Where do dialectical rules derive their authority, and who enforces them? The answer to these questions is simple, if a little disquieting in its ultimate implications. Although there are special circumstances in which there may be a Chairman, a Judge, or others whose job it is to control proceedings, in ordinary discourse there is no such person. The control of each dialogue is in the hands of the participants themselves.

There is clearly room for dispute between participants over how their dialogue should be conducted, and no dialogue will be possible at all unless there is a certain minimum of procedural agreement. We cannot legislate against all the possible abuses of dialectical procedure, and there would be little point in doing so even if we could. Provided, however, disagreement is not extreme, participants will often resolve their differences by means of one of a set of procedures that are themselves characteristic of a language or culture. There are accepted forms for the lodging of objections, their debate, and their resolution. The point of order, or procedural locution, is as much a part of ordinary language as it is of the formal rules of meetings and committees. 'What shall we discuss?', 'That doesn't follow', 'Let us leave that aside for the moment', 'Proceed!', 'I don't understand', 'That is irrelevant', 'Wait, you're going too fast for me', 'It's not for me to say' -- these and similar locutions contribute not to the subject or topic of the dialogue but to its shape. We could call them 'metalinguistic' if it were not that this word is too broad and invites confusion with other dialectical phenomena such as locutions in quotation and remarks on the dialogue by onlookers. (The concept of 'metalanguage' needs rethinking in a dialectical context.) I shall call them simply points of order, and contrast them with topic points. For various reasons we cannot always make a clear distinction between the two kinds of point, but it is useful to be able to do so on occasion.

Now it should be clear that the commonest use of logic and logical terms is as an aid to the making and debating of points of order. One's primary reason for needing to say that one statement implies, contradicts, supports, generalizes or exemplifies another is that one wishes to attack or justify certain dialectical moves. Among logical terms I include, for this purpose, the names of fallacies. Not the least of the merits of a really good classification of fallacies would be that it could be used in the formulation of appropriate points of order; and, if existing classifications are not much used for this purpose, so much the worse for them. It should be made possible in principle, as Bentham wished, that the perpetrator of fallacy be greeted with 'voices in scores crying aloud "Stale! Stale! Fallacy of Authority! Fallacy of Distrust!" and so on'.

In so far as there are accepted forms for the raising of points of order it must be possible, in principle, to reduce the forms to rule. The infinite regress latent in such a move may be avoided by various means, and does not, in any case, constitute a fundamental objection so long as we properly understand that our 'rules' are not to be conceived as rigidly enforceable in practice but rather as in the category of norms or conventions. As elsewhere in Dialectic, we have a choice where to draw the line between regarding a locution as entirely outside the permitted forms, or regarding it as legal but open to a particular reply; in this case, the raising of a point of order. From one point of view, communication may be said to have broken down between participants when one of them is regularly resorting to, say, argument in a circle; but, from another, the minimum conditions of communication are satisfied so long as properly formulated objections are listened to and answered. All that follows is that there is a certain openness about any set of rules.

These remarks prepare us for an assault on the Dialectic of Aristotle's first class of Fallacies, the Fallacies Dependent on Language; for the problem of formulating sets of rules in respect of meaning-constancy and equivocation is a difficult one, and such rules as can be offered are, as we shall see, essentially rather tentative and 'open'. But first we should reflect on why it is that the discussions of the previous chapter did not touch on these Fallacies. In describing entities denoted by letters 'S', 'T', 'A', 'B' as 'statements' we were relying on twentieth-century logical tradition to give that term meaning. The denotata of the letters are 'statements' in the sense that they are two-valued, may be combined in truth-functions, concatenated in proofs, 'asserted' as theorems, and so on. But in taking over this tradition we have also used these entities as components of 'locutions' of various kinds, and some but not all of these locutions have themselves been identified as 'statements' under a criterion involving their role in dialogue. We should not try to have things both ways. If the letters and formulae of the statement calculus really represent statements this must be because of their potential or hypothetical role as locutions in a dialogue and the part they would play if so used. This, at least, is the methodological assumption we should make as we explore dialectical systems; though some people, no doubt, would prefer the converse thesis that statements in a dialogue are such because they represent entities with the logical properties of those of statement calculus.

At all events, the smooth phrase 'Let S, T, U, . . . be statements of a logical system' begs this question and can be justified only in a preliminary exposition. If we want to lay bare the foundations of Dialectic we should give the dialectical rules themselves a chance to determine what is a statement, what a question, and so on. This general idea is familiar enough from Wittgenstein.¹ I do not think, however, that it has ever been worked out in any detail. The programme is too large a one to be undertaken here but certain features of it are of fundamental importance for us.

