Sentence-sense

Frege’s notion of Sinn is almost always debated in the straitened context of his essay Uber Sinn und Bedeutung, and when it is debated in that way the discussion almost always terminates in the half-hearted acceptance or inadequately founded rejection of the idea that different names and referring expressions may present one and the same object by means of different manners of presentation. The idea of Sinn is eked out by various metaphors about telescopes and so on, and by an intuitive feeling that something like Frege’s doctrine must surely be correct for different definite descriptions of one thing - if only because such definite descriptions may plainly differ in meaning. But if discussion and interpretation of Sinn are confined within the context of Uber Sinn und Bedeutung, then it is only after proper names and definite descriptions have set the scene that much notice is taken of the fact that Frege assigns sense and reference to sentences and predicates as well as to proper names. As I shall shortly show, there is an interpretation of the theory which even puts us in a position to assign a Sinn to morphemes. If we take things in the order in which the interpretation I am caricaturing takes them, however, then the fact that Frege allows both Sinn and Bedeutung to expressions such as sentences has to be accommodated by supposing that Frege tried to explain sentence-sense from the primitive idea of naming, mentioning or designating ordinary objects such as particular planets, men, mountains, etc. Frege’s whole and complete explanation of sentence-meaning is then supposed to have been built on the idea (which is Frege’s but does not, I believe, play this explanatory role) that the proposition is to an object called its truth-value, The True or The False, as the sense of the words ‘planet seen as the brightest body in the evening sky’ is to the planet Venus. It is then complained - and with every appearance of justice - that the analogy is quite inadequate to explain sentence sense or elucidate its alleged equivalence ‘'with the proposition.

The dissatisfaction some have felt here is unsurprising, because neither the mode of presentation idea, nor any of the metaphors of Uber Sinn und Bedeutung, are robust enough to sustain the weight which such an analogical extension from the idea of reference would necessarily shift onto them. Designation and reference cannot both undertake the whole explanation of what it is to say something and ab initio explain saying as referring to The True or The False. There are difficulties in any case in Frege’s doctrine that sentences are names of truth-values,¹ but it makes a great deal of difference whether we are simply asked to put up with it as a consequence of a theory of meaning already clear and already explanatory, or whether we are asked to accept it from the very outset as something upon which further explanation can be securely based - which is what would happen on the interpretation I am attacking. To treat it as the second sort of doctrine legitimates awkward questions. If the target is The True, it might be complained, and if different senses or thoughts are just different avenues to The True, then why not choose the easiest way of hitting the target and always say snow is white, or the cat is on the mat, or anything at all that is as a matter of fact true? Is there one message always and an indefinite number of media? Or if there is more than one message, is the message the medium itself? The interpretation leaves Frege’s doctrine without resources to explain the evident absurdity of these suggestions. (And if one is to take fully seriously the idea that one can refer to The False while aiming at The True, there will never apparently be room for anything but a rigidly extensional theory of designation and reference. What then of sense without reference?)

It will be clear that the point of Frege’s theory has got lost somewhere, if one reflects that this whole line of interpretation, which starts from the ‘ mode of presentation’ notion of sense, seems to be incompatible with Frege’s insistence (in the Grundlagen and elsewhere) that reference itself is unintelligible outside the context of a complete sentence or thought. Frege took care to imbed the theory of sense and reference of proper names in a larger context, and it is in that context that we must seek it out. So seen, and in its proper universality, the theory is first and foremost a general theory of language, and its foundation and basis is not naming at all but the notion of sentence-sense itself. The interpretation I shall commend, which turns the other interpretation upside down and makes sentence-sense, not reference, the point of leverage, explains the production and understanding of familiar and unfamiliar utterances by an account of how the constituents of sentences can systematically contribute to the meaning of the complete sentences within whose structure they figure. And if sentence-sense is where the theory begins, no analogy has to bear the weight of explaining the meaning of sentences. A fortiori no problematic passage from designation of planets to designation of truth-values has to explain saying, saying being where the theory really started. But with a going theory of saying and sentence-sense and an account of how words contribute to it, an analogy (which is only an analogy) does then become possible between the way in which an arithmetical function ( )² determines value 4 for an argument 2 and the way in which a predicate ( ) is wise determines the truth-value True for argument Socrates. But the explanatory dividend of this analogy primarily concerns predicates not sentences. It explains them as a species of sentence-functor. Knowing their meaning is just a matter of knowing what they do within complete sentences possessed of this or that complete meaning.² As for the senses of referring expressions, these are simply a special case of senses in general. It is often supposed that Frege devised the theory of sense in response to the special problem of informative identity statements. This may or may not be a historical fact. The value of the general theory of sense manifestly transcends the grave difficulties with which e.g. genuine proper names confront his solution to that special problem.³

