1
An early version of this paper was published in Savas L. Tsohatzidis, ed., Foundations of Speech Act Theory:
Philosophical and Linguistic Perspectives (London & New York: Routledge 1994). The present, revised, version
appears in Savas L. Tsohatzidis, Truth, Force, and Knowledge in Language: Essays on Semantic and Pragmatic Topics
(Berlin & Boston: De Gruyter 2020).
Speaker meaning, sentence meaning, and metaphor
Savas L. Tsohatzidis
1. Introduction
Two widely held assumptions in contemporary discussions of meaning are, first, that a
distinction deserves to be drawn between what sentences of natural languages mean and
what speakers of those languages mean by uttering those sentences; and, second, that this
distinction largely determines the distinction between semantics and pragmatics as
domains of inquiry, with semantics being dedicated to the analysis of sentence meaning
and pragmatics being concerned with the analysis (usually along lines inspired by the
Gricean theory of conversational implicatures) of the various kinds of meaning that are
instances of speaker meaning rather than of sentence meaning.
Metaphorical meaning is among the kinds of meaning that those who subscribe to
these assumptions (including Grice himself) regard as a kind of meaning that is an
instance of speaker meaning rather than of sentence meaning. But since metaphorical
meaning can hardly be taken to be a pretheoretically obvious case of meaning that is
speaker-based rather than sentence-based, and since there is a considerable number of
not obviously flawed theories, both ancient and modern, where the phenomenon of
metaphor has been regarded as a semantic rather than as a pragmatic phenomenon,
1
attempts to analyse metaphor within the context of the two assumptions can—and
must—be evaluated in at least two ways. First, by considering how well they manage to
distinguish metaphor from other presumed cases of speaker meaning, assuming that it is a
case of speaker meaning. And second, by considering how well they motivate their initial
assumption that metaphor is a case of speaker meaning rather than of sentence
meaning—in other words, that it is a pragmatic rather than a semantic phenomenon. In
what follows, I will be referring to these two questions as the “internal” and the
“external” question, respectively.
My purpose in this essay is to examine, from the two perspectives just indicated,
some claims of John Searle’s paper on metaphor (Searle 1979: 76–116), which is one of
the best-known attempts to analyse metaphorical meaning as a special case of speaker
meaning rather than of sentence meaning, within what Searle regards as a broadly
Gricean framework. I will argue that Searle’s attempt—which, in the respects that are
relevant here, is representative of most other pragmatic approaches to metaphor
2
—fails
both as an attempt to distinguish metaphor from other presumed cases of speaker
meaning, and as an attempt to motivate the assumption that metaphor is in fact a case of
speaker meaning. To the extent that the essay succeeds, then, it suggests that, as far as
1 Cf., among relatively recent works, Cohen and Margalit (1972), Cohen (1979), Lakoff and Johnson (1980), Ross
(1981), MacCormack (1985), Kittay (1987).
2 And also of some well-known approaches which do not label themselves “pragmatic”, such as Davidson’s (1978).
2
the linguistically central phenomenon of metaphor is concerned, the widespread
tendency to reinterpret as many semantic phenomena as possible in pragmatic terms,
may have to be firmly resisted.
2. The internal question
A speaker speaks literally, according to Searle, just in case what he means by uttering a
sentence is identical with what the sentence he utters itself means. (Thus, a speaker who,
in saying, “There are prime numbers”, means that there are prime numbers is a speaker
who speaks literally, since what he means is identical with what the sentence he utters—
namely, the sentence “There are prime numbers”—itself means.) Cases where what a
speaker means by uttering a sentence is not identical with what the sentence he utters
itself means come, according to Searle, in two varieties. On the one hand, a speaker who,
in uttering a sentence, means not only what the sentence he utters itself means but, in
addition, something else as well is a speaker who speaks indirectly. (For example, a speaker
who, in saying to his hearer, “Can you tell me the time?”, means both that he would like
to know whether the hearer can tell him the time and that he would like to know what
the time is, is a speaker who speaks indirectly, since, although he does mean what the
sentence he utters itself means—namely, that he would like to know whether the hearer
has the ability to tell him the time—means, in addition, something else that is not meant
by the sentence he utters—namely, that he would like to know what the time actually is.)