The thesis that I shall adopt is that all properties of linguistic entities are 'dialectical', in the sense of being determinable from the broad pattern of their use. We might call this thesis the Dialectical Theory of Logical Form or, perhaps, the Dialectical Theory of Meaning. And I should emphasize, also, that an interest in formalized dialogues does not in itself commit us to this thesis. It is perfectly possible for someone to feel that there is much to be gained from studying the combinations of symbols in dialogue -- in something the way Carnap envisaged 'pragmatics' as an extension or special application of 'syntactics' and 'semantics'² -- while holding that meanings are states of mind, or that implications are relations between objective fact, or that logical truths are unchangeable and are perceived by a special intuition. However, I think it is true that someone who holds one of these other beliefs will be unable to take a dialogue quite seriously as a logical phenomenon and will tend to regard it as, at best, a public performance which mirrors or plays out essentially private or non-immanent processes. Such a person could countenance most of our previous chapter, but will disagree with part of what I have to say in the present one. Moreover, I shall not here rehearse any of the arguments which might convert him. If anything is here done to convert such a person, it will only be in the demonstration that the task of applying this point of view to our present problems can be plausibly carried through.

Now Equivocation, if we think of the meanings of sentences or terms as extralinguistic entities, becomes in essence the association of a single sentence or term with two or more such entities instead of one; but an approach such as that of the previous chapter, by locating most of the properties of the locutions in propositional letters such as 'A', 'B', 'S' and 'T' smuggles in the (what I take to be) fiction that the question of meaning can be isolated from that of dialectical properties. When the letter 'S', say, is used twice or more in a given example it is by convention the case that it has the same meaning at each occurrence; but, if meanings are to be allowed to change with context, and to be determined by the extended context, the question of whether the meaning of a given symbol changes is to be answered a posteriori and the question should not be begged by writing in an assumption of constancy. Where there is meaning-constancy it must be possible to deduce the fact from external considerations, and where there is equivocation the same must apply.

What are the external criteria of meaning-constancy?

(A) The most obvious way of finding out what a speaker means is to ask him. When he says 'By S I mean X',³ provided we are not in doubt what he means by X, this normally settles the matter and, if the context is an argument in which S might have been equivocal, we can turn our attention to the actual truth or falsity of the statements in it, with S interpreted as X. For example, a military commander can 'locate' one of his units either in the sense of finding out where it is, or in the sense of deciding where to put it; but, if a staff-officer ambiguously reports 'Battalion B has been located at grid-reference G', he can always be asked which of the possible senses he intends. We can imagine that there might be a slippery communication-situation in which someone, or some chain of people, effectively reasons:

	(1) Battalion B has been told to go to G;
that is	(2) Battalion B has been located at G;
therefore	(3) Battalion B must be at G.

Battalion B may not be at G, the enemy may strike and win the battle; but the bearer of (2), if he has survived, can be asked 'Did you mean "located" in sense such-and-such or sense so-and-so?', and all may then be made plain. Possibly, of course, no one will ever ask him, and in this case we shall never know for certain what he meant; but this is a practical point that need not concern us so long as we are sure that a suitable answer would have been forthcoming if the question had been put.

Let us formulate this first attempt at a dialectical meaning-criterion, then, as 'What a person means by his utterances is what he tells or would tell you he means if asked'. This is a better and more powerful criterion than it at first appears, for most of the objections that are most naturally lodged against it are quite easily answered. Consider, first, the objection that 'By S I mean X' merely defines one linguistic entity in terms of another, so that we remain imprisoned in language. This complaint arises from a too narrow conception of what it is to explain the meaning of a word: there are well-understood means such as ostensive definition which do get outside language for us, and it is not at all unusual for the meaning of a family of terms to be made clear simply on the grounds of their mutual relations. Otherwise, no one could ever learn a language by the 'direct' method.

Secondly, it could be objected that locutions such as 'By S I meanX' occur only at a very sophisticated level of language, and that we could never come to learn to use them until we had independently learnt the meanings of large numbers of more elementary terms and idioms. The answer to this objection, however, must be straight denial; for although the meaning of the word 'mean' may perhaps be a sophisticated matter, the processes of explaining a meaning, giving a synonym, and so on, are among the most primitive processes of language. It is hardly conceivable that there should be a practically usable language that did not contain, or have associated with it, idioms permitting the explanation of meanings.