What then is sentence-sense ? In the Grundgesetze der Arithmetik 1.32 Frege wrote:

But to all names properly formed from our primitive signs there belongs not only a reference (Bedeutung) but also a sense (Sinn). Every such name of a truth-value ex¬ presses a sense, a thought (Gedanke). It is determined by what we have laid down under what conditions every such name designates The True. The sense of this name, the thought, is the sense or thought that these conditions are fulfilled. A sentence of my concept¬ writing or symbolism consists then of the assertion sign and a name. . .of a truth-value . . .By such a sentence it is affirmed that this name designates the True. Since it at the same time expresses a thought, we have in every well-formed sentence of the symbolism a judgement that a thought is true. So there simply cannot fail to be a thought [associated with such a sentence]. . .

The simple or composite names of which the name of a truth-value consists con¬ tribute to the expression of the thought. This contribution (Beitrag) of each is its sense. If a name is a part of the name of a truth-value, then the sense of the former name is a part of the thought which the latter [name of a truth-value] expresses.

Wittgenstein had no difficulty in rescuing the sentences I have italicized from the technicalities of the context (Tractatus Logico-Philosophicus 4.021-4.024, cp. Philosophische Bernerkungen iv, 43) and nor should we. The suggestion is this (cp. Dummett, P.A.S. 1958-9). If we will simply take the notion of ‘ true ’ as clear enough for the purpose - not for all purposes, but for this one - then we can say that, for arbitrary sentence s, to know the meaning of s is to know under what conditions the sentence 5 would count as true. Unlike theories constructed in terms of belief and intention after the fashion of H. P. Grice’s ‘Meaning’ (in this volume), or theories which place trusting confidence in a notion of use which they have still to delimit in a way which does not reimport all the problems, this theory of meaning offers us not a way into the circle of semantic terms but a connexion between two of them. This may be the best theory we shall have for some time, and it is certainly some sort of basis from which to speculate about word-sense and difference of word-sense. But before I can get to the philosophical theory of dictionaries, there are some problems to clear away.

(α) The intensionality of the notion of meaning creates difficulties for the doctrine.

(ß) Problems arise from the association of the definition (since the time of Frege) with various positivisitic doctrines, from the limitations of the definition, and from its concentration on the indicative mood.

(y) There is a problem in Frege’s definition arising from its apparent confusion of sentences with statements.

(δ) There is need for a rather careful statement of the connexion between truth-conditional definitions for sentences and meaning-definitions for the words that are constituents of sentences.

(α) The notion of meaning is an intensional one, whereas the notion of truth is not. How then can we catch the meaning of means by defining it in terms of true ? Frege’s proposal seems to be that s means that p if and only if (s is true if and only if p). But if we fix the meaning of a sentence s by simply saying sentence s is true if and only if snow is white then, pending an account of the if which does not give us back the whole problem (and others too), nothing can prevent a critic from supplanting snow is white in such a context by another sentence with the same truth-value - e.g. King Charles was beheaded. What results will still be true. Matters are quite different with sentence s means that snow is white. If that is what s means it certainly does not follow that it means that King Charles was beheaded. That is not at all what we had in mind for the truth-condition. It is notwhat one might call the designated condition.

