Against Metaphorical Meaning
Ernie Lepore • Matthew Stone
Ó Springer Science+Business Media B.V. 2010
Abstract The commonplace view about metaphorical
interpretation is that it can be characterized in traditional
semantic and pragmatic terms, thereby assimilating metaphor
to other familiar uses of language. We will reject this
view, and propose in its place the view that, though metaphors
can issue in distinctive cognitive and discourse
effects, they do so without issuing in metaphorical meaning
and truth, and so, without metaphorical communication.
Our inspiration derives from Donald Davidson’s critical
arguments against metaphorical meaning and Richard
Rorty’s exploration of the diverse uses of language. But
unlike these authors we ground our discussion squarely in
distinctions about causal mechanisms in cooperative
activity developed by H.P. Grice and others.
Keywords Coordination Á Dead metaphor Á Hinting Á
Joking Á Metaphor Á Pragmatics Á Semantics Á Simile Á
Speaker meaning
Speakers can use metaphors to make their views plain.
Matt Groening, with (1), presents an unmistakable characterization
of the feelings of love.
1. Love is a snowmobile racing across the tundra and
then suddenly it flips over, pinning you underneath.
At night, the ice weasels come.
Groening’s imagery has an absurd particularity, of course,
yet many readers still resonate to his portrayal of the
exhilaration of a new relationship, and in particular, to the
gnawing, trapped, painful feelings that too often appear
once the exhilaration subsides. Suppose we grant that love
really does follow a trajectory from excitement, to
misadventure, to torment. Is that what Groening means
by (1)? Is he, thereby, using (1) to communicate something
true?
It is tempting to answer these questions in the affirmative.
Once we do, we can then describe metaphorical
interpretation in the traditional semantic and pragmatic
terms, thus, assimilating metaphor to other familiar uses of
language. Moreover, there are a range of familiar theoretical
tools to allow us to implement this view. Perhaps,
metaphors supply new, context-dependent meanings to
words and phrases (Stern 2000); or enrich and adapt our
ordinary meanings to more relevant alternatives (Carston
1991; Bezuidenhout 2001; Recanati 2001). Maybe their
uses carry implicatures beyond literal meaning (Grice
1989; Searle 1979); or let us construct propositions by
distinctive metaphorical thinking, substituting for the
ordinary meaning in further steps of semantic and pragmatic
interpretation (Black 1955; Hills 1997; Camp 2003).
In each of these cases, metaphorical usage carries a
meaning its speaker is trying to communicate.
In this paper, we will reject this commonplace view, and
propose in its place the view that, though metaphors can
issue in distinctive cognitive and discourse effects, they do
so without issuing in metaphorical meaning and truth, and
so, without metaphorical communication. Our inspiration
E. Lepore (&)
Department of Philosophy and Center for Cognitive Science,
Rutgers University, New Brunswick, NJ 08901, USA
e-mail: lepore@ruccs.rutgers.edu
M. Stone
Department of Computer Science and Center for Cognitive
Science, Rutgers University, New Brunswick, NJ 08901, USA
e-mail: mdstone@cs.rutgers.edu
123
Topoi
DOI 10.1007/s11245-009-9076-1
derives from Davidson’s critical arguments against metaphorical
meaning (1979) and Rorty’s exploration of the
diverse uses of language (1987). But unlike these authors
we ground our discussion squarely in distinctions about
causal mechanisms in cooperative activity developed by
Grice (1957, 1969) and others.
Language figures in many cooperative activities where
speakers are understood only when their intentions are
recognized, but not all such exchanges constitute communication,
at least not as conceived by Grice, and so, not all
of them involve speaker meaning: for example, our practices
of hinting, joking, trash-talking, flirting, and flattering,
among others, involve crucially linguistic behavior; each
normally includes linguistic communication as traditionally
understood, but essential aspects of these practices
cannot be reduced to, or characterized in terms of, the
communicative enterprise. In what follows, we will elaborate
on these vague claims as well as argue that metaphor
is best catalogued along with these practices and not with
communicated meaning. This position on metaphor is
compatible with a wide variety of positive proposals about
the processes of metaphorical understanding and their
cognitive effects, any one of which affords appealing
responses to commonly-aired criticisms of the Davidsonian
idea that metaphors lack meaning.
1 Metaphor and Communication
We are unaware of any argument for or against metaphorical
meaning and truth that includes a discussion of
speaker meaning.1
This is surprising, since anyone who
claims metaphorical usage carries meaning, in one manner
or another, say, to underwrite assertion, truth, disagreement,
or interpretation, must be assuming speakers who
employ metaphors mean something by them, something
that exceeds their literal meaning. Yet the standard progression
from Aristotle forward has been to explain how to
assimilate metaphorical meaning to standard semantic and/
or pragmatic accounts and mechanisms. All such efforts are
premature until it has been established that whatever point
a metaphorist intends to make with his usage rises to the
level of speaker meaning. To discern whether it does, we
will adopt the familiar perspective that speaker meaning is
a species of intention in action.
The connection between speaker meaning and intention
traces back to the landmark discussion of Grice (1957).
What distinguishes speaker meaning from other intentions
an agent might have is the kinds of effects his action aims
for, and how these effects are to be achieved. Grice
describes the distinctive effects of meaning in terms of
eliciting a decision in audience members that is under their
control. As canonical examples, Grice cites inducing an
audience to adopt a belief, in the case of an assertion, and
inducing an audience to commit to an action, in the case of
an imperative. In these cases, if p is the content of the
belief that the audience adopts, or a description of the
action that the audience commits to, then we can say p
gives the intended effect the speaker is aiming for.
Grice captures the distinctive process of communication
by requiring that any intended effects be elicited in part by
an audience’s recognition of the speaker’s intentions. In
other words, in order for a speaker’s intention to count as
what his utterance means, this intention must envisage that
its recognition will provide an audience with the decisive
reason to commit to the very effect the speaker is aiming
for. To take a concrete illustration, in order for a speaker to
mean that it is raining with an utterance of ‘‘It is raining,’’
he must intend that his audience believe that it is raining;
and further that they recognize that the speaker has this
intention that they should believe that it is raining; and
finally, that it be by means of this very recognition that they
actually come to believe that it is raining. Generalizing,
then, Grice’s explication of speaker meaning comes to (2):
2. In order for a speaker S to (speaker) mean p with
an utterance u is for S to produce u with the
intention to elicit the effect given by p in an
audience by means of their recognizing that this is
his intention (cf. Grice 1989, p. 220).
