Grice's Meaning Project: Wayne A. Davis
Grice's Meaning Project: Wayne A. Davis
Grice's Meaning Project: Wayne A. Davis
Wayne A. Davis
RESUMEN
El proyecto que Grice concibió sobre el significado consistía en explicar qué es
el significado “no-natural” en términos del significado del hablante y el significado
del hablante en términos de intenciones. Examino aquí el intento de Grice de llevar a
cabo este proyecto examinando las definiciones particulares que propuso. Estas defi-
niciones tienen numerosos problemas muy difíciles de tratar, que han llevado a los es-
tudiosos de Grice a concluir que todo el proyecto era impracticable. Muestro que la
teoría del significado de la expresión que he desarrollado resuelve los problemas ofre-
ciendo definiciones alternativas del significado del hablante y del significado de la
expresión. Los recursos teóricos adicionales más importantes que se necesitan son la
convención, el pensamiento ocurrente como estado mental distinto de la creencia y la
estructura del pensamiento. El proyecto de Grice es todavía el enfoque más promete-
dor para explicar en qué consiste el significado lingüístico.
ABSTRACT
Grice’s meaning project was to explain what ‘non-natural’ meaning is by defin-
ing expression meaning in terms of speaker meaning and speaker meaning in terms of
intention. I review Grice’s attempt to carry out this project by examining the particu-
lar definitions he proposed. These have numerous intractable problems, which led
scholars to conclude that the basic project was flawed. I briefly show how the expres-
sion theory of meaning I have developed solves the problems by offering alternative
definitions of speaker and expression meaning. The most important additional theo-
retical resources needed are convention, occurrent thought as a mental state distinct
from belief, and thought structure. Grice’s project is still the most promising approach
to explaining what linguistic meaning is.
41
42 Wayne A. Davis
Grice’s first move was to show that the question needs to be refined.
For there are many different senses of meaning. He called the senses illus-
trated in (1) ‘natural’.
(1)
a. Tiger Woods means to win. (Intention)
(2)
a. By ‘Palmer beat Nicklaus’, Tiger means that Palmer defeated
Nicklaus. (Speaker Meaning)
While the meaning of ‘means’ in (2) seems related in some way to its mean-
ings in (1), we cannot replace ‘means’ with ‘intends’, ‘indicates’, or ‘shows’
in (2) without completely changing the truth conditions of the sentences.
Grice observed that in (1)(a) and (b), we cannot infer that anyone meant any-
Grice’s Meaning Project 43
thing by those spots or dark clouds. He also observed that whereas we can re-
formulate (2) using quotation marks, we cannot do that in (1). Thus (2)(b) is
equivalent to ‘Palmer beat Nicklaus’ means “Palmer defeated Nicklaus”, but
(1)(c) is not equivalent to Those spots mean “He has measles”. Human activ-
ity is clearly entailed by both sentences in (2).
Grice went on to say that “the distinction between natural and non-
natural meaning is, I think, what people are getting at when they display an
interest in a distinction between ‘natural’ and ‘conventional’ signs”[Grice
(1957), p. 379]. Hobbes [(1655), §2.3] noticed that while we can say both
that dense clouds signify rain, and that stone markers signify the boundaries
of fields, one sign is completely natural, the other conventional. Whereas
dense clouds would indicate rain even if humans had never existed, a pile of
stones indicates the boundary of a field only because humans use stones for
the purpose of marking boundaries. Tiger’s saying ‘Palmer beat Nicklaus’ is
a conventional rather than natural sign that he is thinking the thought that
Palmer defeated Nicklaus. Note, though, that Hobbes is using ‘signify’ as a
synonym of ‘indicate’ or ‘show’ in both cases. What Tiger said is a causal
consequence of what he thought, so the latter can be inferred from the former.
Hobbes’s distinction between natural and conventional signs is a distinction be-
tween things with what Grice calls natural meaning.2 Note also that while natu-
ral meaning does not entail human action, it does not exclude it either.
Grice was concerned to analyze non-natural meaning. He was primarily
interested in what it is for words, phrases, and sentences to mean something,
and in what it is for a speaker to mean something by uttering linguistic ex-
pressions. Grice’s fundamental thesis was that expression meaning can be de-
fined in terms of speaker meaning, and speaker meaning in terms of intention;
hence non-natural meaning can be defined in terms of natural meaning.