It might go some way to meet this objection to amend the suggested schema by adding a ‘necessarily’ and stating the revised theory like this:

s means that p if only if (necessarily (s is true if and only if p))

Leaving aside some perhaps surmountable difficulties about mention and use and the presumed universal contingency of the fact that any sentence (as typographically identified) is assigned the sense which it is assigned, this suggestion and any emendation of it would still have the important disadvantage of leaving all sentences expressing logically or mathematically necessary truths with the same meaning. This would have to follow unless subsidiary doctrines were somehow imported to distinguish these senses from one another by reference to the syntactic structure of the sentences and the word-senses of these sentences’ constituents. (The last could be fixed via contingent sentences, for which the difficulty does not arise.) As it stands, however, the proposal certainly does not solve the original problem.

These reflections about procedures for mating sentences with the conditions under which they are true, and the reliance of such procedures on the contribution of constituents, do however serve as a reminder that the difficulty is less a practical difficulty about particular meaning definitions than a theoretical difficulty about meaning as such. We understand fewer things than we need to understand about either the theoretical constraints on, or the canonical form for, the truth-conditions which our defining procedure will mate with the English sentences whose sense they explain. But it is evident that, since there are an infinite number of English sentences to be assigned truth-conditions, such conditions can only be produced by a method of systematic decomposition of the given English sentences into the basic structures and components from which they were built up in the first place. (Such a decomposition must exist. No speaker learns to produce or understand the infinite number of sentences he can produce or understand by learning their senses one by one. And for the same reason only a finite number of basic structures and components can be available within any language.)⁴ Now in practice this requirement, which is a substantial and non-trivial one, combines with the minimal requirement of material equivalence to make it virtually inconceivable that any satisfactory set of recursive procedures for assigning truth-conditions could serve up an irrelevant or non-designated condition for a sentence.

It is of paramount importance to claim neither too little nor too much importance for this. On the one hand, it seems to me to leave the theoretical problem of meaning and the intentionality of means com¬ pletely unaffected. We really are almost as far as ever from a definition or analysis of means itself.⁵ On the other hand the difficulty does nothing to obscure the insight that assigning or discovering a sense for s at least involves assigning to it or dis¬ covering for it some correct truth-condition or other. I conclude that we have made one piece of progress, both as concerns what is theoretically involved in giving truth-conditions and as concerns the general theory of making particular worddefinitions. However negligible our progress with the special case of an analysis and dictionary definition of the word means, we can oppose to the extreme vagueness of the requirement that the truth-condition for a sentence be a designated condition the full austerity of another requirement. The truth-condition must have been produced by the operation of a systematic, general, and uniform procedure competent to analyze any sentence in the language into semantic components drawn from a finite list of such components (i.e. a vocabulary or dictionary). And the procedure must account for the semantic structure of the sentence by showing how it could be generated by a finite number of semantically interpreted modifications or steps from one of the finite number of semantically basic sentence-forms of the language. (Whether, as I should continue to hope, the generative syntax of the school of N. Chomsky could be transposed or transformed into such a ‘generative semantics’ I cannot judge).

I come now to (ß). As he abandoned some of the objectives of the Tractatus Logico-Philosophicus Wittgenstein gave a more and more operationalist or positivistic slant to the insight which he had taken from Frege’s Grundgesetze (v. Philosophische Bemerkungen, passim). And the process which he began was completed by the Vienna Circle and A. J. Ayer in their doctrine The meaning of a sentence is the method of its verification. This formula had tendentious uses and perhaps for that reason it is now insufficiently recognized that it did represent a piece of thought, which was just as serious as it was crude maybe and unsatisfactory, about what it is to know the sense of a sentence which one uses or to know what one is saying.⁶ But, genuine though this problem is, Frege’s truth-conditions theory itself is neutral between, and has a value quite apart from, particular answers to it, and what concerns me are other questions.

I have objected to the truth-conditions theory (see (α)) that it is in danger of letting in too much, but the standard objection to it is much more likely to be that it lets in too little. ‘ Even in the case where the indicative reigns, the truth-condition account leaves out almost everything of some sentences’ communicative significance. And for other sentences, the theory is based on a hopelessly special case and can never come to terms with the complexity and multiplicity of other grammatical moods and functions than the indicative.’