(2) locates speaker meaning squarely in how an agent’s
actions impact on others. When an audience considers
meaning in utterances, they are, according to Grice, asking
what was the speaker aiming for and how did he envision it
would transpire.
Grice’s aim, in part, is to distinguish meaningful signaling
from other ways in which we share information
with one another. On the one hand, Grice is concerned
with explaining (away) cases where actions clearly provide
us with information but where we fail to locate any
speaker meaning in them. To take Grice’s vivid example,
Herod’s action of presenting Salome with the head of
John the Baptist clearly informs her that John is dead; but
it would be a mistake to conclude with his action Herod
means that John is dead; his act of showing her the head
doesn’t signal this information; it’s rather seeing John’s
severed head that leads her to believe he is dead. It is true
Herod intends Salome to learn that John was dead, and
she knows this, but, again, she learns he is dead from
seeing his severed head, not from recognizing Herod’s
intention.
1
Searle (1979), of course, in his well-known discussion of metaphor
just assumes metaphorical interpretations are speaker meant but he
offers no argument for this assumption other than his arguments that
the few other accounts he considers are flawed.
E. Lepore, M. Stone
123
This doesn’t mean language is essential to meaning.
Grice sees meaning in language as part of a larger phenomenon
of communication, and not just something that
depends distinctively on the use of linguistic forms or on
the existence of linguistic conventions or grammar. (Grice
would actually prefer to reduce our knowledge of meaning
in language to our knowledge of psychology: what a sentence
means, in his view, is just what speakers generally
mean when they use an utterance of the sentence.) This is
why his analysis is formulated in terms of fundamental
notions from collaboration and not from linguistic theory.
For whenever we participate in joint activities, we intend
our contributions to be recognized; often we go further and
rely on specific processes and mechanisms in doing
our part.
Information can be contributed in ways besides through
language, and many of these ways depend on just the sort
of coordination Grice associates with intention recognition.
Grice cites as an example the information retrieved from a
bell on a tram. When the conductor rings it three times, this
means the tram is full. Lewis (1969) considers the signals
from the ride of Paul Revere. When the sexton of the Old
North Church hangs one lantern in the tower, this means
the Redcoats are coming by land. In both cases, an agent
intends to convey information to an audience; and in both,
the audience must recognize the relevant intention in order
to tap into the information the agent is contributing.
In this paper, we build on the distinctions first targeted
by Grice, and evaluate whether metaphor is best classified
in this light as generating speaker meaning. The details as
Grice lays them out in (2) are tendentious; the literature
offers a number of difficult cases. This requires us to clarify
the status of these alleged counter-examples before we can
settle questions about metaphorical meaning sharply.
Broadly speaking, three responses are possible. Grice
might be right: properly understood, the notions of deciding
on an effect and acting on the basis of a recognized
intention dissolve these alleged counter-examples. If this is
your view, you can push our arguments through with
Grice’s original characterization (2). Alternatively, you
might think Grice’s view needs tweaking. Grice himself
issued a range of emendations and qualifications up to the
William James Lectures, published as (1989, p. 114),
giving secondary conditions on the kinds of intentions that
count as issuing in speaker meaning and the way in which
those intentions must be recognized and used. His final
account involves some six or seven bullet items, and takes
almost a page to state; it is not particularly natural. Nevertheless,
if you endorse it, you will be able to push our
arguments through. And, lastly, you might want to precisify
(2) invoking specialized constructs from pragmatics; in
particular, following Lewis (1969, 1979) and Thomason
(1990), by giving explicit and direct theoretical
characterizations of the notions of coordination and of
contributing information to an interaction. These constructs
don’t necessarily make for a simple or elegant philosophical
analysis, and they undercut some of the metaphysical
aims under which Grice proposed (2). For that reason, they
may prove more congenial to linguists or psychologists
interested in the status of metaphor in the science of
communication. But if you agree with them, that will
enable you to push our position on metaphor through. To
see what we have in mind, consider a counter-example to
(2) that concerns the aims of meaning.
When a student answers a question on an oral exam, she
surely means what she says. But how can we square this
intuition with Grice’s analysis? If her answer is correct,
mustn’t the examiner already have this information? But if
her aim is for the examiner to credit her with a correct
answer, isn’t his response a function of the standards of the
exam, not of his recognition of her intention to be judged
correct? How does her answer contribute this information
to the interaction? Such cases highlight a fundamental
difficulty in characterizing the effects of meaning concretely
in terms of decision or belief, or even, following
Grice (1969), generally, in terms of the production of a
suitable response by the audience. In this regard Lewis’s
(1979) account fares better. He recommends that utterances
be characterized as providing updates to the conversational
record, an abstract ‘scoreboard’ that tracks interlocutors’
contributions to their interaction.
Speaker meaning in this framework still involves an
intention, one to update the conversational record. This
record describes actions taken overtly by the interlocutors as
contributions to the interaction as well as the evident, public
effects those actions engender. This permits us to think
precisely about what goes on the record, and to connect the
conversational record to the mental states of interlocutors.
Normally, when interlocutors are well-informed and attentive,
they track the record, and so, come to mutually suppose,
for the purposes of the conversation, that things are as
recorded (Stalnaker 1978; Thomason 1990).2
Since invoking the conversational record gives an
independent characterization of the effects speaker meaning
can aim for, we can solve the puzzles about the presence
and limits of meaning in cases such as examinations.
2
Stalnaker and Thomason in fact propose to reduce the conversational
record to the mutual attitudes of interlocutors. However, one
can also locate the conversational record by reference to the norms
which interlocutors subscribe to in participating in conversation; this
lets us narrate public actions as taking the same effects on the
scoreboard even in defective cases where interlocutors may fail to
understand one another, because they are ignorant about the facts of
meaning or must work through temporary misunderstanding or
ambiguity (Steedman and Stone 2006; Lepore and M. Stone 2007;
DeVault 2008).
Against Metaphorical Meaning
123
In that case, the conversational record for the examiner and
his student simply tracks questions asked and responses
given; in this framework, the interlocutors’ activity begins
with a blank slate, not with shared background knowledge
and present beliefs. The student’s utterances are intended to
register specific content on the record; each utterance seeks
publicly to establish a specific proposition as the official
answer to a stated question. Establishing an answer in this
sense arises through the pure coordination of student and
examiner, through the kind of causality Grice’s definition
aims to pinpoint. Putting the answer on the record is separate
from any auxiliary responses the examiner must make
(and typically, must register publicly in turn, through a
suitable overt acknowledgment) to judge the answer and
evaluate the student.