Grice’s project was then to produce those definitions. A satisfactory definition
must be noncircular. Grice’s analyses will meet this requirement if non-
natural meaning can be defined in terms of natural meaning as expected.
Grice offered an analysis of intention elsewhere, doing so exclusively in
terms of nonlinguistic notions like belief and desire. The elements added by
other leading analyses are similarly non-linguistic.3 If successful, Grice’s pro-
ject would reduce semantics to psychology.
It might be thought that Grice’s project is doomed from the start. For
what I call the ‘naive theory’ of speaker meaning is very plausible. It is natu-
ral to think that S meant ȝ by e iff (i) S uttered e and (ii) e means ȝ. This na-
ive theory maintains that (2)(a) is true because (2)(b) is true. It reduces
speaker meaning to expression meaning, the opposite of what Grice’s thesis
maintains. The naive theory is plausible to the extent that people do typically
mean ȝ by uttering words that mean ȝ. But it takes only a few moments of re-
flection to realize that this simple theory is indeed naive. What a speaker
means may depart from what the sentence she utters means in many cases,
44 Wayne A. Davis
including slips of the tongue, ellipses, figures of speech, and nonce meanings
or codes. (2)(a) and (2)(b) are logically independent: neither entails the other.
I might leave B’s handkerchief near the scene of a murder in order to induce the
detective to believe that B was the murderer; but we should not want to say that
the handkerchief (or my leaving it there) meantNN anything or that I had meantNN
by leaving it that B was the murderer. Clearly we must at least add that, for x to
have meantNN anything, not merely must it have been “uttered” with the intention
of inducing a certain belief but also the utterer must have intended an “audience”
to recognize the intention behind the utterance [Grice (1957), pp. 381ff].
Grice’s second shot was therefore: S meant that p iff S intended to produce in
some audience A the belief that p by means of recognition of this intention.
Grice’s Meaning Project 45
Despite what seemed clear to Grice, there are many cases in which a speaker
means something without intending an audience to recognize his intentions.
In addition to the private diary case, and that of expected misunderstanding,
people often send unsigned notes or talk to their babies and pets. Many
Griceans compounded the problem by interpreting Grice’s definiens as a re-
flexive intention, a very difficult thing for anyone to understand, let alone for
everyone to have who means something.5 Grice himself took ‘this intention’
to refer to the intention to produce in A the belief that p, period; see Grice
(1969), p. 151. Others followed Grice’s lead, but required an infinite regress
of intentions — also difficult to imagine anyone having.6
These problems can be avoided by defining speaker meaning in terms
of both kinds of natural meaning, thereby melding Grice’s approach with the
basic insight of Hobbes (1655) and Locke (1690). On my view [Davis
(2003), Ch. 2], the relevant intention is to provide an indication that one has
a particular belief. Thus Tiger meant that Palmer defeated (rather than
struck) Nicklaus only if he intended his utterance to be an indication that he
believes Palmer defeated (rather than struck) Nicklaus. In either case, the in-
tended indication would be a conventional rather than natural sign of the belief.
Grice did not mean that B was the murderer by leaving the handkerchief in the
case he had in mind because Grice did not intend the handkerchief to indicate
anything about what he believed. He only intended it to be evidence that B was
there. Grice would mean that B was the murderer in a variant of the case in
which he did intend the handkerchief to be an indication that he believed B was
the murder (imagine the detective asking ‘Who is the murderer?’ and Grice
then overtly producing B’s handkerchief as an answer). A speaker may intend
to produce an indication of belief even if he knows that the audience already
has that belief, that the audience will misunderstand him, or that he has no au-
dience. As a first approximation: S meant that p by e iff S produced e as an in-
dication that he believes that p. Given the definition of expression I offer
[Davis (2003), Ch. 3], it follows that S meant that p by e iff S used e to express
the belief that p. I refine this basic definition to handle some unusual cases, but
the refinements are not relevant to understanding what expression meaning is.
Ziff [(1967), p. 446] and Schiffer [(1972), pp. 2-3)] noticed that there
are two distinct kinds of speaker meaning, as illustrated by (3).
