On H. P. Grice’s account of meaning¹

Because I believe the coin is counterfeit, because it seems to be gaining currency, I mean to examine and attempt to discredit an account of meaning circulated some time ago by H. P. Grice²

Those among us concerned with problems of semantics are much concerned with the sense(s) or meaning(s) of the morpheme ‘mean’ in (1) and (2):

(1) The sentence ‘ Snow is white ’ means snow is white.

(2) The adjective ‘ ungulate ’ means having hoofs.

Other senses or meanings of ‘ mean ’ are of interest in semantics primarily only in so far as they have some bearing on ‘mean’ in either (1) or (2).

Grice’s paper is entitled ‘ Meaning ’. It appears to be an account of meaning that is supposed to have some bearing on the senses of ‘mean’ in (1) and (2).

Grice apparently says that ‘mean’ in (1) and (2) is used in what he calls ‘non-natural ’ senses of the verb. He uses the abbreviation ‘ mean-nn ’ to mark the ‘ non-natural ’ senses of ‘ mean ’. He offers something of an analysis of ‘ mean-nn ’ (and of the morphological variants, ‘meant-nn’, ‘means-nn’, and so forth).

Does ‘mean’ in (1) or (2) have the sense(s) indicated by Grice’s analysis?

It will simplify matters to adopt the following convention: when the expression ‘ mean-nn ’ (or any of its variants) is used here, that expression is to have the sense(s) indicated by and in conformance with Grice’s analysis. The problem of this paper can then be stated in a simple way.

Consider (1nn) and (2nn):

(1nn) The sentence ‘ Snow is white’ means-nn snow is white.t

(2nn) The adjective ‘ungulate’ means-nn having hoofs.

Is (1nn) simply a restatement of (1), (2nn) of (2)?

After an ingenious intricate discussion, Grice arrives at the following ‘ generalizations ’ (p. 58):

(i)‘A meant-nn something by x ’ is (roughly) equivalent to ‘ A intended the utterance of x to produce some effect in an audience by means of the recognition of this intention ’.

(ii)‘ x meant-nn something ’ is (roughly) equivalent to ‘ Somebody meant-nn some¬ thing by x ’.

(iii) ‘x means-nn (timeless) that so-and-so ’ might as a first shot be equated with some statement or disjunction of statements about what ‘ people ’ (vague) intend (with qualifications about ‘ recognition ’) to effect by x.

It is indicated in the discussion that the letter ‘A ’ is supposed to be replaceable by the name of a person, that the letter ‘ x ’ may be but need not be replaced by a sentence.

Thus it is evident that Grice’s account is supposed to apply to sentence (1) and so to (inn). Although he mentions words in the course of his discussion none of the ‘ generalizations ’ appear to apply to words. Thus apparently he is not concerned to supply, in the paper in question, an account of the sense of ‘means’ in (2). Consequently, I mean to forget about (2) and (2nn) and be concerned here only with (1) and (1nn).

Of the three ‘ generalizations ’ stated, (iii) is the only one that directly applies to (inn). Unfortunately, (iii) is not particularly pellucid. (Even so, what emerges from the fog will be sufficient to establish certain points.) Let us begin by examining (i) and (ii) in the hope of gaining insight into (iii).

On being inducted into the army, George is compelled to take a test designed to establish sanity. George is known to be an irritable academic. The test he is being given would be appropriate for morons. One of the questions asked is: ‘ What would you say if you were asked to identify yourself? ’ George replied to the officer asking the question by uttering (3):

(3) Ugh ugh blugh blugh ugh blug blug.

According to the dictum of (i), George meant-nn something by (3): he intended the utterance of (3) to produce an effect in his audience by means of the recognition of his intention. The effect he intended was that of offending his audience. The accomplishment ofthis effect depended on the recognition of his intention. (The case in question is also in accordance with the various caveats noted by Grice in the course of his discussion: the officer testing George could ‘refuse to be offended’ (p. 58), thus the intended effect was in some sense within the control of the audience; George’s intention to offend was his ‘primary’ intention (p. 58); and so forth.) Consequently, as far as one can tell, (3) fills Grice’s bill.