On the other hand, a speaker who, in uttering a sentence, does not mean at all what the
sentence he utters itself means but something different altogether is a speaker who
speaks figuratively. (For example, a speaker who, in saying, “Sally is a block of ice”, means
that Sally is unemotional, or a speaker who, in saying, “Sally was very kind to me”, means
that Sally was very rude to him, are speakers who speak figuratively, since they do not
mean at all what the sentences they utter themselves mean—namely, that Sally is a block
of ice or that Sally was very kind, respectively—but something else altogether—namely,
that Sally is unemotional and that Sally was very rude, respectively.) Finally, a speaker
who speaks figuratively—that is to say, a speaker who does not mean at all what the
sentence he utters means, but something different altogether—can, according to Searle,
be doing the one or the other of two different kinds of things. He may be speaking
ironically, in which case not only does he not mean at all what the sentence he utters
means but means the opposite of what the sentence he utters means; or he may be
speaking metaphorically, in which case he does not mean at all what the sentence he utters
means, but neither does he mean the opposite of what the sentence he utters means.
(Thus, a speaker who, in saying, “You have been very kind to me, Sally”, means that Sally
has been very rude to him is a speaker who speaks ironically, because not only does he
not mean what the sentence he utters means, but means the opposite of what that
sentence means—given that a person’s being rude is the opposite of that person’s being
kind; whereas a speaker who, in saying “Sally is a block of ice”, means that Sally is
unemotional, is a speaker who speaks metaphorically, because, though he does not mean
at all what the sentence he utters means, neither does he mean the opposite of what that
3
sentence means—given that an object’s being a block of ice is not the opposite of that
object’s being unemotional.)
For the sake of argument, let us now assume, with Searle, that metaphorical meaning
is strictly a matter of speaker meaning rather than of sentence meaning. The internal
question to ask about Searle’s proposal can then be split into two sub-questions. First,
whether it provides a basis for a proper distinction between metaphor and irony. And
second, whether it provides a basis for a proper distinction between metaphor and irony,
on the one hand, and non-literal but not figurative (that is, indirect) uses of sentences, on
the other.
If Searle’s way of distinguishing ironical utterances from metaphorical utterances
were correct, there would not be any utterances that one could properly describe as
simultaneously ironical and metaphorical, since saying of an utterance that it is
simultaneously ironical and metaphorical is, within Searle’s theory, equivalent to saying
that what the speaker of that utterance means both is (in view of its ironical character)
and is not (in view of its metaphorical character) opposite to what the sentence he utters
itself means. However, there certainly exist utterances that are properly describable as
being ironical and metaphorical at the same time. If, for example, I wish to suggest that a
certain person that someone has just described as very dependable is not dependable at
all, I may succeed in doing so not only by saying, ironically, “He is very dependable
indeed!”, but also by saying, equally ironically,
(1)
Sure, he is a rock!
But this last utterance would not only be ironical but also metaphorical—as would be any
other utterance in which an animate object would be described as an inanimate one. And
since, on Searle’s theory, speaking ironically is meaning the opposite of what the sentence
one utters means whereas speaking metaphorically is not meaning the opposite of what
the sentence one utters means, anyone uttering the metaphorical sentence in (1) ironically
would have to be counted as meaning and not meaning at the same time the opposite of
what (1) means. Similarly, if I wish to suggest that a man that someone has just described
as very brave is, in fact, not brave at all, I may succeed in doing so not only by saying,
ironically, “That man is very brave indeed!”, but also by saying, equally ironically,
(2)
That man is a lion, of course!
But this last utterance would be not only ironical but also metaphorical, as would be any
other utterance in which a human being would be presented as belonging to a class of
non-human beings. In these and in numerous other cases of the same kind, then, Searle’s
account would be forced to translate the obviously correct observation that the same
utterance can be simultaneously an instance of metaphor and an instance of irony into
the incoherent claim that what the speaker of that utterance means both is and is not
opposite to what the sentence he utters means. Consequently, Searle’s proposed way of
distinguishing ironies from meta-phors cannot be maintained, even if we grant that the
latter are, like the former, manifestations of speaker meaning rather than of sentence
meaning (an assumption that the examples just considered already make suspect, since
4
they suggest that metaphorical interpretations are context-free in a way in which ironical
interpretations are not: one needs quite specific information about the context in order
to determine whether “He is a rock!” or “That man is a lion!” are meant ironically or
non-ironically; but no such information is required in order to determine that they are
meant metaphorically, and it is, indeed, hard to think of a context in which the metaphors
they contain would have not been identified and in which they would still be counted as
acceptable utterances).