(3)
a. By uttering ‘Palmer beat Nicklaus’, Tiger meant that Palmer de-
feated Nicklaus. (Cognitive)
In the most typical cases, (3)(a) and (b) are both true or both false. But there
are cases in which one is true but not the other. Imagine, for example, that
Tiger meant “defeat” by ‘beat’, but was using irony. In that case, (3)(b) is
true but (3)(a) is false. In the same case, Tiger meant that Nicklaus defeated
Palmer even though he meant “Palmer defeated Nicklaus” by the sentence he
uttered. I account for this distinction by defining cogitative speaker meaning
as the direct expression of thoughts rather than beliefs. I can think the thought
that there are only five planets even though I do not believe that there are
only five, and I can believe that there are more than five planets even when I
am not thinking that thought. The truth of (3)(b) requires that Tiger intended
the words he uttered to be an indication that he is thinking the thought that
Palmer defeated Nicklaus, not an indication that he believes it. Moreover, Tiger
must intend his utterance to indicate that thought directly, not by indicating an-
other thought. In contrast, Tiger also intended his utterance to indicate that he is
thinking the thought that Nicklaus defeated Palmer, but indirectly, by indicat-
ing the opposite thought. Hence: S meant “p” by e iff S produced e as a direct
indication that he is thinking the thought “p.” Again, this is only a first ap-
proximation, but the refinements are immaterial to what follows.7
For instance, ‘Time flies’ means that time goes by quickly is true because Peo-
ple mean that time goes by quickly by ‘Time flies’. Similarly, ‘2 + 2 = 4’ means
that two plus two equals four because People mean that two plus two equals
four by ‘2 + 2 = 4’. Grice’s theory is thus a version of the Wittgensteinian the-
sis that meaning is use. What words mean, on Grice’s view, is determined by
what they are used to do. Formulated this generally, the thesis is a platitude.
Grice’s specific formulation was much more plausible than others, however,
because of which use he took to be determinative of meaning. A sentence
means that p in virtue of being used specifically to mean that p. This is liable
to sound vacuous, unless we remember that it is defining expression meaning
in terms of speaker meaning.
Wittgenstein himself took the use of an expression to be ‘the special
circumstances, or “surroundings”, in which it is spoken or written’ [Malcolm
(1967), p. 337]. This is too general to provide a theory: which of all the cir-
Grice’s Meaning Project 47
With this correction, a Gricean theory becomes more like that of Hobbes and
Locke in a second way.
A second flaw is that Grice’s initial definition of sentence meaning
counts common implicatures as meanings. In another seminal work, “Logic
and Conversation” (1975), Grice observed that speakers can properly use a
sentence to mean things other than what the sentence means. Suppose John
asks Jane ‘Does Mary want to go to the movies?’ Jane might answer ‘She has
to study’. Typically, one of the things Jane would have meant by saying this
is that Mary cannot go to the movies. Jane meant something that the sentence
she uttered does not mean, illustrating again the need to distinguish speaker
meaning from expression meaning. Grice (1975: 24ff) introduced the techni-
cal term ‘implicature’ for the phenomenon of meaning one thing by saying
something else. Jane implicated that Mary cannot go to the movies, and peo-
ple who say ‘Some cats are black’ commonly implicate “Not all cats are
black”. The problem for Grice’s definition of sentence meaning is that it
wrongly counts the things people implicate as meanings. But ‘She has to study’
does not mean “She cannot go to the movies,” and ‘Some cats are black’ does
not mean “Not all cats are black”. There is no inconsistency in ‘Some cats are
black, indeed all are’.
The requirement that meaning be conventional usage does not help
here. There is nothing unconventional about Jane’s usage. Indeed, her usage
illustrates a conventional way of implicating something [Davis (1998), Chs.
5-6]. Jane’s usage would be unconventional if she meant “cannot go to the
movies” by the phrase ‘cannot study’. But this involves cogitative, not cogni-
tive speaker meaning. This second problem with Grice’s analysis does have a
simple fix, however. We can avoid it by defining sentence meaning as direct
speaker meaning. ‘Some cats are black’ does not mean “Not all cats are
black” because speakers mean that not all cats are black by saying and mean-
ing that some cats are black. Similarly, Jane means that Mary cannot go to
the movies by meaning that she has to study.
in terms of sentence meaning.10 But no attempt to show how this can be done
has been remotely successful, and it should be clear on reflection that it can-
not be done. The information that ‘Į ȕ’ means that dogs bark is insufficient to
determine whether ‘Į’ means “dogs” and ‘ȕ’ “bark” or ‘ȕ’ “dogs” and ‘Į’
“bark”, nor even whether ‘Į’ and ‘ȕ’are meaningful components like words
or meaningless components like letters. Moreover, the sentential primacy
thesis contradicts compositionality by getting the order of determination
backwards. ‘Dogs bark’ means what it does because ‘dogs’ and ‘bark’ mean
what they do, not vice versa.11
Grice addressed the problem of word meaning very briefly in “Utterer’s
Meaning and Intentions” (1969). He did so as part of his attempt to define
cogitative speaker meaning in terms of cognitive.