But even though it is clear that George meant-nn something by (3), it is equally clear that George did not mean anything by (3). Grice seems to have conflated and confused ‘A meant something by uttering x ‘, which is true in a case like (3), with the quite different ‘A meant something by x’, which is untrue in a case like (3).

The malady just noted in connection with (i) of course at once infects (ii). For even though it is clear that George meant-nn something by (3) and hence (3) meant-nn something, it is equally clear that (3) did not mean anything. Indeed, had

(3) meant anything, that would have defeated George’s purpose in uttering (3).

The preceding case admits of the following variation. On being given the test over again by another officer, instead of uttering (3), George uttered (4):

(4)pi.hi.y pi.hi.y

Again in accordance with (i) we can say that George meant-nn something by (4) and what he meant-nn was precisely what he meant-nn by (3). (For we may suppose that he had the same intention in each case, expected the same reaction, and so forth.)

But in this case, not only did George mean something by uttering (4), he also meant something by (4): even though he rightly expected his utterance to be treated as though it were mere noise, what he meant by (4), and what he said, was that he didn’t know. George was perversely speaking in Hopi.³ Here one need not confuse ‘ George did not mean what he said ’, which is true, for he did know the answer to the question, with the quite different ‘ George did not mean anything by (4) ’, which is untrue.

That George meant-nn something by (4) is wholly irrelevant to the question whether George meant something by (4). And the fact that (4) meant-nn something (in virtue of (ii)) is wholy irrelevant to the question whether (4) meant anything. This should be obvious from the fact that what (4) meant had nothing whatever to do with what George intended to effect by uttering the utterance and hence had nothing whatever to do with what (4) meant-nn.

The curious character of (ii) can be further displayed by the following sort of cases. Consider (5):

(5) Claudius murdered my father.

and let us conjure up three contexts of utterance: (a) George uttered a sentence token of type (5), thus he uttered (5a), in the course of a morning soliloquy; (b) George uttered another such token, (5b), in the afternoon in the course of a discussion with Josef; and (c) George uttered another such token, (5c) in the evening while delerious with fever. Now consider (6) and (6nn):

(6) (5a) meant the same as (5b) which meant the same as (5c) which meant the same as (5a).

where (6nn) is the same as (6) save that for each occurrence of ‘ meant ’ in (6), (6nn) has an occurrence of ‘ meant-nn ’. Thus (6) says that the three tokens in question all had the same meaning; (6nn) says that they all had the same meaning-nn.

Although (6) is true, according to Grice’s account (6nn) must be false (here taking (6) and (6nn) to stand for statements). That this is so can be seen as follows.

According to (ii), a sentence S meant-nn something (roughly) if and only if somebody meant-nn something by it. Thus (5a), (5b), and (5c) meant-nn something (roughly) if and only if somebody meant-nn something by them.

Did anyone mean-nn anything by (5c)? Evidently not. For (5c) was uttered while George was delerious with fever, unaware of any audience. Hence (5c) was not intended to produce any effect in an audience. (Here one need not confuse ‘What George said meant nothing ’, which may be true in one sense of ‘ what George said ’, with ‘The expression which George uttered meant nothing’, which is untrue.)

Did anyone mean-nn anything by (5b)? Presumably so. Since (5b) was uttered by George in the course of a discussion with Josef, if anything fits Grice’s account, (5b) does.

Did anyone mean-nn anything by (5a)? It would seem not, for since George uttered (5a) in the course of a soliloquy, it could hardly have been intended to pro¬ duce an effect in an audience. (But perhaps Grice would wish to maintain that, in so far as George was speaking to himself, he was his own audience. But then could he intend to produce an effect in himself by means of a recognition on his own part of his own intention? These are mysteries we may cheerfully bequeath to Mr. Grice.) Evidently (6nn) is untrue even though (6) is true.