Thirdly, it might be objected that our formulation does not apply to the meanings of the linguistic entities themselves but only to 'what the speaker means'; and that language is not tied to any given one of its users or occasion of its use. There are two answers to this. The first is that the thesis as it stands is actually a little narrower than it need be, but that it can easily be broadened. The meaning of a term is, it is true, not just a matter of what a speaker or writer intends it to mean but also a matter of what a hearer or reader understands by it, what an average speaker or an average hearer would mean by it in normal circumstances, and so on; but we can determine what a hearer takes a word to mean by asking him, and we can determine what an average speaker, hearer, reader, or writer means by conducting a poll. We can, perhaps, determine what a fictional speaker or hearer means by devising an imaginary explanation-request and answer in accordance with the canons of fiction. Thus 'hearer's meaning', and the others, can be defined analogously to 'speaker's meaning'. But a second answer is appropriate to those who remain unsatisfied with this extension of the criterion and insist on asking for a definition of 'the meaning'. It is to ask: What other kind of meaning can there be? Besides the various possible explanations of meaning that could be elicited by asking users, it is difficult to see that any other source of explanations is possible. This pure Platonism must be rejected.

Thus the criterion stands up reasonably well to some of the possible objections. Unfortunately, there are difficulties of another order. It should be clear, for one thing, that there are often cases in which we feel that what a man means is not what he says he means; and this implies some different test of meaning from the one we have given. More disturbingly, it sometimes happens that when a speaker is asked to clarify an ambiguous statement he is unable to do so because the confusion is more than verbal. Let us return to the example of the military commander. Staff officers and their aides are apt to conceive units as chessmen that can be moved at will, and someone whose reasoning is directed by this model will be apt, to this extent, to recognize only one sense of the word 'located'. Asked to distinguish two senses he will at first be at a loss. If so, it seems to follow that a person may sometimes not know what he means or, at least, that he may be unable to give the kind of detailed account of his meaning that is necessary to resolve questions of meaning-constancy or equivocation. In short, we shall have to abandon this criterion and look for some other one which includes an analysis, among other things, of what it is for someone's thinking to be model-dominated.

(B) If, then, we cannot rely on what a speaker says he means, to what extent can we determine meaning from use in (what must be called) 'zero order' contexts? If a word W is under study and we disregard its occurrence in such contexts as 'By W I mean so-and-so' and turn instead to its primary application in such contexts as (if it is a common noun) 'All (W)s are (Y)s' and 'So-and-so is a (W)', is it here that questions of meaning are resolved? (In these formulations the parentheses represent syntactical 'anti-quotes', such that when the expression within them refers to a word the expression including parentheses refers to the word's reference.)

There are two well-known answers to this question in recent philosophical literature. Both involve the recognition that some statements, more than others, tend to reflect the meanings of the words in them in the sense that their truth or falsity is relatively little dependent on empirical or extralinguistic fact. Quine ('Two Dogmas of Empiricism'), though he denies that there is a coherent distinction to be drawn between 'analytic' and 'synthetic' statements, erects, in its place a distinction between a higher or lower degree of 'corrigibility' or 'immunity to revision'.⁴ Given a body of beliefs, and an experience in some way at variance with them, we always have a choice of different ways of adjusting our beliefs to accommodate the experience; and in making this choice we give favouritism to certain beliefs and prefer, if possible, to adjust certain others. The favoured, or relatively 'incorrigible', beliefs include the higher-order best-entrenched theoretical ones, particularly those of Mathematics and Logic. The less-favoured or 'corrigible' ones are the ones whose reversal creates no great strain in the system of beliefs as a whole, such as those of ill-confirmed immediate experience.

Grice and Strawson ('In Defense of a Dogma'), though apparently defending the distinction between analyticity and syntheticity against Quine's attack, given an account of it which also makes it a relative matter. This time, however, it is not the entrenchment or corrigibility of statements that is emphasized but rather the comprehensibility of their negations. The man who says 'My neighbour's three-year-old son does not yet understand Russell's theory of types' says something synthetic since, however surprised we would be to find it false, we can still give some meaning to the hypothesis that it should be so; but 'My neighbour's three-year-old child is not yet an adult' is such that, if someone were to deny it, we would find his statement incomprehensible. 'You mean that he is unusually well-developed?' we would say, or 'You mean that he has stopped growing?'; but if glosses of this kind were all rejected we would have no option but to regard the statement as meaningless altogether. Grice and Strawson do not intend their account as a complete one but, for what it is worth, it reduces analyticity to a behaviour-pattern of the hearer in reaction to an actual or hypothetical denial, and is generically similar to Quine's. Both accounts are 'dialectical', in that they refer their respective explications of analyticity or incorrigibility to patterns of verbal behaviour. Quine, it is true, thinks in terms of an average or corporate behaviour of modern scientists, and Grice and Strawson think rather of individual idiosyncracies; but in both cases it is clear that questions of meaning are to be resolved into questions of analyticity or incorrigibility of verbal formulations, and these, in turn, into behaviour patterns.