Thus by ‘beat’ Tiger meant “defeated” when he said ‘Palmer beat Nicklaus’
if Tom meant that Palmer defeated Nicklaus. Grice’s suggestion is that a
speaker means something by a word because he uses a sentence containing
that word to meant that something is the case. Grice can then say, along the
lines sketched in the previous section, that a word x means ij iff people di-
rectly and conventionally mean ij by x.
Grice’s attempt to reduce word meaning to sentence meaning is clever,
but quite unsatisfactory. First, it fails completely when the language U is us-
ing is different from the language we are using to say what U means. Sup-
pose U is a French woman who says ‘C’est la vie!’ with its usual meaning.
Then by ‘vie’ she meant “life.” But the result of replacing ‘vie’ with ‘life’ is
not a sentence of English. So it makes no sense to say that by uttering ‘C’est
la vie’ she meant that C’est la life. Second, Grice’s analysis of word meaning
entails that a speaker cannot mean anything by a word that is not used as part
of a sentence. But in the title of this paper, I meant “project” by ‘project’.
Third, the analysis produces nonsensical results in indefinitely many cases.
Suppose Lindbergh says ‘The aeroplane is lucky’, meaning that the airplane
is lucky. Then Grice’s analysis rules that by ‘ero’ Lindbergh meant “ir.” We
know this makes no sense because we know that ‘ero’ and ‘ir’ have no mean-
ings. But Grice’s analysis provides no basis for distinguishing ‘ero’/“r” from
‘vixen’/“female fox.” Or suppose Ted says ‘Mary’s car is red and Jane’s is
blue’, meaning that Mary’s car is red and Jane’s car is blue. Then Grice’s
50 Wayne A. Davis
analysis rules that by ‘is’ Ted meant “car is.” Fourth, Grice’s proposals gives
totally wrong results for pronouns. It rules, for example, that by ‘I’ Tiger means
“he” given that by uttering ‘I won’ Tiger meant that he won.12
We can handle word meaning within the general Gricean framework by
focusing on cogitative speaker meaning rather than cognitive. As noted in
§II, (3)(b) means that Tiger produced ‘Palmer beat Nicklaus’ as a direct indi-
cation that he is thinking the thought that Palmer defeated Nicklaus. I have
argued at length that thoughts are complex, and use the terms ‘concept’ and
‘idea’ interchangeably for thoughts and thought parts [Davis (2003), Part III].
The thought that Palmer defeated Nicklaus has at least three parts, the idea of
Palmer, the idea of Nicklaus, and the concept “defeated.” We think a thought
when the thought occurs to us. A thought occurs to us only if all of its parts
do. The thought that Palmer defeated Nicklaus occurs to Tiger only if its
three component concepts occur to him in a particular arrangement. The
thought that Nicklaus defeated Palmer occurs to Tiger only if the same con-
cepts occur to him in a different arrangement (the concept of Palmer trading
places with the concept of Nicklaus). The thought that Palmer struck Nicklaus
occurs to Tiger if the concepts of Palmer and Nicklaus occur with the concept
“struck” the way they occurred with the concept “defeated” in the other
thought. We are said to ‘conceive’ a concept when it occurs to us. The idea of
Palmer occurs to us when we are thinking of Palmer.
We can now define cogitative speaker meaning for words as easily as
for sentences: S meant ij by e iff S produced e as a direct indication that he is
conceiving the concept ij. Or using my definition of expression, S meant “ij”
iff S directly expressed the idea “ij.” Tiger means “defeated” by ‘beat’ when
Tiger uses ‘beat’ to directly express the idea “defeated.” He means “struck”
when he uses ‘beat’ to directly express the idea “struck.” The French woman
meant “life” by ‘vie’ because she used ‘vie’ to directly express the idea of
life. Lindbergh did not mean anything by ‘ero’ because he did not use it as a
direct indication that he is conceiving any concept. Speakers of English do
not use ‘I’ to express the idea “he.”13 We can say in general that individual
word meaning is direct and conventional speaker meaning. Hence the word
‘beat’ means “defeated” because speakers conventionally and directly use the
word ‘beat’ to mean “defeated” — that is, to express the idea “defeated.”