Sentence token (5c) exemplifies a case in which even though it was not true that the speaker meant anything by the token, the token nonetheless meant something. One can also produce cases in which a speaker did mean something by an utterance and yet the utterance itself did not mean anything.

George has had his head tampered with: electrodes have been inserted, plates mounted, and so forth. The effect was curious: when asked how he felt, George replied by uttering (7):

(7) Glyting elly beleg.

What he meant by (7), he later informed us, was that he felt fine. He said that, at the time, he had somehow believed that (7) was synonymous with ‘ I feel fine ’ and that everyone knew this.

According to (i), George meant-nn something by (7), and according to (ii), (7) must have meant-nn something. But (7) did not mean anything at all.

The preceding examples should suffice to indicate that Grice’s equivalences (i) and (ii) are untenable. But their extraordinary character can be made even plainer by the following sort of case.

A man suddenly cried out ‘ Gleeg gleeg gleeg! ’, intending thereby to produce a certain effect in an audience by means of the recognition of his intention. He wished to make his audience believe that it was snowing in Tibet. Of course he did not produce the effect he was after since no one recognized what his intention was. Nonetheless that he had such an intention became clear. Being deemed mad, he was turned over to a psychiatrist. He complained to the psychiatrist that when he cried ‘ Gleeg gleeg gleeg! ’ he had such an intention but no one recognized his intention and were they not mad not to do so.

According to Grice’s equivalence (i), the madman meant-nn something by ‘ Gleeg gleeg gleeg! ’ and so, according to (ii), the madman’s cry must have meant-nn something, presumably that it was snowing in Tibet. But the madman’s cry did not mean anything at all; it certainly did not mean it was snowing in Tibet. Had it meant that, there would have been less reason to turn him over to a psychiatrist.

On Grice’s account, good intentions suffice to convert nonsense to sense: the road to Babble is paved with such intentions.

It is time to turn to Grice’s suggestion of an equivalence, (iii). Consider sentences (8),(9) , and (9nn):

(8) He’s a son of a stickleback fish.

(9) Sentence (8) means the male referred to is a son of a small scaleless fish (family Gasterosteidae) having two or more free spines in front of the dorsal fin.

where (9nn) is the same as (9) save that (as in (6nn)) ‘means’ has given way to ‘ means-nn ’.

I take it that there is no reason whatever to suppose that the sense of ‘ means ’ in (9) differs in any way from the sense of ‘ means ’ in (1). Both (9) and (1) are simply of the form: sentence S means m. But if ‘ means ’ in (9) has precisely the same sense as ‘means’ in (1), it follows that (1nn) is not simply a restatement of (1). For (1) and (1nn) differ as (9) and (9nn) differ, and according to Grice’s account, (9) and (9nn) are radically different. That this is so can be seen as follows.

I am inclined to suppose that (8) has been uttered only rarely. Nonetheless (taking (9) and (9nn) to stand for statements), I am reasonably certain that (9) is a reasonably correct statement of the meaning of (8) and I am being reasonable in being so certain.

If it is a correct statement of the meaning of (8), and if (9nn) is simply a restatement of (9), since (9nn) is, (9) must be equivalent to some statement or disjunction of statements about what ‘people’ intend to effect by (8). So Grice evidently maintains ; for despite its vagueness, that is what is indicated by (iii).

But the question ‘ What do people intend to effect by (8) ? ’ would not be a sensible question. Since hardly anyone has ever uttered (8) before, or so I suppose, one can hardly ask what people intend to effect by it.

‘ Then what would people intend to effect by (8) ? ’: the question is somewhat idle. What people would intend to effect by (8) is a matter about which one can only speculate, vaguely. However, since the obvious emendation of Grice’s account invites such speculation, let us speculate.

What would people intend to effect by uttering (8)? Given the acoustic similarity between (8) and a familiar form of expression, given that sticklebacks are known to be tough fish, given that the sex of a fish is not readily determined by the uninitiated, most likely by uttering (8) people would thereby intend to denigrate a contextually indicated male person.