Expanded to meet our present preoccupations, the account must go something like this: Meanings of words are, of course, always relative to a language-user or a group G of language-users. The meanings any group G attaches to words are determinate in terms of the (zero-order) statements that are relatively incorrigible or analytic to G. Two words W and X have the same meaning for G if members of G, in the face of any experience, consistendy and stubbornly allocate the same truth-value to any zero-order statement containing X as they do to the corresponding statement containing W and, in particular, if they consistently and stubbornly maintain the truth of 'All (W)s are (X)s and all (X)s are (W)s'; and, perhaps (though this is independent), express puzzlement when faced with statements such as 'Not all (W)s are (X)s' or 'This A is an (X) but not a (W)\ The words W and X have different meanings for G if there is a zero-order statement containing X to which they are quite prepared to allocate a truth-value different from the corresponding one containing W (etc.).

There is a reverse side to this doctrine, which needs to go as follows: Since the language-behaviour of some person or group may be unsystematic or incoherent, it is not necessarily the case that questions of meaning are resoluble. If some members of a group assent to, and others dissent from, certain statements it may not be possible to say, for that group, either that W and X are synonymous or that they are not. Furthermore, the statements even of an individual may be mutually 'inconsistent', in a primitive sense of that word, namely, that they do not fit together into a pattern. It is only in so far as a regular pattern of use can be determined that it is possible to make suitable judgements about meaning.

Now let us turn to equivocation. What, in the zero-order uses of a word W by a group G, could lead us to judge that W is equivocal? The question is not easily answered, and we must subdivide it a little. Let us start by asking how we could determine that a word W has two or more different meanings. Some obvious tests present themselves: chairs and questions can both be 'hard', but it is easy enough for us to sort out two meanings on the grounds of categorial difference using, if necessary, the refined methods of Sommers ('The Ordinary Language Tree' and 'Types and Ontology'). Even when no relevant categorial difference presents itself as, perhaps, in the case of 'bank' in 'I'm going down to the bank', it may be the case that the pattern of use splits sharply into two sub-patterns: the occasions on which people talk about financial institutions and the things they say about them seldom overlap with the occasions of their references to, or their statements about, river-verges. In such cases the distinguishing of two meanings is easy but seldom necessary and hence in a certain sense gratuitous.

Equivocation differs from double-meaning because, first, we must assume the existence of an invalid argument based on meaning-shift and, secondly, because we must assume that the perpetrators of the argument either deceive themselves, or set out to deceive other people, into thinking the argument valid. But equivocation, as we remarked in chapter 1, may be of two kinds, gross and subtle. It is, presumably, a sufficient account of the gross kind that the double-meaning involved should be as just described. Since it is only per stupiditatem that anyone is ever deceived by them, equivocations of this sort are of litde interest to the logician.

The subtle variety is a different matter. Let us revive the example we used in chapter 1. Someone says 'Ignorance of the law is no excuse', and says it disapprovingly, in such a way as to leave no doubt that he regards ignorance as morally blameworthy. We may suppose, in fact, that he regularly and systematically fails or refuses to make a distinction between legal and moral obligations, and accepts the consequences. His reasoning contains arguments that others consider equivocal; for example

All acts prescribed by law are obligatory.
Non-performance of an obligatory act is to be disapproved.
Therefore, non-performance of an act prescribed by law is to be disapproved.

The point is a quite general one. When someone reasons, syllogistically, say, and we wish to condemn his argument as an equivocation on the middle term, our grounds for doing so will be that we consider the premisses actually or possibly true and the conclusion actually or possibly false; but these cannot be grounds which will appeal to the person who puts the argument forward, since he must be supposed to be deceived by, or out to deceive others with, the argument. If, on this theory, we suppose an argument to be capable of deeply deceiving someone, we thereby suppose it to be capable of creating for him a whole pattern of use of the words involved in it; and thereby destroy the supposition that the words are, for him, equivocal.

Our conclusion could be that the theory that a man's zero-order use of language determines what he means by his words is untenable. But it could also be that there is no such thing as a deep or subtle equivocation. Before exploring further, let us see what other dialectical tests of meaning-constancy are possible.

(C) Sometimes cases arise in which we want to say that someone is deceived by an argument 'temporarily', or 'against his better judgement', and that later or quieter reflection leads him to a reappraisal in which he sees the fault; perhaps, an equivocation of the middle term.

This kind of case needs to be mentioned but, as a test for equivocation, it is really outside our terms of reference. The temporary nature of his assent to the conclusion is not in itself an argument for its invalidity; the fact that his repudiation of it was later, or quieter, is not an argument for its correctness; and we only beg the question if we refer to the argument as a 'deception' or say that he later 'sees the fault'. We have not yet shown that these terms have a dialectical analysis.