Whereas I use the terms ‘idea’ and ‘concept’ interchangeably to denote
thought parts, these terms can also be used to denote conceptions (systems of
belief about an subject), and ‘idea’ has often been used to mean sensory im-
ages. Neither images nor conceptions are expressed the way thoughts and
their parts are. The word ‘necessity’ is meaningful even though we cannot
form an image of necessity. And two people may use the term ‘China’ with
the same meaning even though they have very different conceptions of
China. Meaning cannot be defined in terms of images or conceptions. Since
thinking a thought is not itself a linguistic act, and does not entail anything
Grice’s Meaning Project 51
with linguistic meaning, defining the meaning of a word in terms of the ex-
pression of a thought part is in no way circular or regressive.14
One of the most serious defects of Grice’s analysis of word meaning is
that it precludes compositionality. If Grice’s analysis of what it is for a word
to have meaning were correct, then it could not be true that sentences mean
what they do because of what the words in the sentences mean. It could not
be true that ‘Palmer beat Nicklaus’ means “Palmer defeated Nicklaus” in part
because ‘beat’ means “defeated.” Grice’s definition of word meaning does
not assign anything to words that they can contribute to the meaning of sen-
tences. The alternative I have developed does. Words have meaning in virtue
of being used to express thought parts. Meaningful sentences express com-
plete thoughts. Except for idioms and the like, the words in a sentence ex-
press parts of the thought expressed by the whole sentence.
V. PRODUCTIVITY
because: (i) its component words have certain meanings and (ii) it has the
grammatical structure ‘S V O’. Grice’s account of sentence meaning cannot
be complete because it lacks projective devices [Ziff (1967), p. 450].
I account for compositional meaning, and thereby the semantic produc-
tivity of languages, by giving a recursive definition of linguistic meaning
[Davis (2003), Ch. 10]. Recursive definitions have a base clause and a recur-
sion clause. The base cause of my definition is provided by the modification
of Grice’s thesis I sketched above: e means ȝ if e is used to conventionally
and directly mean ȝ — that is, to conventionally and directly express the idea
ȝ. This holds for both individual word meaning and idiomatic sentence mean-
ing. Thus ‘flies’ means “travels through the air on wings” because speakers
conventionally and directly use ‘flies’ to mean “travels through the air on
wings.” And ‘Time flies’ has the idiomatic means “Time goes by quickly” be-
cause people conventionally and directly use ‘Time flies’ to mean “Time goes
by quickly.”
The recursion clause of my definition of linguistic meaning is made
possible by the fact that there are conventions to use sentences with a given
(surface) structure to express ideas with a particular structure. In general,
there are conventions whereby word structures express idea structures. Let us
use c(V)[c(S), c(O)] to represent the structure that the thought that Palmer de-
feated Nicklaus has in common with the thought that Mary kissed Jack. Let
us use c(defeated)[c(Palmer), c(Nicklaus)] to represent the thought that
Palmer defeated Nicklaus. Then we can say that it is conventional among
speakers of English to use sentences of the form ‘S V O’ to express thoughts
with the structure c(V)[c(S), c(O)]. This is a common practice, which serves
the purpose of communication and perpetuates itself in part through prece-
dent. And as natural as it seems to English speakers to do things this way, the
practice could have been to use sentences of the form ‘O V S’ or ‘S O V’ to
express thoughts with that structure; word order could have a different signifi-
cance, and does in other languages. As a first approximation, the recursion
clause of my definition is: e expresses i if speakers use E[x1, x2,..., xn] conven-
tionally to directly express I[c(x1), c(x2),..., c(xn)], where e = E[e1, e2,..., en] and
ei expresses c(ei) for all i from 1 to n. Thus ‘The president beat Beethoven’ ex-
presses the idea (and thus means) “The president defeated Beethoven” because
(i) speakers conventionally use sentences of the form ‘S V O’ to express the
idea structure c(V)[c(S), c(O)]; and (ii) ‘beat’ expresses c(defeated), ‘The
president’ expresses c(the president), and ‘Beethoven’ expresses c(Beethoven).
If English were an OVS language like Hixkaryana of Brazil, then ‘The presi-
dent beat Beethoven’ would mean that Beethoven defeated the president. For in
that case it would be conventional to use sentences of the form ‘S V O’ to ex-
press the idea structure c(V)[c(O), c(S)]. This brief exposition ignores many de-
tails and complications, but it should suffice to indicate how Grice’s project can
be carried out without departing from his fundamental vision.