A complementary refinement involves Grice’s account
of the processes involved in meaning making. Difficult
cases are not distinguished sharply by the kinds of information
or reasoning that interlocutors use to understand
one another. They are more clearly distinguished by the
circumstances that prevail in the world, and set the stage
for interlocutors’ joint activity. Accordingly, we may
choose to be more precise about the processes involved in
speaker meaning not just in terms of intention recognition,
or the interrelated attitudes of interlocutors, as in (2) and
preserved through Grice (1969), but directly in terms of the
kinds of situations where agents face signaling problems,
following Lewis (1969). The key notion is coordination.
Coordination problems are situations where multiple
agents must match their behavior to one another, and
multiple ways of doing this are possible. To act successfully,
agents must use their expectations about one another
to select one solution to the exclusion of others.
Updating the conversational record, so as to register a
specific proposition as part of the presumptions of the
interaction, is a coordination problem in Lewis’s sense.
There is no inherent connection between utterances and
contents, so a speaker could use any utterance to register
any proposition, so long as her audience used a corresponding
strategy to track the propositions so contributed.
Interlocutors solve the coordination problem by adhering to
what Lewis calls a convention, that is, a particular scheme,
implicit in interlocutors’ mutual expectations about how
they will speak and how they will understand, for aligning
their efforts to track the information each contributes.
More precisely, with acting in a coordination problem,
each agent adheres to a strategy that maps situations to
chosen actions and desired outcomes. This strategy is, in
effect, an intention. In fact, it is one possible intention
among many. Other agents respond according to their
expectations about the strategy their partners adhere to.
That is, they recognize their partners’ intentions and choose
a response that fits what their partner has manifested—in a
situation where multiple alternatives are possible. Equilibrium
demands that neither agent would deviate from his
decision making given more information about his partner’s
strategy. Focusing on coordination problems in
Lewis’s sense thus gives a rationale for Grice’s analysis of
meaning in terms of intention. Nevertheless, characterizing
coordination problems directly allows us to see meaning as
arising fundamentally from the general kind of decision
making interlocutors are doing, and general properties of
circumstances in which interlocutors act, not—as in Grice’s
definition—from the specific psychological processes
that lead interlocutors to particular decisions in particular
cases. This gives much sharper grounds to delimit coordination,
and, in turn, meaning.
To concretize these various ideas, consider this case
from Lewis. It is a convention that we drive on the right
hand side of the road, or indeed, that we walk on the right
hand side in a narrow, busy corridor. In choosing to go
right, we are acting on our expectations of how others will
go. We think they will also go right. If we thought they
would go left, we would go left too; we wish to avoid
congestion or collision. The coordinated expectations that
send us all to the right—as we do in the US—do not refer
to anything in the world that settles our preferences or
strategies for us. There is, in fact, another equilibrium. We
could use the left—as in Britain or Japan. Either strategy
serves equally well, so long as all agree.
Institutions and circumstances can break the symmetry
which constitutes coordination. This changes the status of
our decisions. In the US, there are sanctions for driving on
the wrong side of the road. Lewis judges—and we agree—
that we drive how we do because of our expectations about
what others will do, given that we wish to avoid accidents.
We do not drive on the right to avoid tickets. But if, in fact,
we acted based on a pure preference to drive on the right,
irrespective of what others did, because of the sanctions in
force, then even if that strategy happened to result in an
efficient solution to the problems of joint action, we would
not have a problem of coordination, or a convention, in
Lewis’s sense. We would act efficiently, but independently.
To take a less extreme example, many long, narrow busy
corridors are fitted with two moving sidewalks, one headed
toward each end of the corridor. We have a preference to
take the sidewalk that goes in our direction of travel. Those
with the opposite direction have the opposite preference.
We all avoid collision and congestion by riding our
respective sidewalks, but we do not coordinate. The world
settles our preferences and strategies for us. In anticipating
others’ strategies we need only assume that they act in their
interests, irrespective of any consideration they give to
their interactions with us, and their expectations about us.
They can reason likewise. In thinking about such cases, it
may be easier to apply Lewis’s notion of coordination to
E. Lepore, M. Stone
123
the circumstances that prevail in the world, rather than to
work with Grice’s definition (2) and try to judge directly
whether agents act on the basis of recognizing one
another’s intentions.
Applied to language, paralleling the example above in
discussing Grice, in order that a speaker mean that it is
raining with an utterance of ‘‘It is raining,’’ roughly, she
must intend to contribute that it is raining by coordination.
This requires her audience and she to add that it is raining
to the conversational record in order to satisfy their joint
interests, including their interest in agreeing on the record;
and that her audience will do so in virtue of their mutual
expectation that if she says ‘‘It is raining,’’ then both parties
agree on the record only if they all add the information that
it is raining to it; and lastly, had different expectations
prevailed and had she followed a different strategy, she
could have instead done something else, thereby leading
her and her audience to add that it is raining to the record,
and that would thereby have served their joint interests just
as well. Generalizing, we can summarize with (3):
3. For a speaker S to mean p with an utterance u is for
S to utter u with the intention of coordinating with
an audience to add p to their conversational record.
The key difference between (2) and (3) is that instead of
looking to analyze when and how recognition forms the
basis for a belief, this account looks to what it means to add
something to the conversational record and what kinds of
circumstances and outcomes describe interlocutors’
activities.
We now have a sense of what’s involved in applying
Grice’s analysis of speaker meaning as intention more
precisely and systematically. With this in place, we can
revisit the key virtue of definitions like (2) and (3) for the
present argument. These definitions of speaker meaning
enable us to see why we do not take everything we can
infer from someone’s utterance as speaker meant.3
Two
specific kinds of cases are worth highlighting: first, cases
where interlocutors contribute information (by establishing
a suitable belief or updating the record), but do not coordinate;
and second, cases where interlocutors do coordinate
(by recognizing one anothers’ intentions or drawing
appropriately on mutual expectations), but their coordination
does not aim to convey information.
First, we must distinguish what Stalnaker (1978) calls
‘‘secondary effects’’ of utterances from speaker meaning.
We can learn much from utterances beyond what they
mean. Here is a simple case: when a speaker utters u, we
learn that she is the speaker of u. But she does not speaker
mean that she is the speaker of u. The information clearly
becomes public on the basis of the utterance itself, and so
the audience comes to believe it; it goes on the record.