This charge may be expected to come with accusations of committing what Austin called the descriptive fallacy, but caution would be advisable in invoking his authority at this point. For so put the charge is dubiously consistent with something else which Austin also insisted upon, at a slightly later stage of his thought, - his distinction between the meaning of a sentence and its force. To draw this distinction he relied upon the notion of a locution (which he opposed to an illocution and a perlocution); and locution he glossed in the Fregean terms of sense and reference. If Austin knew anything about Frege, whose Grundlagen he had after all translated, he must have been aware that at least for indicative sentences (Austin might have said ‘ at least for straightforwardly indicative sentences ’) the meaning-force distinction, if glossed in this way, would tend to limit their proper meaning to all and only that which bears upon their truth-conditions. And Austin was surely aware that Frege himself had laid great emphasis upon precisely this kind of distinction when he insisted that logic could not concern itself with the coloring or illumination of language, which was something Frege considered quite irrelevant to sense, or with anything at all that did not bear directly upon truth.⁷ Grice has looked for similar distinctions (of greater generality than Frege’s)⁸ and he distinguishes between the implications of a sentence s, what follows from it if it is true, and the implicatures of it which are distinct from its implications - the situational import of a speaker s saying s. (This import typically arises not from the meaning of s but from the standard working of the perfectly general conventions of helpfulness, sincerity, etc., which Grice argues govern the conduct of speech-exchanges.)

There is a clear point in attempting to make such distinctions as these. The total communicative content of an utterance, construing ‘content’ comprehensively, is something too complex to be accounted for by any one-level theory. An utterance or speech-act considered as a whole may be horizontally complex, e.g. in respect of the syntactic structure of the sentence used in the making of it, and it may also be vertically complex. In saying that p a man may warn someone that q. In warning him that q he may threaten that r, and in threatening that r he may insinuate, what is even worse perhaps for the listener, that s. And in doing all this he may bring off what Austin called perlocutionary effects. He may alter the listener’s behavior in some way he precisely intended to alter it. If we persist in lumping all these things together in an undifferentiated notion of meaning it seems hopeless to look for a systematic theory to account for such ‘ meaning ’. We must rather unpack the speech-act layer by layer. And at the first layer we must, I think, try to isolate all and only what is strictly said before we can explain how circumstances, conventions, and whatever else add implicatures, forces, or illocutions; and how these in their turn secure perlocutionary effects.

The one element with a claim to be really fundamental or central to linguistic communication is this element consisting of what is strictly said. It is this which we must begin by refining, and the suggestion implicit in Frege’s theory of sense, and in Grice’s and Austin’s theories, is that there relates to the strict meaning of a sentence s all and only that which bears logically upon the truth of s. From which we recover an intelligible and satisfactory-sounding doctrine -that truth must be the central notion of semantics, and that the boundary between what does and what does not bear logically on the truth of what is strictly said must be the boundary between the science of semantics and the science of the further effects obtaining in a speech-exchange.⁹ There is a precedent for calling the latter pragmatics, but obviously this name is still little more than a catch-all for what does not bear on strict meaning.

Even if this defence of Frege’s theory of sense were accepted by the objector he might still claim that an enormous number of apparently indicative utterances were not really indicative in meaning - insisting still on the heinousness of what Austin stigmatized as the descriptive fallacy - and he might claim that an even larger body of utterances were not even apparently indicative. A full reply to this would take the form of (i) a thorough scrutiny of the class of so-called performative utterances,¹⁰ and (ii) an appraisal of the claims (which still seem to me to be passably good) of the indicative mood’s claims to priority over other moods.¹¹

However many concessions the objector succeeded in wresting from a defender of the Fregean doctrine it is difficult to believe that they could imperil the following minimal contention - That any satisfactory theory of meaning (whether or not exempt from the allegedly crippling defects of the descriptive fallacy) must entail the following proposition:

To know the sense of an indicative sentence s it is necessary to know some condition p which is true if and only if s is true and which is the designated condition for s.

But from this somewhat exiguous condition of adequacy I think it is possible to squeeze enough to get started on word-sense. Not even one who believes there is a descriptive fallacy waiting to be committed (and certainly then committed by this writer) will want to suppose that sentence-constituents, i.e. words, generally have different senses according to whether they occur in indicative or ‘ performative ’ or imperative or optative sentences.

But two more difficulties remain, (y) and (δ) , before we can exploit the power of this Fregean doctrine.