Grice’s Meaning Project 53
Assuming that ‘Fido’ is an Į part and ‘shaggy’ a ȕ part, Grice further sup-
poses that the following correlations hold:
Grice claims that the two procedures together with the two correlations de-
termine the following resultant procedure.
What is so surprising about this claim is that it is hard to see how RP could re-
sult from C1 and C2 unless we assumed a referential theory of meaning — that
is, unless we assumed that C1 and C2 told us what ‘Fido’ and ‘shaggy’ mean.
But since ‘Fido’ and ‘Jones’ dog’ have different meanings even if they have
the very same referent — Frege’s (1892) basic point, as well as Kripke’s
[(1972), II], this cannot be right. If RP did result from the given procedures,
then so should the following:
To use ‘Fido is shaggy’ to mean that the dog that attacks Jones’s mail-
man has more hair than a hairless thing.
But these procedures assign different meanings to ‘Fido is shaggy’ if they as-
sign any at all.
What Grice set out to explain is how the meaning of a sentence is con-
sequential on what the words in it mean. The only feature of words Grice
looked at, however, were reference and denotation. The meaning of a sen-
tence is not consequential on the reference or denotation of the words com-
posing it. Sentences with very different meanings can be composed of words
with the same reference or denotation.
Since Grice’s first basic procedure tells us nothing about how sentence
meaning is consequential on word meaning, his second basic procedure is
carrying all the weight. But as a semantic postulate it is completely implausi-
ble. ‘R-correlate’ and ‘D-correlate’ are technical terms of Grice’s theory,
terms that he did little to explain. We cannot assume that ordinary speakers of
any natural language have those concepts. So we cannot assume that ordinary
speakers have the intentions required for such a meaning procedure. More-
over, the meaning P2 assigns to a sentence is metalinguistic: But when ordi-
nary speakers of English use ‘Fido is shaggy’, they are not talking about the
words ‘Fido’ and ‘shaggy’. So for a second reason, P2 does not describe the
conventional practice of speakers who use ‘Fido is hairy’. The most fundamen-
tal problem is that P2 assumes precisely what Grice is trying to explain. It is the
meaning of ‘is’ that determines that a sentence of the form ‘Į is ȕ’ can be used
Grice’s Meaning Project 55
to mean that Į is ȕ. It is because ‘is’ means “is” that ‘Fido is shaggy’ cannot be
used to mean that Fido isn’t shaggy, nor that Fido loves shagginess.
Grice found compositionality and productivity intractable because
he limited himself to cognitive speaker meaning, and had no theory of thoughts
as complex objects. By using cogitative speaker meaning instead, and taking
into account thought parts and thought structure, the Gricean project can be
completed.
A final problem Grice never saw arises from the fact that words have
different meanings in different languages. The five-letter sequence ‘lance’
means “hazzard” in Spanish, “long spear” in English. Whose conventions de-
termine the meanings of words in English? If we say that the meaning of a
word in English is determined by the usage of English speakers, and define
English speakers as those who use words to mean what they mean in English,
then we will have defined ourselves into a circle.
Grice briefly acknowledged the relativity of meaning in ‘Utterer’s
Meaning, Sentence Meaning, and Word Meaning’ (1968). Following Neale
[(1992), p. 553], I will simplify Grice’s formulation by ignoring his attempt
to accommodate non-indicative moods.
For population group G, complete utterance-type X means “p” iff (a) at least
some (many) members of G have in their behavioral repertoires the procedure
of uttering a token of X if they mean that p, and (b) the retention of this proce-
dure is for them conditional on the assumption that at least some (other) mem-
bers of G have, or have had, this procedure in their repertoires. [Simplified from
Grice (1968), p. 553]
This can be generalized to cover word meaning along the lines I sketched in
§IV. We can see in Grice’s formulation the rudiments of the idea that linguis-
tic meaning is conventional speaker meaning. Clause (a) gets at the idea that
linguistic meaning is determined by what is commonly meant, and clause (b)
at the idea that precedent is one of the mechanisms whereby such practices
perpetuate themselves. It is characteristic of conventions that they vary from
group to group. It is conventional to drive on the left in Britain, on the right
in Spain. So Grice’s formulation does predict that what words mean will vary
from group to group. Grice’s gloss provides no help, however, in understand-
ing what it is for ‘lance’ to mean “hazzard” in Spanish and “long spear” in
English. It does enable us to draw some conclusion about what words mean
in the English speaking population. But these predictions give us little insight
into linguistic relativity. For example, when Grice’s formulation is generalized
to words, it entails that ‘lance’ means both “hazzard” and “long spear” in the
English speaking population. For many members of the English speaker popu-
56 Wayne A. Davis
lation (those in subpopulations that also speak Spanish) do have the procedure
of using ‘lance’ to mean “hazzard,” and the retention of this procedure is condi-
tional on the fact that other members of the English speaking population (those
in their Spanish speaking subpopulation) use ‘lance’ to mean “hazzard.”