Interlocutors must track this information; interlocutors
need it to interpret, for example, tokens of the word I,
which refer to the speaker of the utterance. But interlocutors
do not coordinate on who the speaker of u is. When u
occurs, facts in the world, not intention recognition or
interlocutors’ expectations about one another, determine
who said u.
Similar considerations establish that when a speaker
utters u, we learn that u was made, but the speaker does not
speaker mean that u was made. Interlocutors register what
sounds have been produced, and how those sounds are
organized into words and phrases. This is clearly public
information interlocutors must track; they need it to
interpret clarifications that refer to what has been said, to
interpret locutions like the former and the latter that evoke
expressions used.4
But interlocutors do not coordinate the
exchange of this information. Interlocutors may coordinate
on the performances they use to realize specific words: as
with tomato and tomahto, different speakers, by convention,
say the same word in different ways. But, given that
the word has been used, there is no further problem of
coordination about how interlocutors will take stock of the
fact that the word has been used. The speaker has said
something specific, and the fact of this utterance is manifest,
so the audience comes to believe that the utterance has
been made; it goes on the record. But no intention recognition
or mutual expectations are involved. It is not a
convention that you register this utterance, instead of
another. Inevitably, a record of what has happened can
include only what has happened.
This forces our attention to a second kind of case—
where interlocutors coordinate on actions but not on
information. Consider Alice and Bob, who are shopping for
a dinner that they will cook together. They agreed that
Alice will buy food and Bob wine. They have a coordination
problem. They will be pleased with many different
main courses, and many different bottles, but the best
outcome requires that the meal and the wine match. This
time, Alice chooses to buy a steak, rather than the salmon,
which looks very good too. This is a public contribution to
their activity; it goes on the record that Ann has bought the
steak; and Bob knows it. It follows that Bob must buy red
wine, rather than white. Again, this is because of the state
of the world and Alice and Bob’s preferences about dinner,
not because of a signal from Alice. Bob’s reason for buying
red wine is that he knows Alice has bought steak, not that
he recognizes Alice’s intention for what he should do. Bob
3
This was, of course, a large part of Grice’s motivation in
introducing (2).
4
And even, if Stalnaker (1998) is right, to follow the referential
mechanisms that underpin the coherence of discourse generally.
Against Metaphorical Meaning
123
in fact chooses a fine Bordeaux, so Alice and Bob solve
their coordination problem.
The structure of this interaction highlights the difference
between solving coordination problems and making
meaning. Alice does not mean that Bob should buy red
wine by her purchase of steak. She is only doing her part
for dinner. Her purchase figures in the solution of their
coordination problem and has real-world consequences,
including updating the state of Alice and Bob’s joint
activity. But Alice and Bob are not coordinating on what
information should characterize the state of their joint
activity, given what they have done: and only coordinating
on adding information to the record underwrites attributions
of meaning.
In fact, our utterances often contribute to realizing our
practical projects, just like any other instrumental action.
Thus, following Austin (1962), we distinguish the illocutionary
effects our utterances conventionally engender
from the perlocutionary effects they bring about through
other causal mechanisms. ‘‘Saying something will often, or
even normally, produce certain effects upon the feelings,
thoughts, or actions of the audience, or of the speaker, or of
other persons’’ Austin writes, ‘‘and it may be done with the
design, intention, or purpose of producing them’’ (1962, p.
101). Austin’s familiar examples include convincing, persuading,
deterring, alarming, or surprising the audience.
Such effects depend not only on the content that utterances
contribute to the conversation, but also on the circumstances
that obtain, on our preferences for the future, and
on the flow of our ideas and emotions. We do not coordinate
on these broader mechanisms. They are general and in
no way conventional. Austin highlights this generality: ‘‘It
is characteristic of perlocutionary acts that the response
achieved, or the sequel, can be achieved additionally or
entirely by non-locutionary means: thus intimidation
may be achieved by waving a stick or pointing a gun’’
(1962, p. 119).
We may coordinate on perlocutionary effects, of course,
just as we coordinate on any other outcome we value. You
may be working to convince me, and I may be working to
be convinced. But this is a change in me, occasioned by the
make-up of my psychology; it is not information contributed
to our interaction through our coordination. Your
argument does not mean that you convince me, even
if you do.
We can now see what’s at stake in positing metaphorical
meaning. A speaker successfully communicates p with an
utterance only if he intentionally elicits an effect given by p
in his audience by means of recognizing the speaker’s
intention to elicit this effect; or if the utterance contributes
to coordinating with the audience to add p to their conversational
record. If metaphors contribute content, then,
this content must be identified either with what a speaker
intends her audience to recognize her as trying to communicate;
or as what must be added to the conversational
record in consequence of coordination between her and her
audience.
Metaphorical meaning cannot be merely content ‘gotten
across’ nor can it be merely behaving recognizably as
coordinating. It requires an audience to recognize the
specific content a speaker wants to get across, and to use
the signal of the metaphor as the basis for the uptake of that
content. Since we deny this must happen in normal confrontations
with metaphors, we therefore reject metaphorical
meaning.5
We are not denying that metaphors can be used with the
intention of drawing a hearer’s attention to similarities, but
this is not an intention to convey propositional content. We
concede that a metaphor succeeds, if it does, from the
cognitive effort an audience puts into exploring the similarities
suggested by the metaphorical imagery, but we
deny that this derives from recognizing a speaker’s intention
to convey propositional content.6
Taking this perspective
allows us to afford a specific place for metaphor in
the architecture of cooperative interaction without ascribing
any meaning to a metaphor or its uses beyond the
literal.
How are we to take stock of this proposal? We have thus
far considered several cases whereby utterances bring
about effects that are not part of what the utterances or their
creators mean. None of these cases provides a close analogue
to what happens in metaphor. The so-called secondary
effects of utterances are a poor fit. When a speaker
uses an English utterance, his audience learns that the
utterance has been made; they learn that the speaker speaks
English. Their inferences track what they know must be so
when a speaker uses certain words. But appreciating a
metaphor represents a different kind of inference. If we
take away from (1) that love follows a trajectory from
excitement, to misadventure, to torment, it’s not because
we think love must follow this trajectory for Groening to
have formulated and produced his utterance.
Familiar cases of perlocutionary effects also provide a
poor fit. When a speaker’s utterance intimidates the audience,
it’s because its content triggers certain thoughts and
emotions in the audience. The psychological dynamics run
their course. The audience need not recognize the speaker’s
intention to intimidate—if the speaker even has such an
intention—nor must the audience participate in any special
5
Thanks to Zoltan Szabo for emphasizing the importance of this
distinction to us.