(y) As it stands in Frege and at many points in my exposition, the theory seems to confuse sentence with statement or proposition. If the s which figures in the theory that to know the meaning of s is to know under what conditions s is true is a statement then it is the right sort of thing to be true or false, but we have not yet been given a route back to sentences and their constituent words. If s is a sentence then there is no problem of this kind, but it may be said that a sentence by itself is not true or false. An amended theory would then have to rule that, for arbitrary sentence s, to know the meaning of sentence s is to know the truth-conditions of the statement which s makes, which apparently makes the theory presuppose a whole prior account of stating. But if we have that, it may be said, then all the hard work is over already. Nor can the dilemma be laughed off by those who say they cannot countenance propositions unless ‘proposition’ is a gratuitous synonym for sentence. If truth is accounted a predicate of sentences then, if context-dependent utterance is to be explained, it has to be defined for sentence-tokens. The dilemma is then that a theory of meaning needs to be or include a theory of sentence-type meaning.

It looks like the beginning of an answer to state the theory thus:

To know the meaning of a sentence-type s is to know under what conditions the utterance of a token t^s of the type s is factually licensed [token-true],

s confers an intial purport on t^s which it enjoys in abstraction from all the circumstances of its utterance, but in working out its meaning there is also the context in which it is lodged to take into account. Only this can determine what the notion of factual licence or token-truth necessarily imports, e.g. the references of such referring devices as demonstratives. To understand s, then, is, amongst other things, precisely to understand how such extralinguistic factors are to determine the truth conditions for t^s. (They must of course be all and only the extralinguistic factors which bear on any truth-condition for t^s.) The speaker knows the truth-conditions for t^s, and he identifies a particular statement, only if he knows how to work out the contribution of, for example, these demonstrative or identificatory factors. So the truth-conditions theory does not provide any effective retrieval of all and only the semantic purport of t^s until the nature of these contextual inputs is fully described and we are in a position to add to what we already have such additional theses as this:

To know the meaning of a sentence type s is (a) to know how in principle to work out the demonstrative purport of its indexical expressions, in context, and (b) to know under what conditions the utterance of a token t^s with contextual determination d is factually licensed [token-true].

I recognize that to make this retrieval effective we should need actually to possess a complete theory of demonstrative reference. And for a completed theory we should have to specify exactly what contextual determination amounted to. I expect that this would be tedious and at some points very difficult but I believe that we have enough feeling for what it would be like to make it unnecessary to say more here.¹²

The following diagram may serve to summarize some of the conclusions and the answer to (ß).

(1) Sentence type s with assigned and determinate grammatical structure and generic sense S.

(2) Tokens t^s₁, t^s₂. . .produced in particular speech episodes E(t^s₁), E(t^s₂). . .

(3) Demonstrative (and any other) inputs to (4), these being determined by the context of utterance-episode E(t^s₁). The demonstrative purport of any referring phrases in t^s₁ (plus any other semantically relevant purport as yet unaccounted for).

(4) The statement made or proposition propounded by the speaker in E(t^s₁) - what he says, this being determined by what has to be the case for the speaker to count as saying truly (i.e. saying something true).

(5) Truth-value of (4).

(6) Situational factors bearing on (7).

(7) What the speaker means in or by saying (4).

(δ) It has been maintained by Ryle and others that words have sense in an only derivative manner, that they are abstractions rather than extractions from sentence-sense. There is something we must acknowledge and something we must reject in this doctrine. What we must concede is that when we specify the contribution of words we specify what they contribute as verbs or predicates or names or whatever, i.e. as sentence-parts, to a whole sentence-sense. Neither their status as this or that part of speech nor the very idea of words having a sense can exist in isolation from the possibility of words’ occurrence in sentences. But this is not yet to accept that words do not have sense as it were autonomously. And they must. If our entire understanding of word-sense were derived by abstraction from the senses of sentences and if (as is obviously the case) we could only get to know a finite number of sentence senses directly, there would be an infinite number of different ways of extrapolating to the sense of sentences whose meanings we have to work out. But we do in fact have an agreed way of working them out. This is because word-senses are autonomous items, for which we can write dictionary entries. Quine has put the point in dispute so elegantly and concisely that it is enough to quote him:

The unit of communication is the sentence and not the word. This point of semantical theory was long obscured by the undeniable primacy, in one respect, of words. Sentences being limitless in number and words limited, we necessarily understand most sentences by construction from antecedently familiar words. Actually there is no conflict here. We can allow the sentences a full monopoly of ‘ meaning ’ in some sense, without denying that the meaning must be worked out. Then we can say that knowing words is knowing how to work out the meanings of sentences containing them. Dictionary definitions are mere clauses in a recursive definition of the meanings of sentences.¹³

It is easy to see that by the account towards which we are working, the sense of a word being the precise contribution (Beitrag) which its presence makes to the true utterance conditions of the complete sentences in which it figures, a word will be ambiguous if and only if the dictionary which defines it requires more than one entry for the word. It will require this in order that the procedure of recursive definition (of which the dictionary is one part and the syntax of the language is another) should account correctly for the meaning or meanings of the sentences in which it occurs. Sometimes we can see straight off that a word has more than one definition (I say some more below about certain hazards associated with ‘just seeing’), but the view I am advancing or commending makes much more hang on our being able to recognize that a sentence may be read or heard in one way so as to be true and read or heard in another way so as to be false. If the syntax of the language can find only one grammatical analysis of the sentence then we must try to account for the different readings by postulating a word ambiguity and try to write a multiple dictionary entry in order to account for the distinct truth-conditions for the sentence.¹⁴

¹ Some would maintain that it is true and profound that a rose is a rose is a rose. The vexatious or linguistically intolerant will maintain that this is strictly speaking an empty, meaningless, and ungrammatical assertion. Frege’s theory is alone in insisting that the assertion is clear, grammatical - and false. For the truth-value True (= a rose is a rose) is not a rose. P. T. Geach’s example. Even more serious is the fact that the equivalence relation = needed to abstract truth-values as objects presupposes sentences already possessed of the established properties of.meaning and truth or falsehood.

² Frege does in fact allow predicates or concept-words a reference as well as a sense, but one who upholds this feature of Frege’s theory of language, imported only by the need to quantify over properties, need not maintain that coming to understand a predicate is best or most illuminatingly described as a matter of coming to be able to identify its reference. It is a matter of coming to understand what sentences with what sense the predicate is a functor in. I dwell on this case because the word-class whose senses I shall principally be concerned with is precisely this class of sentence-functors.

³ The difficulties are especially evident in the case of true identities of the form (proper name n) + ‘ = ’+ (proper name m), where m and n have to have different senses for Frege’s solution to work. Unfortunately this seems to be impossible without collapsing m and n into definite descriptions, and I myself should now want to see the problem as a special case of the paradox of analysis. It is simply that paradox applied to names with identical senses, and the problem it raises exactly parallels the problem of the supposed difference in sense of the two sentences oculist = df eye doctor and oculist = oculist. I welcome the opportunity to disown my ‘ Identity-Statements’ in Analytical Philosophy (2nd series, edited by R. J. Butler), which (amongst a hundred other faults) failed to look hard enough at the possibility of supposing (a) that ‘ if anything is Hesperus then Hesperus = Phosphorus ’ is an informative necessary statement, empirically discovered, and (b) that the paradox of analysis has and was bound to have an analogue for proper names. I now think that all the peculiarities which arise under (a) and (b) can be accepted and that they are indeed to be deduced from any correct theory of reference by proper name. (If contingent information is required to fix the sense of a proper name and this is the mark of reference to particulars - cp. Strawson, Individuals, Part 11 - then we must not be too surprised to find senses giving rise to an empirically discovered logical necessity.)

⁴ See Donald Davidson, ‘ Theories of Meaning and Learnable Languages’ in Proceedings of the 1964 International Congressfor Logic, Methodology, and Philosophy of Science (North Holland, Amsterdam, 1965, ed. Bar Hillel). Had I seen and taken proper note of this and his seminal ‘Truth and Meaning’ (in Synthese, 1967) much of the present paper could have been differently and more elegantly put by means of Davidson’s illuminating allusion to Tarski’s truth definition. A certain complementarity and a quantum of historical background information do however result from leaving my contentions in what is substantially their original form.