Let us now turn to Searle’s proposed elucidation of the difference between
figurativity and indirection. On his account, what distinguishes an utterer who speaks
figuratively from an utterer who speaks indirectly is that, although they both mean
certain things that are different from those meant by the sentences they utter, the former
does not mean at all the thing meant by the sentence he utters, whereas the latter means
in addition the thing meant by the sentence he utters. If Searle’s way of drawing this
distinction was correct, then, there would not exist any utterers that one could properly
describe as simultaneously speaking figuratively and indirectly, since that description
would entail that they simultaneously mean and do not mean what the sentences they
utter mean. However, there certainly exist utterers whose utterances can properly be
described as being simultaneously instances of figurativity and instances of indirection.
For example, just as I can indirectly suggest to someone that her baby needs a bath by
saying to her “Your baby is full of dirt”, I can indirectly suggest to her that her baby
needs a bath by saying to her,
(3)
Your baby has become a piglet.
But the fact that this last utterance can be the vehicle of an indirect suggestion hardly
prevents it from being a vehicle of metaphor, and it is, in fact, only when its intended
metaphorical interpretation has been established that its indirect force of suggestion can
be properly attributed. Similarly, just as I can indirectly express my admiration for a
certain painting by asking the question, “Isn’t that painting admirable?”, I can indirectly
express my admiration for the same painting by asking the question,
(4)
Isn’t that painting a jewel?
But the fact that the latter utterance is, just like the former, a possible vehicle of
indirection hardly prevents it from being, unlike the former, a vehicle of metaphor, and it
is, once again, only when its intended metaphorical interpretation has been determined
that its indirect force can be properly attributed. Faced with these and numerous other
cases of the same sort, however, Searle would be forced to translate the obviously correct
observation that a certain utterance is at the same time an instance of figurativity and an
instance of indirection into the incoherent claim that its speaker has achieved the
impossible task of having meant and of not having meant at the same time what was
meant by the sentence he has uttered—for, the first of these features is, on Searle’s
account, a necessary feature of every instance of indirection and the second a necessary
feature of every instance of figurativity. It seems, then, that Searle’s proposed way of
distinguishing between figurativity and indirection is no more successful than his
5
proposed way of distinguishing between the two main kinds of figurativity that he takes
to be represented by irony and by metaphor, respectively. Consequently, the two internal
sub-questions that his pragmatic account of metaphor was supposed to be able to resolve
can certainly not be regarded as having been successfully resolved. Let us now proceed to
the external question that any such conception would have to be able to answer in the
affirmative if it were to ensure that it is not fundamentally misconceived: is there any
good reason for thinking that metaphor is a matter of speaker meaning rather than of
sentence meaning—in other words, that it is a pragmatic rather than a semantic
phenomenon?
3. The external question
Searle’s answer to that question appears to be that it is obvious that metaphorical meanings
are speaker-based rather than sentence-based. But the fact that the question whether a
given conveyed meaning is speaker-based rather than sentence-based has obvious
answers in certain cases does not entail that it has an obvious answer in every possible
case, and this, as noted, is especially clear in the case of metaphor, where many people
before and after Searle have regarded as far from obvious the thing that he takes to be
obvious. It would appear, then, that reference to some independently available criterion
would be needed to settle this fundamental question. And, fortunately for Searle, the
Gricean framework within which he sees himself as operating provides a criterion, the
so-called “cancellability test”, which does purport to properly motivate decisions as to
whether a given conveyed meaning is speaker-based or sentence-based. It appears,
however, that Searle has neglected to apply the cancellability test before making his
decisions concerning the analysis of metaphor. For, as I will now argue, that test, applied
to relevant cases, clearly contradicts his (and many others’, including Grice’s own)
assumption that metaphorical meanings are speaker-based rather than sentence-based.