6
To say metaphors make us see similarities is not to say metaphorical
content is a simile, though we may be able to describe the effects
of a metaphor with a simile. But of course this no more requires the
metaphor to mean that one thing is like another than Herod’s showing
John heads requires us to have his act speaker mean that John is dead.
E. Lepore, M. Stone
123
way in calling up the thoughts and feelings involved, in
bringing the effects about. By contrast, we cannot appreciate
the suggestions of (1)—e.g. that love begins its trajectory
with excitement—merely by submitting the
proposition that love is a snowmobile to our ordinary faculty
of inference, critical thinking, and deliberation. Love
is not a snowmobile! To suggest otherwise is absurd at face
value. Reasoning normally from (1) would just rediscover
the absurdity. We must recognize that (1) is a metaphor,
and shape our psychological response accordingly.
Other sorts of acts, however, upon reflection do offer
closer parallels with metaphor. By drawing these parallels
explicitly, we hope to clarify the status we are ascribing to
metaphor.
Davidson likens speaking metaphorically to joking.
Consider the conversation in (4), as reported by Lampert
and Ervin-Tripp (2006, p. 54).
4. Friends talk about Diet Pepsi.
Vi: There is a problem with aspartame, but I can’t
remember what it is
Joe: It affects your memory
Vi: Hah! That’s a good one
As Vi and Joe spoke, Vi was drinking a Diet Pepsi. The
drink is sweetened with aspartame. Vi takes Joe to be
joking here. The sentence that he uses clearly means that
aspartame affects memory, but his point in using it is not to
contribute this information seriously to their interaction. In
fact, despite Joe’s utterance, it seems he makes no
commitment whatsoever, one way or another, as to whether
aspartame affects memory, or even whether it is associated
with any other health problems.
What gives Joe’s utterance its humor is the imagery
behind it. The joke conjures up the imaginative possibility
that Vi is already suffering from the ills of aspartame, but
hasn’t yet noticed the symptoms or their connection to the
beverages she drinks. The effect is to ridicule whatever
worries might lie behind Vi’s broaching the topic of the
problems of aspartame. Of course, in order to get Joe’s
joke, Vi must recognize that he is offering his utterance
merely to develop the imagery behind the joke, not to
contribute this information. She must participate in
exploring the consequences of this imagery to discover that
it portrays her as suffering from the ills of aspartame, and
so, to discover that this joke offers Joe a way of chiding her
for worrying about it. Thus, the joke requires special
cognitive effort from Vi, prompted in part by her recognition
of its status as a joke.
It seems clear that the effects that Joe achieves with (4)
do not fit the account of speaker meaning from (2) or (3).
Joe’s goal here seems to be to ridicule Vi, not to assert that
(he thinks) she is ridiculous. This outcome—creating a
specific kind of interaction between people—is simply the
wrong kind of effect for speaker meaning. Moreover,
intention recognition plays a very limited role in the elicitation
of this effect. It ends once Vi has recognized Joe’s
utterance as a joke. From then on, the humor in the utterance
is determined by the imaginative possibilities it portrays, not
by Joe’s intentions, Joe and Vi’s mutual expectations, or the
rules of language. These considerations seem so overpowering
that we think few people would even be tempted to say
that Joe’s joke has a special, non-literal meaning, to the
effect that Vi’s concern with aspartame is laughable. If
speakers mean anything by their jokes, they mean nothing
other than what their words mean.
The situation is exactly parallel, we claim, with metaphor.
Groening’s metaphor conjures up the imaginative
possibility that love is a snowmobile. To get its point, a
reader must recognize Groening is offering his metaphor
merely to develop the imagery behind it, and of course, not
to contribute its literal meaning to the conversational
record. On confronting the metaphor, an audience must
participate in exploring the implications of its imagery.
That effort may lead the audience to appreciate certain
similarities between love and a tragic snowmobile ride:
exhilaration, misadventure and torment. The goal of the
utterance is for this appreciation to occur, not for specific
information to be exchanged, and interlocutors do not
coordinate on the information itself or derive it directly by
intention recognition. Thus, even when a metaphor succeeds,
it does not result in special, non-literal meaning. The
parallel between metaphors and jokes, and the obvious
appeal of the explanation as it applies to jokes, adds to the
plausibility of our view of metaphor.
Analogously, consider hinting. Here is an instructive if
not subtle example. A student brings a completed exam to
the examiner’s desk at the final for a small, friendly class.
The examiner looks its over, finds an error and responds
with (5).
5. You might want look at your answer to problem
three again.
Under normal circumstances, an examiner would say (5)
only because the student’s answer to problem three is
wrong. The student knows this. In fact, in this case, the
examiner intends that the student revisit her answer, notice
her error and correct it. The student knows this as well. But
from this it does not follow that the examiner speaker
means by his utterance of (5) that the student’s answer to
problem three is wrong. Indeed, if the examiner intends
that the student should discover the error, then both the
analyses in (2) and (3) predict that this is not a case of
speaker meaning. The student finds that the answer is
wrong by looking at her paper, not by recognizing the
examiner’s intentions.
Against Metaphorical Meaning
123
This is a coordination problem. Both student and
examiner prefer the student to fix her answer, and so, do
better on the exam. It is likely the examiner has many ways
to clue his student in, and the student has many strategies
for pinpointing and fixing her error. But the relevant
coordination, like Alice and Bob’s over dinner, involves
action in the world—in this case, particularly, the cognitive
and problem-solving actions undertaken by the student.
In interpreting (5), the student must recognize the
examiner’s intention and act based on that recognition.
Plausibly, this is an indirect instruction, and by (5), the
examiner does mean for the student to revisit problem
three. Ultimately though, the student discovers the error
on her own. So, the theory articulated in (2) or (3) predicts
that the examiner does not mean with an utterance
of (5) that problem three is wrong. He merely hints it is
wrong.
Most hints are not as explicit as (5) in what they convey.
The speaker prompts the hearer to reconsider some question.
It really is up to the hearer what she learns from this
inquiry. This gives most hints the open-ended texture found
also in metaphor. If one has a specific message to impart, a
hint is often a poor way to get it across. D. Stone et al.
(1999) make this point particularly effectively, using
example (6), spoken by a wife to a husband who has taken
up golfing every Saturday morning.
6. Is golf so important that you need to play it this
often?