In fact, the whole question of a dialectical theory of truth and falsity might be said to be open. The simplest of all demonstrations that an argument is invalid -- and hence, perhaps, equivocal -- is the demonstration that its premisses are true and its conclusion false. Dialectical considerations, however, do not provide such tests since they do not provide criteria of truth or falsity for more than a very restricted class of the statements we make.

If there is a dialectical theory of truth it must run something as follows: 'It is true that S' means (very nearly) the same as 'S', and 'That is true', 'That is false' are phrases used in dialogues to indicate agreement and disagreement. 'Is it true that S?' is (very closely) the same question as 'S'?, and so on. When the abstract nouns 'truth' and 'falsity' are used they are translatable, not always very directly, into more elementary terms, and hence dissoluble into locutions relevant to actual or projected cases of agreement or disagreement. 'S and T have the same truth-value' means (approximately) '(I agree to) T if and only if (I agree to) S'.

Now the assumption that the premisses of some projected argument are true and the conclusion false is an assumption we non-participants make and which cannot commit the participants. If, on the other hand, our hypothetical arguer who trusts his better judgement comes to this conclusion about truth-values, this says nothing except that his use of words, in so far as it is a quiet and reflective one, is not such as to lead him to regard the argument as valid, even though his use in the heat of the moment may have been different.

(D) Whatever theory is adopted, it must explain a certain asymmetry between 'Yes' and 'No' answers to questions of meaning-constancy. Although there is no contradiction in supposing of quite ordinary words in quite ordinary contexts that they are equivocal and misleading, we almost never suppose any word to be equivocal until we get into trouble with it. When two people disagree over some kind of inference, charges of equivocation may begin to flow; but, when we are in agreement and think that there is nothing to disagree about, to envisage simultaneously that there is or might be a hidden equivocation is inappropriate for us because it is unnecessary. The onus of proof is on the side of the person who makes such an assumption, and the assumption is equivalent to an assumption of the possibility of disagreement.

Consequendy, a theory of meaning-constancy that can be applied independently of whether the constancy actually matters or makes a difference must be fundamentally on the wrong tack. There is, as we might put it, a presumption of meaning-constancy in the absence of evidence to the contrary. The presumption is a methodological one of the same character as the legal presumption that an accused man is innocent in the absence of proof of guilt, or that a witness is telling the truth: it is not, of course, itself in the category of a reason or argument supporting the thesis of meaning-constancy, and least of all is it an argument for the impossibility of equivocation. Dialectic, however, has many presumptions of this kind, whose existence is related to the necessary conditions of meaningful or useful discourse. It is a presumption of any dialogue that its participants are sober, conscious, speak deliberately, know the language, mean what they say and tell the truth, that when they ask questions they want answers, and so on.⁵

Now the presumption of meaning-constancy -- and we might say that this presumption applies with especial force in the case of two adjacent uses of a word in the statement of an argument -- is not inconsistent with the doctrine that the meaning of a word is the pattern of its use, or even with the doctrine that a man's meaning is what he says it is; but it gives any such doctrine a new character and a special twist. We may have to say, for example, that in so far as there is a presumption that W is constant in meaning there is a presumption that any given use of W is part of a pattern, or that the user's explanations of his meaning are mutually coherent. And if a given use of W by person P seems not to be part of a pattern of use by P, it is always possible that there is a developing pattern and that a retrospective assessment may alter the verdict. In conjunction with these other doctrines the presumption of meaning-constancy implies that an apparent equivocation always carries with it a presumption that there is an actual or developing meaning of the locutions in question, of such a character as to render the suspect locution non-equivocal.

This makes the formal logician's assumption of the meaning-constancy of his symbols at once both more and less reasonable; more, because his assumption is our presumption; less, because it assumes that meaning-constancy is axiomatically and absolutely achievable.

When we reflect that charges of equivocation are themselves dialectical locutions, and that they are procedural in nature -- that is, points of order -- we may also be led to wonder whether a better analysis of them might not be given in terms of their procedural role, resting on features not realizable at the topical level. Points of order are our means of pressing an opponent to adopt this or that procedure or to give the dialogue this or that shape, and they may also be our means of pressing others to adopt certain uses of words rather than others. We perhaps assume too readily that a point of order, when it raises an objection to some locution, stigmatizes a prior fault. Though it may be appropriate to say so of an objection to a gross equivocation, the subtler variety should sometimes be regarded less as a fallacy than as an idiom in disuse.

There are two ways in which 'points of order' might be incorporated, at an elementary level, into a formal, dialectical language. The first would be to treat each kind of point of order on its merits, as a special locution with prescribed functions; as, in committee procedure, there is one, stylized form of closure, 'I move that the motion be now put'. The other would go to the other extreme and provide a procedural metalanguage with all or most of the facilities of the object language but including also means of referring to locutions of the object language and dissecting their relevant properties. Neither is quite realistic. Procedural locutions do not, as in the first model, lack a grammar; but neither are they, as in the second, to be regarded as debatable in quite the same sense as topical ones. The consideration of points of order is essentially perfunctory. There is something wrong with a dialogue in which points of order are endlessly debated, or in which they are subject without restriction to yet further points of order.