The principle of the cancellability test—whose basic idea comes from Grice (1975,
1989), and which was employed by Searle himself on other occasions (for example in
Searle 1979: 30–57)—is simple and, I think, sound: If a speaker who has uttered a
sentence S, and has been interpreted as having thereby meant that p, can without
linguistic oddity cancel that interpretation of his utterance, then his having meant that p by
S is not part of what S itself means; if, on the other hand, a speaker who has uttered a
sentence S, and has been interpreted as having thereby meant that p, cannot without
linguistic oddity cancel that interpretation of his utterance, then his having meant that p
by S is part of what S itself means. (Thus, if a speaker who has said, “I am a man”, has
been interpreted as having meant, among other things, that he is human, his having
meant that he is human must be taken to be part of what the sentence he has uttered itself
means, since he could not without oddity have said, “I am a man, but that doesn’t mean
that I am human”; on the other hand, if a speaker who has said, “I am a man”, has been
interpreted as having meant, among other things, that he would like to meet a woman,
his having meant that he would like to meet a woman cannot be taken to be part of what
the sentence he has uttered itself means, since he could without oddity have said, “I am a
man, but that doesn’t mean I would like to meet a woman”.)
6
Now, many of the non-metaphorical cases where, for Searle as for many others, a
speaker’s meaning diverges from a sentence’s meaning are easily confirmed to be cases of
such divergence by the cancellability test. Thus, the test shows that, if a speaker who says,
“I am hungry”, is interpreted as asking for food, the interpretation in question cannot
legitimately be supposed to be part of what the sentence he uttered itself means, since the
speaker could without oddity have cancelled that interpretation by producing an
utterance like (5):
(5)
I am hungry—but please don’t give me any food: I am on a diet.
Similarly, the test shows that, if a speaker who says, “Why should I ever divorce my
wife?”, is interpreted as expressing the opinion that there are no reasons why he should
divorce his wife, the interpretation in question cannot legitimately be supposed to be part
of what the sentence he uttered itself means, since the speaker could without oddity have
cancelled that interpretation by producing an utterance like (6):
(6)
Why should I ever divorce my wife?—I don’t mean to suggest that no reasons
could ever be found; I simply want to be told what these reasons are.
What Searle and many others (including Grice) have failed to notice, however, is that,
when applied to metaphorical utterances, the cancellability test gives results that
contradict the thesis that metaphorical meanings are speaker-based rather than sentence-
based. Suppose, to adapt one of Searle’s favourite examples, that a speaker says, “My
wife is a block of ice”, and is interpreted as meaning metaphorically that his wife is, say,
unemotional. If this or any other metaphorical interpretation is not part of what the
sentence the speaker utters means, then the speaker should be able to cancel without
oddity all interpretations of his utterance except the one that is strictly identical to its
literal meaning. Suppose, then, that the speaker attempts to block all metaphorical
interpretations of his utterance, by speaking as if his wife is literally a block of ice and
nothing else—by saying, for example,
(7)
My wife is a block of ice—so, please help yourselves to my wife if you need ice for
your drink
s.
This utterance, I submit, would elicit the one or the other of two types of reaction.
Either it would be immediately rejected as semantically anomalous, or it would be
accepted as semantically well-formed provided that one would have managed to interpret
the speaker as still speaking metaphorically when describing his wife as something that
his interlocutors might choose to add to their drinks. But this means that the attempt to
cancel without oddity every metaphorical interpretation of “My wife is a block of ice”
cannot possibly succeed: either the result will be an utterance that is rejected as
semantically odd, or it will be an utterance that is accepted as semantically not odd
precisely because the metaphor has not been cancelled. And this in turn means that, as far
as the cancellability test is concerned, metaphorical interpretations of uttered sentences
must be supposed to be just functions of what the sentences themselves mean rather than
7
functions of what speakers may choose to mean by uttering those sentences. Suppose, to
take one more, familiar, example, that the sentence “Time is money” receives, on a
particular occasion of utterance, one of its usual metaphorical interpretations. If these
interpretations are not determined by what it means but rather by what its speaker has
chosen to mean by uttering it, then its speaker should be in a position to block without
oddity all metaphorical interpretations of his utterance by going on to speak as if time
was, literally, a kind of money and nothing else—by saying, for example,
(8)
Time is money—so, how much of your time have you got in your bank account?
But this utterance would be either rejected as semantically anomalous or accepted as
semantically well-formed provided that its hearer would have managed to interpret its
speaker as still speaking metaphorically when describing time as something that can be
deposited in a bank account. This means that the result of the attempt to prevent without
oddity the metaphorical interpretation of “Time is money” will be either an utterance
that is odd or an utterance whose metaphorical interpretation has not been prevented.
And since, according to the cancellability test, it is sentence meanings, rather than
speaker meanings, that cannot be prevented without oddity, the test’s verdict must, as in
the previous case, be that the metaphorical meaning of “Time is money” resides in what
it means rather than in what any speaker might have chosen to mean by uttering it.