⁵ And so far as this problem goes Brian Loar has persuaded me that the difficulties of keep¬ ing out non-designated conditions are intimately connected, in ways which I shall leave it to him to expound in due course, with the answer to the difficulties of amending and completing Grice’s 1957 definition (to leave room for the fact that people sometimes say what they don’t mean etc.). Proponents of a truth-definition of meaning have not found a way to render them¬ selves simply exempt from all of the problems which arise here.

⁶ I surely cannot say or explain what All mimsy were the borogroves means by saying that this sentence will be true if and only if everything satisfies the open sentence if X is a borogrove then X is mimsy. And it is certainly a part of what would still be lacking in this explanation that it gives no idea at all of what investigations with what outcome would count for or against the assertion. This is not however to say that discovering or fixing any sense for the sentence or its predicates would be a matter of finding or fixing anything like the method of verification or falsification - let alone the method of conclusive verification or falsification - for the sentence or for the satisfaction of one of these predicates.

⁷ Frege’s unconcern with the illumination (Erleuchtung) or coloring of language is so total and he is so uncurious about what does not straightforwardly pertain to truth he prevents himself from examining what lies on the other side of his distinction or drawing it in terms which match the generality of Austin’s or Grice’s. But I should say that his view of the sense side is for the most part congruent with theirs. There are complications Frege would have to take account of, however.

⁸ In ‘The Causal Theory of Perception’, P.A.S.S. (1962) and in subsequent work as yet, unfortunately, unpublished.

⁹ For some more of what needs to be taken account of here see the discussion of the definition of snub.

It seems much more important to settle a hierarchy or ordering of the considerations which need to be taken into account in working out the total communicative content than to establish a firm frontier here. For example if the minimal connectedness of p with q were somehow shown to be a universal condition of uttering if p then q (where p and q are in the subjunctive) - if so much were shown - then it might come to seem quite unimportant whether this was due to the meaning of subjunctive if (semantics) or due to the operation of a convention for making a speech act (pragmatics) which was too universal or too deeply entrenched to admit of waiver.

¹⁰ There are two aspects of this problem. Some still hold, it seems, that it is a part of the meaning of ‘ There is a bull in this field ’, said (under suitable circumstances) as a warning, that the utterance is a warning and equivalent to the* ‘ explicit form’ (so-called) ‘ I warn you there is a bull in this field’. Austin himself would have preferred to hold that warning was only a part of the illocutionary force and not any part of the meaning of ‘There is a bull in this field’. But he would certainly have held that its having this force was determined by convention. Both Austin’s convention view and the other view of ‘There is a bull in this field’, like any alleged equivalence between ‘explicit’ and ‘non-explicit’ forms, seem to be undermined by the obvious reflection that what makes ‘There is a bull in this field’ into a warning is a starkly extra-linguistic matter which a hearer would have to know already, and have to combine with a pre-existing understanding of what is already said by ‘There is a bull in this field’, in order to understand it as a warning. I should make a similar claim about ‘ I will be there’ construed or misconstrued by the hearer as an illocutionary act of promising. This cannot be equivalent to ‘I promise to be there’.

The other aspect of the problem of performatives is the status of the ‘explicit’ forms, e.g. ‘ I warn you there is a bull in this field’ and ‘ I promise to be there’. I persist in taking these as straightforward statements.

The least unimpressive objection to construing ‘ I promise. . . ’ as a statement is that it could hardly be tantamount to the self-descriptive running commentary ‘I am promising’. But ‘ I love you’ cannot be converted into ‘ I am loving you’ either. What this shows is something about what sort of verb ‘love’ is. It doesn’t show one can state or declare that one loves someone. The performative theorist may retort that what he is arguing about are cases like promising something where something does go on and a man does do an action, and what the man says doesn’t appear to describe that action. It makes all the difference, he will say, that the continuous present ‘ I am promising’ is available, and that it still does not give an equivalent of ‘ I promise’.