The evolutionary character of living languages provides the solution to
the linguistic relativity problem [Davis (2003), Ch. 11]. Languages in general
are systems of ‘modes of expressions’, which pair expressions with ideas or
other mental states. Living languages like English and Spanish are those that
depend on the conventional usage of an evolving lineage of speakers in certain
ways. A natural language is discovered and named when a group of speakers is
identified with a distinctive way of expressing ideas. As the conventions per-
petuate themselves over time, they are passed to new members of the speech
community. Since conventions are seldom perfect regularities, they can change
over time. Previously unconventional uses can spread and become conven-
tional. What words mean in a living language today is determined by the con-
ventional usage of the linguistic descendants of prior users. Language commu-
nities often divide, with their subgroups becoming isolated from each other.
The original language then generally evolves differently in the separate line-
ages, as when a large group of Dutch speakers migrated to South Africa begin-
ning in the seventeenth century. We can then consider contemporary
Netherlands Dutch and Afrikaans to be different dialects of Dutch, or separate
languages that evolved from a common origin. What a word means in Afri-
kaans today depends on the current meaning conventions of the lineage of
speakers that can be traced back to the original Dutch settlers of South Africa,
and what a word means in Netherlands Dutch today depends on the meaning
conventions of the lineage of speakers that continued living in the Netherlands.
While linguists and philosophers of language are most interested in
what it is for words in living, natural languages to have meanings, that does not
exhaust the phenomenon of linguistic meaning, as Grice [(1989), pp. 298ff],
recognized. Living languages die when use of the language becomes too un-
common to perpetuate itself. In artificial languages, words mean what they
do as a result of stipulation rather than convention. And in idiolects, words
mean what they do because of personal practices of the individual.
A complete foundational theory of meaning along Gricean lines is more
complex than anything Grice committed to print, but probably not more intri-
cate than what Grice suspected was the case.
Philosophy Department
Georgetown University
215 New North37th and O Streets NW,
Washington, DC 20057, USA
E-mail: davisw@georgetown.edu
Grice’s Meaning Project 57
NOTES
1
See M. Green (2003); Self-Expression.
2
Grice [(1957), p. 379] observed that not everything with meaning (natural or
non-natural) counts as a sign. Whereas the utterance of a sentence may be a sign, the
sentence itself is not.
3
See Grice (1971); Audi (1973); Davis (1984); Bratman (1987). I rebut argu-
ments that intention must be defined in terms of meaning in Davis (2003), Ch. 22.
4
For a full survey and critique, see Davis (2003), Ch. 4.
5
See e.g. Bach & Harnish (1979), 15, who require the speaker to intend the ef-
fect consisting of the hearer recognizing the speaker’s intention to produce that effect.
Until we know what ‘that effect’ refers to, we cannot know what intention the hearer
is intended to recognize. And until we know what intention the hearer is intended to
recognize, we cannot know which effect consists of the hearer recognizing that inten-
tion. Since ‘that effect’ is anaphoric on ‘the effect consisting of the hearer recognizing
[...]’, we cannot understand anything.
6
See Schiffer (1972), p. 63; Grice (1982), p. 242; (1989), p. 302.
7
Grice’s [(1969), pp. 148-150] analysis of cogitative speaker meaning was in-
adequate in many says. See Davis (2003), § 4.5; and below, §IV.
8
Compare and contrast Grice (1982), pp. 238ff; (1989), pp. 298ff.
9
See Davis (2003), Ch. 9. This definition derives from Hume (1739), p. 490)
and David Lewis (1975), pp. 4-5).