We should now remind ourselves that much of what we have been saying is implied in Sextus's criticism of the concept of fallacy. For Sextus, even proper names do not have meanings independent of the context of their use, as he makes clear with the example of the two servants called Manes. The order 'Fetch Manes' is unambiguous if it is given when only one of the two servants is on duty, but otherwise the person to whom the order is given will have to ask 'Which one?'. In the same way, the order 'Bring me some wine' can be carried out without question if there is only one kind of wine available, but needs to be re-referred back when uttered by a man rich enough to have two. The name 'Manes' and the phrase 'some wine' are, in themselves, neither ambiguous nor unambiguous and neither, we must assume, is any other word or phrase. What matters is that locutions involving them should play their appropriate, or demanded role; and, provided they all do this, it makes no sense to explore their meaning any more finely.

So much, at least, Sextus says about the question of what a word means on a particular occasion of its use. But it is a reasonable guess that this is also what he would say about the question of the meaning of a word: words have no meanings apart from the meanings they bear on particular occasions and though, if a pattern of use develops, we may describe the pattern and discover an average or norm this norm is no more than the expression of the demands for satisfactory communication on the individual occasions from which it is derived.

Sextus, however, goes much further than we have so far done, in that he makes, or seems to make, a criticism not merely of the doctrine that terms can be clearly equivocal or unequivocal but also of the whole concept of a fallacy. To be specific, he makes criticisms which, if they are to be sustained, can be interpreted as affirming the dialectical character not merely of the meanings of terms, but of logical form itself.

Let us consider further his example, quoted in chapter 3, to prove the uselessness of a logical doctrine of fallacies:

In diseases, at the stages of abatement, a varied diet and wine are to be approved.

But in every type of disease an abatement inevitably occurs before the first third day.

It is necessary, therefore, to take for the most part a varied diet and wine before the first third day.

The argument, he tells us, is valid so far as the logician is able to tell; for only the physician with his special knowledge will be able to see that the word 'abatement' is equivocal and refers, in one case, to the general abatement of the disease and, in the other, to the periodic troughs in the fever-cycle. The physician sees this because he knows the conclusion to be false. It is not totally anachronistic to conceive the 'special knowledge' of the physician as a knowledge of empirical fact, unavailable to the logician since he deals only with the a priori. In modern terms, then, what Sextus is saying is that the inferences licensed by the logician carry no authority except what they derive from their conformity to our independently-derived preconceptions of the truth and falsity of the statements occurring in them.

Thus, suppose there were a road leading up to a chasm, we do not push ourselves into the chasm just because there is a road leading to it but we avoid the road because of the chasm.

Properly understood, this says something about the whole status of the logic of inference. We can underline Sextus's thesis by directly raising the question of the application of a logical inference-schema. Let me suppose that I accept a certain premiss or premisses P and a certain inference-process from the application of which I should be led to deduce conclusion Q, but that I have no independent belief either way regarding the truth of Q. Should I accept Q? The question seems to be without sense: assent to P and assent to the inference-process implies assent to Q. But does it? It depends on what is meant, precisely, by 'assent to the inference-process'. There is a possible sense of this phrase such that it is not possible for someone to assent to premisses and inference-process without assenting to the conclusion, or such that a man who accepts premisses but not the conclusion cannot properly be said to have accepted the process. However, acceptance of an inference-process may also be 'formal' only, and it is in this sense that we customarily accept the schemata put before us in Logic-books. Having put up their schemata, the writers of Logic-books are sometimes only too conscious that they cannot, as it were, hand out an unconditional guarantee with them: there must be a saving-clause against improper use. A great deal hangs, then, on the question of what use is 'proper' and what not.

That Formal Logic cannot formalize its own application needs no argument: it takes an enterprise of a different order to do that. But the point that Sextus the sceptic feels bound to make for us is that this new enterprise cannot hand out guarantees either. At least sometimes - and, Sextus perhaps thinks, always - the discovery that a given inference is invalid is made a posteriori, from independent knowledge of the falsity of the conclusion.

The problem is particularly pressing in the case of 'fallacious' arguments since, by definition, they are arguments that seem valid. In the grosser kind of fallacy the deception may be a gross one, which is easily dissipated and seen for what it is; but in subtler fallacies it may not be possible to be sure of the source. In fact, if the test of validity is in any way independent of the actual truth and falsity of premisses and conclusion, the subtle fallacy shades into the valid inference, with no possibility of drawing a dividing line.