I think that the reply to this is that ‘ I promise’ does describe an act but describes it in a way not equivalent to the continuous form ‘ I am promising’. It describes the act in the grammatically perfective aspect instead of the grammatically imperfective aspect. It is the completable act, not the activity of completing it, which is of interest when a man avails himself of the ‘ I promise’ formula, and it is this which makes the instantaneous or aoristic form ‘ I promise’ appropriate and the continuous form inappropriate.

¹¹ To understand, as opposed to merely reacting satisfactorily to, the command ‘shut the door’ I must know what it would be for such a command actually to have been executed, i.e. what it is for it to be true that the door has been shut. (Perhaps I must also know what it is for the door to have been shut by the person commanded because of the command to shut the door.) But to understand what it is for it to be true that the door is or will be shut I need not understand anything about commands or the imperative mood.

If this be thought an important asymmetry it may incline us to allow both the indicative and the notion of saying that a kind of priority. And it may also incline us to suspicion of the whole idea of a propositional content (as it figures in Frege’s and subsequent theories of the assertion sign,'Ⱶ', conceived by Frege as an assertive vertical ‘ | ’ prefixed to a horizontal stroke ‘—’ which heralds a content) which is neutral between and embeddable as a common element in different acts of saying, commanding, questioning, wishing. At very best this common content can only be an artificiality. For if there is no strict parallellism between the indicative and the imperative, optative, and interrogative moods, then the asymmetry thesis combine with the other arguments against 'Ⱶ' to suggest that the common propositional content Гthat pꞀ must really be got by subtraction from the assertion of ГpꞀ , rather than the latter by addition of 'Ⱶ' to ГpꞀ . One might maintain that ГpꞀ automatically says that p unless you obstruct it from doing so. (One way of so obstructing it is to embed it?)

¹² It might be objected to the words factually licensed which figure in the revised theory that this technical locution can only be elucidated in terms of a token’s being used to make a true proposition or statement, which reimports both stating and statement. But I am encouraged to think that a reply could be found to this. It must surely be possible to teach someone a convention of doing something or not doing something (or permitting someone else to do something or not permitting them) according to whether or not a certain condition C obtains. The notion of convention itself is a wider notion than that of saying or stating, and is certainly not automatically or for all purposes ruled out from employment in the elucidation of saying or stating. Now if the idea of convention is carefully handled I believe we can elucidate and designate a class of performances within the convention of saying as in a special way acceptable performances of saying. This will be the class of performances of saying truly. (We need to ascribe structure to saying truly but, so far, the theory need not ascribe more structure than verb + adverb. It is only at a later stage, and for different and slightly more dubious purposes, quantification over propositions, that we need to ascribe to it the structure verb + object.) I say that the idea of convention has to be carefully handled because the notion of true saying will only be effectively fixed here if advantage be taken of the possibility of giving the elucidation in the presence of shared correct belief about whether or not C. Otherwise there will be no effective distinction between the false statement that C obtains made by one who understands what he is saying and a man’s total failure to understand the linguistic purport of the token whose sense is being fixed.

It seems obvious that any thorough discussion would have to touch on or work its way back to Grice’s project. See answer to (α) above.

The effect of the tentative suggestions I have made to answer problem (y) is to distinguish the sense of a sentence (something produced in utterance as a token of its type) from its output, the proposition which the token expresses. Since Frege himself identifies Sinn and Gedanke, modifications become necessary to his theory of indirect Sinn and oratio obliqua when Sinn and Gedanke are distinguished. It looks as if oblique occurrences of an expression E will have to stand for the output (the proposition or proposition factor) not the input, or strict Sinn. But these are complications I shall not enter into at this stage.

¹³ ‘Russell’s Ontological Development’, p. 306 in Bertrand Russell, Philosopher of the Century, ed. R. Schoenman (London, 1967).

¹⁴ Once there are two dictionary entries both classifying some single word as the same part of speech, this may create a much larger number of theoretical alternative readings of some sentences than it will normally occur to us to suppose they have. For some of the theoretically possible readings will be too absurd for them to occur to us as likely or credible. My own view, for what it is worth, would be that this exclusion and this absurdity almost always results from the operation of pragmatic factors, matters of fact, and contextual knowledge rather than from anything lying within the purview of semantics, for which all the absurd alternatives must count as theoretically possible.