10
See e.g. Alston (1964), pp. 36-39; Grice (1969), pp. 148-150; Schiffer
(1972), 6, p. 166; (1987a), pp. 92 y 214-216; Loar [(1981), § 9.9; Chierchia and
McConnell-Ginet (1990), pp. 60-61.
11
See Davis (2003), § 8.4 for the full argument and additional references.
12
Neale (1992), pp. 553ff.
13
I am extending my expression theory of meaning to indexicals in Indexical
Meaning and Concepts.
14
See Davis (2003), Part IV for a full discussion of this point.
REFERENCES
ALSTON, W.P. (1964), Philosophy of language, Englewood Cliffs, NJ: Prentice-Hall.
AUDI, R. (1986), “Intending, intentional action, and desire”, in The ways of desire, ed.
J. Marks, Precedent Press, pp. 17-38.
BRATMAN, M. (1987), Intentions, plans and practical reason, Cambridge, MA,
Harvard University Press.
DAVIS, W.A. (1998), Implicature: intention, convention and principle in the failure of
gricean theory, Cambridge, Cambridge University Press.
— (2003), Meaning, expression and thought, New York, Cambridge University Press.
— (Forthcoming), Indexical meaning and concepts.
FREGE, G. (1892b), “On sense and reference”, in Translations from the philosophical
writings of Gottlob Frege, ed. P. Geach & M. Black, pp. 56-78, Oxford, Basil
Blackwell (1952). From Zeitschrift für Philosophie und philosophische Kritik,
100, pp. 25-50.
58 Wayne A. Davis
GREEN, M.S. (2003), “Grice’s frown: on meaning and expression”, in Saying, meaning
and implicating, ed. G. Meggle, Leipzig, Leipsizger-Universitatsverlag, pp. 200-
219.
— (Forthcoming), Self-Expression, Oxford, Oxford University Press.
GRICE, H.P. (1957), “Meaning”, Philosophical Review, 66, pp. 377-88.
— (1968), “Utterer’s meaning, sentence meaning, and word meaning”, Foundations
of Language, 4, pp. 225-242.
— (1969), “Utterer’s meaning and intentions”, Philosophical Review, 78, pp. 147-
177.
— (1971), “Intention and uncertainty”, Proceedings of the British Academy, 57, pp.
263-279.
— (1975), “Logic and conversation”, in Syntax and semantics, 3: speech acts, ed. P.
Cole & J. Morgan, New York, Academic Press.
— (1982), “Meaning revisited”, in Mutual knowledge, ed. N. Smith, New York,
Academic Press, pp. 223-243.
— (1989), Studies in the way of words, Cambridge, MA, Harvard University Press.
HOBBES, T. (1655), Computation or logic, in Thomas Hobbes: Part I of De corpore, ed.
I.C. Hungerland & G.R. Vick, New York, Abaris Books (1981), pp. 171-331.
HUME, D. (1739), A treatise of human nature, ed. L.A. Selby-Bigge, Oxford,
Clarendon Press, (1888).
KRIPKE, S. (1972), “Naming and necessity”, in Semantics of natural language, ed. D.
Davidson & G. Harman, Dordrecht, D. Reidel, pp. 253-355, 763-769.
LEWIS, D. (1975), “Languages and language”, in Minnesota Studies in the Philosophy
of Language, ed. K. Gunderson, vol. 7, Minneapolis, University of Minnesota
Press, pp. 3-35.
LOAR, B. (1981), Mind and meaning, Cambridge, Cambridge University Press.
LOCKE, J. (1690), An essay concerning human understanding, ed. A.C. Fraser, New
York, Dover Publications (1959).
MALCOLM, N. (1967), “Wittgenstein, Ludwig Josef Johann”, in Encyclopedia of
Philosophy, ed. P. Edwards, vol. 8, New York, Macmillan, pp. 327-340.
NEALE, S. (1992), “Paul Grice and the philosophy of language”, Linguistics and
Philosophy, 15, pp. 509-59.
SCHIFFER, S. (1972), Meaning, Oxford, Clarendon Press.
— (1987), Remnants of meaning, Cambridge, MA, MIT Press.
SEARLE, J. (1983), Intentionality, Cambridge, Cambridge University Press.
ZIFF, P. (1967), “On H.P. Grice’s account of meaning”, Analysis, 28, 1-8, reprinted in
“Readings in the philosophy of language”, ed. J.F. Rosenberg & C. Travis,
Englewood Cliffs, NJ: Prentice-Hall, Inc. (1971).