Let us sharpen up our example and suppose that, from premisses P which I accept, by a process I accept, I am led to a conclusion Q that I reject: What am I to say? When I am asked whether I think Q to be true, the only answer I can conscientiously give is 'No', and to say anything else would be dishonest. Similarly, if I am asked whether I think P to be true, I must say 'Yes', for to give any other answer would be dishonest too. What, now, if I am asked 'But doesn't P imply Q?'? I shall be in trouble if I merely answer 'Yes' to this, but it would be possible for me to say something like: 'It seems to me that the inference from P to Q is valid. Assuming that I am right in thinking P true and Q false there must be something wrong with this inference, but I am unable to say what it is.' What might it be? If our interests are logical we shall, no doubt, examine the details of the inference closely: check the formal inference in accordance with recognized logical systems, double-check our formalization, run through lists of fallacies to see if they give us any clue to the failure. But what if all reasonable efforts fail? One reaction might be to label the inference a 'paradox' and regard it as a 'difficulty' for the logical system within which it is most conveniently analysed, but this would concede the a posteriori nature of logical investigation. It is difficult to conceive of any other reasonable reaction apart from simple suspense of judgement.

What needs to be emphasized is that logic is at least partly a normative enterprise and seeks to impose a certain order on our possibly recalcitrant personal predilections. It is, moreover, possible to conceive even rules such as modus ponens primarily as rules of dialectical procedure. There is, we might say, no contradiction in a man's thinking P to be true, Q to be false and P to imply Q, and this is our occasional fortunate or unfortunate condition; but he must not say that this is the case. He may say it, at least, only with the prefix 'I think'.

These points apply generally in the discussion of any kind of fallacy, but it should be clear that they arise with special force when we consider the Fallacies Dependent on Language, of which we may continue to take Equivocation as typical. It is by now more than ever clear that there are many cases in which the decision to regard a word or phrase as equivocal is come to as one among several possible escape-routes from a threatening contradiction. That this may sometimes be the only reason for the decision is a sceptical suspicion which makes us look, now, for firm ground for our feet. There is, after all, no doubting the procedural importance of charges of equivocation in forcing the clarification, and perhaps creation, of shades of meaning of the words accused in them.

For how do words change their meaning? What dialectical phenomena accompany such change? There is no contradiction in supposing a new use of a word to become current overnight, by magical universal agreement; but we may be sure that this is not the usual case. Again, the history of language reveals a slow drift of meanings imperceptible in the short term; but this kind of change can be of interest only to someone who wishes, as it were, to converse across the centuries. Of far more significance is the inventive use of words: a model catches the attention, its properties are verbalized and, without apology, we have a new use thrust on us. The model may be useful or misleading, permanent or fleeting. If it is a bad model, we shall be in our rights in regarding arguments relying on it as equivocal; but if it is not, to do so would be at best academic and at worst obstructionist.

A charge of equivocation may be thrown back -- 'You are using W in one premiss in sense W and in one premise in sense J and in the other in sense K'. 'No, I am using it in both cases in sense L'. The extent to which one may demand that a word be used in a given sense is variable: we do not, dialectically speaking, have complete 'freedom of stipulation' in the sense in which this sometimes asserted but the justification is, again, pragmatic and no one can make decisions that must be binding on others. 'You are misusing the word W' is a possible form of point of order.

Our thesis might be summed up by saying that equivocation is a procedural, non-topical concept. The question of whether a given term is or is not (subtly) equivocal in a given context or family of contexts is not one that may be decided by looking at the term itself, but rather a question of how one may best order the discourses within which the word appears. We can understand the word 'equivocal' only by seeing that it is not a descriptive term, but rather one that is used in the making of certain kinds of procedural points. 'That is equivocal' does not describe a locution, but objects to it.

A further argument for the non-topicality of equivocation rests on the existence of dialectical paradoxes resembling, though not precisely parallel with, the paradox of the Liar. The simplest of these, in fact, is a dialectical version of the Liar-paradox itself, and provides an argument for the non-topicality of the concept of truth. The man who says of any of his own, current, utterances that it is false presents us with a curious problem; for a necessary condition of a lie is that it should be a prima facie deception but, once a statement is stated to be a deception, it is so no longer. What is openly stated to be false is not, in the large view, a falsehood. 'S -- no, that's false' may be comprehensible as a change of mind, equivalent to 'S -- no, not-S', but 'S and that's false' is a species of nonsense.