They write:
This is a classic example of a statement masquerading
as a question. It’s obvious that the meaning of
the comment is conveyed in the subtext. What is
less obvious is what the meaning is supposed to be.
Your tone conveys anger or frustration. But it’s not
clear what is causing the anger or what your husband
is supposed to do about it. Are you angry that
your husband is engaged in a meaningless sport
rather than community service or household chores?
Are you angry that he’s not taking you along? Are
you angry that you aren’t spending enough time
together? How would he know? (D. Stone et al.
1999, p. 192).
Ultimately, in hinting, a speaker does not achieve speaker
meaning. A speaker invites his audience to participate in a
certain kind of inference: thinking an issue through again,
in light of new evidence that other views might be possible.
This process may lead the audience to notice new things; it
is the speaker’s aim that such discoveries occur, but they
are outside the realm of the speaker’s control—and indeed
that of the audience. The audience makes the discoveries
that they are led to by the critical thinking they do. They do
not make these discoveries merely in virtue of recognizing
their interlocutors’ intentions that they do so.7
We draw the same conclusion about the metaphorist.
While the point of Romeo’s line ‘Juliet is the sun’ might be
to prompt a reader to think of Juliet as radiant and centered,
there should be no temptation to claim the metaphor has a
‘‘special’’ meaning to this effect. Similarly, it is a mistake
7
Some of what Grice called conversational implicatures turn out to
be more like hinting than speaker meaning. He introduces the term
‘‘implicature’’ neutrally:
A asks B how C is getting on in his job, and B replies, ‘Oh
quite well, I think; he likes his colleagues, and he hasn’t been
to prison yet.’ At this point, A might well inquire what B was
implying, what he was suggesting, or even what he meant by
saying that C had not yet been to prison. The answer might be
any one of such things as that C is the sort of person likely to
yield to the temptation provided by the occupation, that C’s
colleagues are really very unpleasant and treacherous people,
and so forth. It might, of course, be quite unnecessary for A to
make such an inquiry of B, the answer to it being, in the
context, clear in advance. It is clear that whatever B implied,
suggested, meant in this example, is distinct from what B said,
which was simply that C had not been to prison yet. I wish to
introduce, as terms of art the verb ‘implicate’ and the related
nouns ‘implicature’ (cf. ‘implying’) and ‘implicatum’ (cf.
‘what is implied’). The point of this maneuver is to avoid
having, on each occasion, to choose between this or that
member of the family of verbs for which ‘‘implicate’ is to do
general duty (1989, p. 24)
Grice has framed this discussion explicitly to avoid the question of
whether implicatures are (speaker) meant. It’s reasonable to conclude
he thought many or even most implicatures were, but not all. Later,
in considering irony, he appeals to considerations that seem to
distance the interpretation of irony from the factors that go into his
typical discussions of speaker meaning.
To be ironical is, among other things, to pretend (as the etymology
suggests), and while one wants the pretense to be
recognized as such, to announce it as a pretense would spoil the
effect. What is possibly more important, it might well be
essential to an element’s having conventional significance that
it could have been the case that some quite different element
should have fulfilled the same semantic purpose; that if a
contemptuous tone does conventionally signify in context that
a remark is to be taken in reverse, then it might have been, for
example, that a querulous tone should have been used (instead)
for the same purpose. But the connection of irony with the
expression of feeling seems to preclude this; if speaking
ironically has to be, or at least appear to be, the expression of a
certain feeling or attitude, then a tone suitable to such a feeling
or attitude seems to be mandatory, at least for the unsophisticated
examples (1989, p. 54)
Thomason (1990) says ‘‘most implicatures are meant’’. He goes on to
explicitly consider the relationship between implicatures and Grice’s
definitions, to show that many implicatures come out as meant. We
endorse this understanding of implicatures and emphasize that our
arguments do not count against the meaningfulness of most implicatures,
particularly Grice’s Group A implicatures, where no maxim
is flouted. This category includes most bread-and-butter examples of
relevance and quantity implicatures: those associated with inferring
coherent interpretations in discourse.
E. Lepore, M. Stone
123
to claim that when Auden writes, ‘‘The hourglass whispers
to the lion’s paw,’’ what he thereby says has a ‘‘special
(non-literal) meaning’’—that time, like the steady flow of
sand in an hourglass, is subtle and unrelenting. There is no
such meaning. In short: the point a metaphorist uses a
metaphor to make is not a property of the metaphor itself.
2 Criticisms of Metaphorical Meaning
This ends our outline of the positive proposal; before
advancing it further, we need to see what’s wrong with its
alternatives. Every account of metaphorical meaning we
are aware of has been criticized in one manner or another,
but nothing like the criticism launched by Davidson. His
central claim in ‘‘What Metaphors Mean’’ (1979) is that
metaphors have only literal meaning. He does not question
their potential effectiveness; only how these effects might
be achieved. Davidson goes further than most critics in
denying that there is a determinate propositional thought a
speaker can reasonably be taken to mean by his use of a
metaphor. A metaphor might ‘‘make us attend to some
likeness, often a novel or surprising likeness, between two
or more things;’’ it might make us ‘‘see one thing as
another,’’8
but this doesn’t distinguish metaphor from a
bump on the head or taking a drug: each can cause certain
effects, like noticing surprising similarities. What he denies
is that the metaphor, like the bump and the drug, means
these effects.
For Davidson, what metaphors convey is largely
non-propositional, unpredictable causal effects of their
utterance-events understood exclusively according to their
literal meanings. This doesn’t mean that when Shakespeare
wrote ‘‘Juliet is the sun,’’ he had nothing in mind. Davidson
is denying that whatever he had in mind must be part of a
metaphorical interpretation his audience is expected to
retrieve. He denies that, associated with his (or any other
metaphor), there is a content every successful interpreter
must grasp (Davidson 1979, p. 43). His conclusion is based
on several arguments; given the nature of speaker meaning
and our positive proposal for what metaphors do, it will be
useful to rehearse some of these criticisms here.
2.1 Ineffability of Metaphorical Content and Rejoinder
Davidson writes:
If a metaphor has a special cognitive content, why
should it be so difficult or impossible to set it out? If,
as Owen Barfield claims, a metaphor ‘‘says one thing
and means another,’’ why should it be that when we
try to get explicit about what it means, the effect is
so much weaker—‘‘put it that way,’’ Barfield says,
‘‘and nearly all the tanning, and with it half the
poetry, is lost.’’ Why does Black think a literal
paraphrase ‘‘inevitably says too much—and with the
wrong emphasis?’’ Why inevitably? Can’t we, if we
are clever enough, come as close as we please?