What people would intend to effect by (8) is a subject for profitless speculation. But if one must say something about the matter, I am inclined to suppose that (9nn) does not convey a correct account of what people would intend to effect by uttering (8). Thus (9nn) is presumably untrue.

If (9nn) were simply a restatement of (9), only a fool would profess to being even reasonably certain that (9) is a correct statement and it would be unreasonable of him to be so certain. But I am reasonably certain that (9) is a correct statement of the meaning of (8) and I am not being unreasonable in being so certain. Therefore (9nn) cannot be simply a restatement of (9) and neither can (1nn) be simply a restatement of (1).

Before allowing Grice’s analysis to rest in peace, the moral of its passing should be emphasized.

His suggestion is stated in terms of what people ‘ intend ’, not in terms of what they ‘would intend’. As such it obviously occasions difficulty with novel utterances. But matters are not at all improved by switching to what people ‘would intend’.

For first, if a sentence is such that people in general simply would not utter it, then if they were to utter it, what they would intend to effect by uttering it might very well have nothing to do with the meaning of the sentence. What would a person intend to effect by uttering the sentence ‘ Snow is white and snow is white and snow is white and snow is white and snow is white’? I conjecture that a person uttering such a sentence would be either a philosopher or a linguist or an avant-garde novelist or a child at play or a Chinese torturer. What people would intend to effect by uttering such a sentence would most likely have nothing whatever to do with the meaning of the sentence.

Secondly, the switch to ‘ would ’ would be of help only if there were a constructive method of determining what people would intend to effect by uttering an utterance. There is no such method. There is not likely to be any (at least in our lifetime).

Ignoring the futility of talking about what people ‘would intend’ to effect by uttering an utterance, one need not ignore the fact that what people generally in fact mean may be altogether irrelevant to a meaning of an utterance.

By the spoken utterance ‘ HE GAVE HIM HELL ’ people generally mean what is meant by the written utterance ‘ He gave him hell ’ and not what is meant by the written utterance ‘ He gave him Hell ’. Quite possibly no one has ever said ‘ I saw the children shooting ’ meaning by that he saws children while he is shooting. That is nonetheless one of the meanings of that remarkably ambiguous sentence. Indefinitely many such examples could be supplied.

To be concerned with what people intend (or would intend) to effect by uttering an expression is to be concerned with the use of the expression. As I have elsewhere pointed out and argued at length, the use of an expression is determined by many factors, many of which have nothing (or have nothing directly) to do with its meaning: acoustic shape is one such factor, length another.⁴

Grice’s analysis rings untrue. It was bound to; his alloy lacks the basic ingredient of meaning: a set of projective devices. The syntactic and semantic structure of any natural language is essentially recursive in character. What any given sentence means depends on what (various) other sentences in the language mean.

That people generally intend (or would intend) this or that by uttering an utterance has, at best, as much significance as a statement to the effect that when ‘ Pass the salt! ’ is uttered, generally people are eating, thus what I have elsewhere called the statement of a ‘regularity’⁵ Not all regularities are semantically relevant: a regularity couched in terms of people’s ‘ intentions ’ is not likely to be.

But even if such regularities were somehow relevant, that would not matter much. A regularity is no more than a ladder which one climbs and then kicks away. An account of meaning constituted by (i), (ii), and (iii) never gets off the ground. There is no reason to suppose it can.⁶

¹ This paper is reprinted from Analysis, XXVIII, 1967, 1-8.

² ‘ Meaning’, The Philosophical Review, LXVI, 3 (1957). Reprinted in this volume, pp. 53-9.

³ See B. Whorf, Language, Thought, and Reality (Cambridge: The Technology Press, 1956), P- II4, for the phonetic significance of (4).

⁴ See my Semantic Analysis (Ithaca: Cornell University Press, i960).

⁵ Ibid.

⁶ I am indebted to D. Stampe for useful criticisms of various points.