Since the view that metaphor is a matter of speaker meaning rather than of sentence
meaning can hardly be regarded as obviously correct, since the cancellability test has been
devised precisely in order to help deciding unobvious cases of this kind, and since it is a
test that appears both to rest on sound assumptions and to give the expected results in
cases where the distinction between speaker meaning and sentence meaning is obvious,
the only reasonable interpretation of the above results is the interpretation according to
which they do in fact show what they appear to show—namely, that metaphors can
legitimately be regarded as functions of what sentences themselves mean, rather than as
functions of what speakers choose to mean by uttering them. (And if this is so, of
course, Searle’s previously encountered difficulties in distinguishing metaphorical
meaning from what he regards as other types of cases of speaker meaning should hardly
appear surprising: for if these other cases are cases of speaker meaning whereas
metaphors are not, it is no wonder that attempts to distinguish the former from the latter
as different species of the same genus cannot succeed.)
4. Conclusion
My purpose in this essay has been twofold. On the one hand, I have tried to show that,
even if we assume that metaphorical meanings are functions of what speakers, as
opposed to sentences, mean, Searle’s proposed bases for distinguishing metaphorical
meanings from other kinds of speaker meanings are unreliable, since they lead to
contradictory statements both in those cases in which an utterance can be simultaneously
an instance of metaphor and an instance of irony and in those cases in which an
utterance can be simultaneously an instance of figurativity and an instance of indirection.
8
On the other hand, I have tried to show that Searle’s assumption that metaphors are
functions of what speakers, as opposed to sentences, mean, is incorrect, at least when
viewed in the light of the test that is most widely accepted as affording a reasoned
decision on the question whether a given conveyed meaning is or is not a function of
what a sentence, as opposed to a speaker, means. If my arguments are well taken, then,
they will be of interest both to those who have been led to suppose that the distinction
between speaker meaning and sentence meaning holds the key to the analysis of
metaphor, and (for different reasons) to those who have long suspected that metaphor is
too fundamental a feature of natural languages to admit of a simple pragmatic, as
opposed to a complex semantic, treatment.
3
References
Cohen, L. Jonathan. 1979. The semantics of metaphor. In Andrew Ortony (ed.), Metaphor
and thought, 64–77. Cambridge: Cambridge University Press.
Cohen, L. Jonathan & Avishai Margalit. 1972. The role of inductive reasoning in the
interpretation of metaphor. In Donald Davidson & Gilbert Harman (eds.), Semantics
of natural language, 722–740. Dordrecht: Reidel.
Cooper, David E. 1986. Metaphor. Oxford: Blackwell.
Davidson, Donald. 1978. What metaphors mean. Critical Inquiry 5. 31–47.
Fogelin, Robert J. 1989. Figuratively speaking. New Haven, CT: Yale University Press.
Grice, Paul. 1975. Logic and conversation. In Peter Cole & Jerry L. Morgan (eds.), Syntax
and semantics, Vol. 3: Speech acts, 41–58. New York: Academic Press.
Grice, Paul. 1989. Studies in the way of words. Cambridge, Mass.: Harvard University Press.
Lakoff, George and Mark Johnson. 1980. Metaphors we live by. Chicago, IL: University of
Chicago Press.
MacCormack, Earl R. 1985. A cognitive theory of metaphor. Cambridge, MA: MIT Press.
Ross, J. F. 1981. Portraying analogy. Cambridge: Cambridge University Press.
Searle, John R. 1979. Expression and meaning: Studies in the theory of speech acts. Cambridge:
Cambridge University Press.
Kittay, Eva Fedder. 1987. Metaphor: Its cognitive force and linguistic structure. Oxford: Oxford
University Press.
Vanderveken, Daniel. 1991. Non-literal speech acts and conversational maxims. In
Ernest Lepore & Robert Van Gulick (eds.), John Searle and his critics, 371–384. Oxford:
Blackwell.
3 For important criticisms of aspects of Searle’s account of metaphor other than those examined in this essay, the
reader will profitably consult Cohen (1979) and Cooper (1986). An analysis of metaphor (and of related topics) that is
directly inspired by Searle, and to which therefore the arguments presented here directly apply, is Vanderveken (1991).
These arguments also apply to the pragmatic analysis of metaphor proposed in Fogelin (1989), even though that
analysis is considerably more sophisticated than—and at certain points justly critical of—Searle’s own.