(p. 42)
Davidson answers his questions:
The reason is …that we imagine there is a content to
be captured when all the while we are in fact focusing
on what the metaphor makes us notice. If what the
metaphor makes us notice were finite in scope and
propositional in nature, this would not in itself make
trouble; we would simply project the content the
metaphor brought to mind onto the metaphor. But in
fact there is no limit to what a metaphor calls to our
attention, and much of what we are caused to notice
is not propositional in character…Since in most cases
what the metaphor prompts or inspires is not entirely,
or even at all, recognition of some truth or fact, the
attempt to give literal expression to the content of the
metaphor is simply misguided (pp. 44–45)
If a metaphor has a speaker meaning, it should be possible
to express it literally. Otherwise, we are left with the puzzle
of how a speaker could formulate, and an addressee could
recognize, an intention to convey a specific proposition,
without any independent way of specifying exactly what
proposition that was. Such puzzles simply do not arise if
we attribute other kinds of intentions to a metaphorist, such
as an intention to get an audience to notice similarities.
One reason why it’s so difficult to paraphrase the metaphor
into literal language is because it can legitimately
‘say’ different things to different hearers. In this regard,
metaphorical interpretation, for lack of a better phrase, is a
creative open-ended affair. When Matt Groening says
‘‘Love is a snowmobile,’’ it occurred to one of us that the
way cold air takes your breath away is a lot like the shock
of being around somebody you have a crush on. That never
occurred to the other one of us. He was instead thinking of
the speed, the smooth ride, the unfamiliar environment.
Still, the first impression is just as valid as an insight the
metaphor delivers. But Matt Groening’s goal was not to get
us to grasp a deep subtle truth, but rather to see something
in a certain way, and this is not something speaker meant.
As Davidson explains:
…if I show you Wittgenstein’s duck-rabbit, and say,
‘It’s a duck’, then with luck you see it as a duck; if I
say, ‘It’s a rabbit’, you see it as a rabbit. But no
proposition expresses what I have led you to see.
(p. 45)8
Davidson (1979), pp. 31, 45; our emphasis.
Against Metaphorical Meaning
123
The difficulty in devising a literal paraphrase of what a
hearer has been led to see is not that the latter is especially
subtle or profound, but rather, that seeing a figure in this (or
any other) way is not amenable to literal paraphrase
(though of course it can be described).
Some authors, like Moran (1997), deny that all ‘‘cognitive
contents’’ can be construed propositionally. Metaphor
might represent the world non-sententially, and so, are
no more amenable to literal sentential articulation than a
painting of a landscape is; and so, to arrive at an interpretation
of a metaphor we are not required to grasp a
meaning beyond the literal. We needn’t grasp a proposition
to whatever effect the metaphor prompts us to see.
Even if Moran is correct about the ontology of content,
his proposal concedes that a speaker cannot successfully
use metaphors to intentionally elicit an effect given by a
propositional content p in his audience by means of their
recognizing this intention to elicit this effect; nor does it
permit the metaphorical use to contribute to coordinating
with an audience to add the propositional content p to their
conversational record. That is, Moran (and others) concede
that however metaphorical use contributes cognitive content
to a discourse it is not through speaker meaning. Nonsentential
representation—content not grounded in a
proposition—is not of the right shape for this.
2.2 Metaphors and Similes
Davidson’s next argument compares metaphors to similes.
…how is it that a simile gets along without a special
intermediate meaning? In general, critics do not
suggest that a simile says one thing and means
another— they do not suppose that it means anything
but what lies on the surface of the words. It may
make us think deep thoughts, just as a metaphor does;
how come, then, no one appeals to the ‘‘special
cognitive content’’ of the simile? (p. 43)
Just as we don’t look for hidden meaning with similes, why
should we with metaphors? There seems, however, to be a
difference: Romeo did not mean Juliet is identical to a
celestial body. So, this argument goes, he must have meant
something else. And so, what he intended to communicate
must be different from what his words literally express.
And it’s this ‘‘something else’’ that is its metaphorical
meaning. So, even if similes can ‘‘work’’ without ‘‘special
cognitive content,’’ the same conclusion should not be
drawn about metaphors.
Davidson, in turn, wonders why we must assume there is
something a speaker who uses a simile means. Why isn’t its
purpose, Davidson asks, just to get a hearer to notice
similarities; not to mean something? The metaphorist’s
intention needn’t be to communicate a proposition
affirming noticed similarities, but rather for his hearer to
understand the proposition literally expressed; and then,
through that understanding perhaps be prompted (contingent
on his background information) to look for certain
similarities. Davidson says, ‘‘a simile tells us, in part, what
a metaphor merely nudges us into noting’’; metaphors
‘‘intimate what the simile declares’’ (p. 36).
When Romeo says, speaking metaphorically, ‘‘Juliet is
the sun,’’ Shakespeare expects his audience to grasp the
literal meaning, and thereby, be prompted to notice similarities
between Juliet and the sun. As with simile, no
‘‘proposition’’ is meant. It might well be that the speaker
herself believes various propositions; she may even convey
them with her utterance. But, so Davidson claims, if she
does, it is only as a ‘‘by-product’’ of getting her audience to
notice similarities. In short, there is no basis for claiming
propositions are speaker meant by her utterance.
Davidson writes, ‘‘Metaphor and simile are merely two
among endless devices that serve to alert us to aspects of
the world by inviting us to make comparisons.’’ The difference
between simile and metaphor—as opposed to
anything else that gets us thinking about similarities—
according to Davidson (p. 38)—is that whereas similes
literally tell us about resemblances, and other forms of
expression can direct our attention to similarities, metaphors
bully us into making comparisons.
2.3 Our Relationship to Davidsonian Views
Davidson’s critical discussion is telling. But his positive
view is merely a sketch, presented through analogies that
are provocative but, we find, misleading. To liken metaphor
to a drug or a bump on the head, for example, suggests
a model where metaphor falls in with perlocutionary
effects like intimidation. And as we have seen, this account
of the relationship between pragmatics and psychology in
metaphor is unsatisfying, because it doesn’t explain the
audience’s need to recognize the metaphor and participate
in its exploration.