A dialectical paradox is generated by a kind of self-reference; not, however, a reference of a given statement to itself, but a reference by a speaker to one of his own current utterances. (A 'current' utterance is an utterance to which he remains committed.) We could characterize the enterprise of the dialectical fallacy-monger by borrowing some terminology from the early Wittgenstein, and saying that he tries to say what cannot be said but can only be shown.⁶ Wittgenstein, after all, indulged in this kind of paradox himself when he wrote at the end of his book (6.54) that the other statements in it were all nonsense. It is possible for me to say S and to know, and to show by some such means as uncertainty of manner, that it is false; but it is not possible to commit oneself both to S and to S's being false.

This being so, something equivalent will apply to 'S and that's true'. The phrase 'and that's true' cannot say more than S says, except in the sense that it may add emphasis, or increase conviction, or otherwise achieve something that could have been shown in other ways, such as by saying 'S' very loudly. Moreover, the second-person locutions 'What you say is false', 'What you say is true' take over some part of the paradoxicality from their first-person equivalents since, considered as topical locutions, they would have to be capable of commanding agreement and disagreement, and this would involve the hearer in committing himself to paradox in the first person. Reference by any participant to another participant's current locutions may also engender paradox. 'What you say is false', and 'What you say is true', may represent elliptical disagreement or agreement with the hearer's locutions, or they may be, or be part of, point-of-order locutions; but they cannot make topical points that essentially involve the concepts 'true' or 'false'.

That 'valid' and 'invalid' are ultimately of the same character could be argued at length, but it would be necessary first to strip from them the meanings they tend to assume within formal systems, in which they indicate topically-discussable conformity with this or that formal norm. I shall confine myself to arguing the case of 'equivocal', which has no clear formal meaning provided we exclude its grosser manifestations. What could 'I am equivocating' mean? It is of the essence of an argument, as we were at pains to insist earlier, that it be put forward in support of its conclusion: otherwise it is merely 'hypothetical'. But someone who argues, say, from premisses P to conclusion Q, and then adds that his argument is equivocal, has implicitly negated the seriousness of his purpose in supporting Q, and no longer really argues. Hence he does not even really equivocate. 'P, therefore Q, and that's equivocal' is a piece of nonsense of the same order as 'S and that's false'. The locution 'That was an equivocation' may, of course, be used to unsay an argument but it cannot be used by the arguer to say something additional. He can show that he is equivocating, as those who equivocate commonly do, by his hesitancy or, more probably, his dogmatism; but he cannot say it. In the same way, 'I am not equivocating' is an empty assertion except for possible subsidiary functions such as adding emphasis or rejecting an objection, and 'You are equivocating', though very much to the point as a point of order, cannot be topical either.

The distinction between topical and procedural locutions can be seen to fulfil the same role for us, in the solution of paradoxes, as theories of levels of language have done for others in connection with paradoxes of a more traditional nature. A dialectical system cannot unrestrictedly admit the predicates 'true' and 'equivocal', together with means of reference to a speaker's or hearer's current locutions, into its topical object-language.

One answer to all of this might be that the logician is never a participant in the dialogues whose locutions he studies, but always an onlooker; and that he may, consequently, make whatever comments he wishes concerning the arguments that participants put forward, without becoming involved in paradox. He can say 'X was equivocating when he said so-and-so'. He can do this impartially just as, when we stand aside from an argument between X and Y, we can say 'X was really right: what Y said was false'. Do logicians, and scientists, speak only in the third person? This is an intellectual fancy that is both comforting and delusive in its implications. It is capable of turning into the claim to be above criticism, like the armchair strategist whose failure to win battles is due to the lack of co-operation of the enemy.

Large-scale and would-be-monolithic enterprises like Logic and the other branches of learning do not superannuate Dialectic or escape its processes, for the onlooker is in an arena of his own. That the interchange may take place in book-length or article-length locutions, with a large and ill-knit group of participants, makes a great deal of difference, no doubt, to the rules; but the rules must have, in common with those that apply to dialogues more properly so-called, the feature that they apply within a variable language. No one will understand equivocation who thinks that words permanently retain their meanings; or, particularly, who thinks that it is his task to make them do so.

The road to an understanding of equivocation, then, is the understanding of charges of equivocation. For this, the development of a theory of charges, objections or points of order is a first essential.

1 The best examples of dialectical analysis are in the 'Brown Book': Wittgenstein, Preliminary Studies for the 'Philosophical Investigations'.

2 Carnap, Introduction to Semantics, pp. 8 ff.

3 The letters 'S' and 'X' may represent terms, whole statements or other linguistic entities; the distinction is not important to the present discussion and I shall make varying assumptions in different examples.

4 Quine 'Two Dogmas of Empiricism'

5 Jurisprudence recognizes three kinds of presumptions, of which these resemble the juris sed non de jure, presumptions 'at law but not with legal force'. See Kenny's Outlines of Criminal Law, § 489.

6 Wittgenstein, Tractatus Logico-Philosophicus, 4.121 ff.