes, evidently without realizing it, back into the familiar justificationist questions of authoritarian epistemology. The result is much closer to the philosophies of Michael Polanyi and T.S. Kuhn than to that of Popper. The reductio ad absurdum of Lakatos's approach is to be found in his contention that Newtonian theory is no more falsifiable than is Freudian (a contention which, Popper decisively rebuts in his response to Lakatos).

Grover Maxwell's interesting paper, "Corroboration Without Demarcation," which I mentioned earlier in the discussion of Goodman (Section XVI), argues that Popper's criterion of demarcation is too restrictive, remarking that a number of statements that seem to play an important role in science are either pure, unrestricted existential statements or "all-and-some" (mixed quantifier) statements which cannot be falsified. Such statements play such an important role in science, he argues, that virtually no scientific theories are falsifiable. Thus Maxwell would prefer to see falsifiability as an important but usually unattainable regulative ideal for theories rather than as a basis for demarcating science from nonscience. One of Maxwell's examples -- "All men are mortal" -- comes from ordinary experience; another from thermodynamics: "Every solid has a melting point," or "For every pure solid at a given external pressure, there is at most one equilibrium liquefaction point."

In replying to Maxwell, Popper states an interesting principle which I do not believe he has published previously or elsewhere: "Whenever a pure existential statement [or all-and-some statement] by being empirically 'confirmed' appears to belong to empirical science, it will in fact do so not on its own account, but by virtue of being a consequence of a corroborated falsifiable theory" (p. 1038). Thus, "All men are mortal" is seen to be a consequence of the falsifiable statement: "All men die before their 150th year." Similarly the unfalsifiable statement about liquefaction may be derived from the falsifiable statement: "For every pure solid substance there is at most one liquefaction point at which mechanical disturbance has an insignificant effect." It is not enough, Popper adds, to subsume a statement under another statement that is strengthened ad hoc: rather, the strengthened theory must belong to science in its own right."125

XXVI.

Theory of Corroboration

I have left until last that part of Popper's theory of, science which he himself links most closely with rationality: degree of corroboration. It is that aspect of his treatment which I find least valuable.

With part of the theory I agree: namely with his argument that degree of corroboration (like degree of corroborability) cannot be a probability and cannot conform to probabilistic theories of continuation. Also valuable is his characterization of degree of corroboration as a critical summary report on the past performance of a theory under severe critical test by experience.

"Belief is never rational: it is rational to suspend belief." (Schilpp,p. 69)

In The Logic of Scientific Discovery, however, Popper writes as follows (and there are similar passages in his later writings):

"degree. of corroboration . . . may be interpreted as a measure of the rationality of our beliefs." (p.414)

"degree of corroboration . . . must not be interpreted . . .as a degree of the rationality of our beliefs in the truth of h . . . Rather, it is a measure of the rationality of accepting, tentatively, a problematic guess." (p.415)

This is worse than confusing. There is no theory of belief in Popper's work into which one might fit and interpret these passages. And although one could, ad hoc, assign them a meaning, they come very close to being, in context, objectively incomprehensible.

This is particularly so since, on a straightforward prima facie interpretation, they are both false and at odds with the rest of Popper's thought. For Popper, in his later work, stresses that metaphysical theories are arguable and rationally criticizable: yet metaphysical theories having very low empirical content, being unfalsifiable cannot be corroborated (although they can be severely criticized using non-empirical means, such as consistency, response to the problem-situation, and such like: see section XVI above). Thus these passages would imply that it would be irrational, or of a very low degree of rationality, to accept -- however tentatively -- any such metaphysical position. Which is in turn incompatible with Popper's clear intention, in his later work, that metaphysical positions be entertained most seriously and rationally.

If one is to avoid unprofitable hermeneutics here, one must ignore these passages. A Popperian account of corroboration could and should be worked out afresh without being bound to Popper's own unfortunate mode of expression. The trouble with Popper's theory of corroboration, as it stands, is that corroboration is only a report on how the theory is empirically tested, and does not consider the other ways in which it may (nonjustificationally) have been examined.

Others particularly Imre Lakatos and Joseph Agassi 126 have objected on similar ground to Popper's theory of corroboration. Lakatos quite fairly calls Popper's theory of corroboration as degree of rationality of belief ''a slip'' and points out, somewhat helpfully : that "theory of rationality must be about rational action, not about 'rational belief.'"127 In his intellectual autobiography, in the Schilpp volume, Popper follows Lakatos on this point, speaking of corroboration there in terms of rational action rather than of rational belief. Yet this distinction between action and belief is itself by no means obvious nor is its meaning clear. There are, for instance, plenty of philosophical accounts of "belief-acts." And in the absence of a theory -- which would provide objective and subjective settings and context for these vague words -- about belief, action, preference, agreement, decision, and the other loose coinage of Popperian discourse about these matters, one hardly can tell how to interpret these discussions or the controversies about them. As it stands, Popper's theory of corroboration is of limited usefulness, and provides a guide neither to rational belief nor to rational action.

In the next instalment of this study -- Part IV -- I shall turn to Popper's contribution to historical and social philosophy and to intellectual history.

CAUFORNIA STATE UNIVERSITY, HAYWARD

HAYWARD, CA. 94542 U.S.A.

W.W. BARTLEY, III

NOTES

There are over 100 notes which are not reproduced here at this stage due to the time required to scan and edit fine print. Rafe Champion is prepared to post hard copy of the notes to readers who do not have access to the journal in a university library. Contact him on rchampATbigpond.net.au

BACK TO THE START OF THE PAPER