Other philosophers also eschew metaphorical meaning,
in ways consistent with Davidson’s criticisms. Rorty
(1987), for example, describes metaphors as ‘‘unfamiliar
noises’’, and emphasizes that an audience, confronted with
a new metaphor, undertakes an active and creative effort of
interpretation. This process distinguishes metaphor from
meaning; it involves no conventions, no signaling. However,
Rorty locates this interpretive effort as part of a
continuum of more or less constrained problems of interpretation.
His continuum runs over many of the distinctions
we have charted, and so, in some respects, it too obscures
the relationships between pragmatics and psychology in
metaphor. The metaphorist intends his audience to shape
their interpretation in distinctive ways appropriate to
E. Lepore, M. Stone
123
metaphor, and they can appreciate the point of the metaphor
only by following through. Rorty locates birdsong in
this continuum, for example. Hearing an unfamiliar howl in
the jungle, we actively explore interpretations that explain
what we must be hearing. But this is to liken metaphorical
interpretation to recognizing Stalnaker’s secondary effects:
what must be the case for an utterance even to be made.
This is qualitatively different, we have seen, from
metaphor.
In short, giving a specific locus for metaphor in pragmatics
and psychology, not just in an undifferentiated
wastebasket of ‘‘use’’, gives us a conceptual framework,
empirical analogues and technical resources for thinking
about metaphor in precise new ways. Our view thus allows
us to endorse many of the criticisms Davidson marshals
against metaphorical meaning, while responding in new
ways to the objections others level against him.
3 Metaphor Without Meaning
There are positive arguments in favor of metaphorical
meaning, and critics have attacked the Davidsonian position
on various grounds.9
Most of them we believe miss
their target but one that most critics urge is that the account
conflicts with the intuitions about how metaphors are
organized, understood, and used (Bergmann 1982; Hills
1997; Stern 2000). We respond to these criticisms here.
3.1 Metaphor and semantics
First, there seem to be cases where metaphors interact by
compositional semantics with the interpretation of other
elements in a sentence. How can metaphors do such a thing
unless they have meanings? We think the data comes in
two families which require different responses. One family
is illustrated by the famous quote by Donne in (7).
7. No man is an island entire of itself; every man
is a piece of the continent, a part of the main.
With (7), of course, Donne compares the social relationships
among people to the geographical relationships
among places. The metaphor may seem to ascribe a new
interpretation to the word ‘‘island’’, associating it with the
property of being so socially isolated as to be devoid of
connections to others. The first clause of (7) denies that any
man has this property. How are we to square this
interpretive interaction with our claim that there is no
metaphorical meaning, that is, no new sense of ‘‘island’’ at
play in (7)?
We deny that the metaphorical thinking involved in
understanding (7) takes scope with respect to the semantic
constructs of Donne’s description. Instead, we suggest, the
reader’s engagement with (7) involves thinking of people
as places once and for all, and interpreting Donne’s words,
with their literal meanings, in the landscape that results.
There are no islands, the reader finds, in this fantasy, only a
patchwork of regions, identified as people in our imaginative
correspondence, forming a single vast continent.
This way of looking at metaphorical interpretation makes
clear that the word ‘‘island’’ is not the unit of Donne’s
metaphor. The entire discourse is. Accordingly, this way of
looking at metaphor dispenses with any metaphorical
meaning for Donne’s words, as we say it must.
Though example (7) involves negation, it’s easy to see
how this explanation would extend to other operators that
take scope within a clause, including tense, modality,
aspect, quantification, and others. Since we can imagine, in
any of these cases, what the world would have to be like for
the clause to be true, we can actively undertake to imagine
things that way, against the background of an imaginative
correspondence initiated by metaphorical thinking.
The second family of potential counter-example is found
in attitude reports such as (8)
8. Does Chris really think no man is an island?
I don’t.
In (8), the speaker is not just asking if Chris subscribes to
the literal proposition that no men are islands. Rather, the
speaker may seem to be asking if his addressee accepts the
point of the metaphor, that people are all interconnected by
their social relationships. This requires the point of the
metaphor to be part of the content of the thought. With a
commitment to metaphorical meaning, we have a ready
explanation: take ‘‘no man is an island’’ semantically to
contribute the proposition that people are all interconnected
by their social relationships. But how can the semantics
derive this requirement on our view, where we claim that
the point of the metaphor is not a matter of semantics?
We propose instead to refine our understanding of the
meaning of ‘‘think’’ and other propositional attitude words.
With (8), it seems that the addressee cannot truthfully
answer ‘‘yes’’ just because Chris agrees that people are all
interconnected by their social relationships. Answering
‘‘yes’’ also requires that Chris accepts the metaphor as apt,
and moreover that Chris is drawn from there by metaphorical
thinking to appreciate that people are all interconnected
by social relationships. The metaphor must be
active in Chris’s thought, and so it must somehow also be
active in the truth conditions of (8).
How can we capture the active role of metaphor? We
have argued that our insights in metaphorical thinking are
prompted just by the literal meanings of utterances.9
See Moran (1989 and 1997) for these and other objections.
Against Metaphorical Meaning
123
Analogously, we will argue that statements such as (8) also
report metaphorical thinking, and the insights that follow
from it, through constructions that appeal just to the literal
meanings of the utterances with which the thoughts are
framed. In other words, a sentence of the form ‘‘A thinks
that p’’, where p is normally understood metaphorically, is
true if and only if A endorses the similarities A discovers
upon entertaining the metaphor p. Finally, then, though (8)
may tell us something about the role of indirection and
metaphor in the meaning of the word ‘‘think’’, it tells us
nothing about indirection and metaphor in the meaning of
‘‘no man is an island.’’
3.2 Metaphor and Pragmatics
Even if we are right, and compositional semantics does not
require metaphorical meaning, others have argued that
pragmatics requires metaphorical meanings. We can give
instructions with metaphors, for example. Swimming
instructors are fond of using imagery like (9) to teach the
butterfly, the strenuous but beautiful stroke where swimmers
must lift their shoulders enough on the recovery to
swing both arms at once forward over the water.
9. With each stroke, let a wave flow along the length
of your body, from your head down your back and
through your legs.
The instruction does not advise the swimmer literally to
bob along with the surface of the water in the pool, but
invites the swimmer to guide their action by the image of a
figurative wave: the sinuous flow the swimmer creates in
her own body. How does an addressee know what to do, on
hearing (9), without extracting an action description that
counts as the meaning of the coach’s metaphor? Indeed,
how else does the addressee even realize that a metaphorical
instruction such as this gives a real command?
Such objections fail to confront the richness with which
our imaginative efforts can guide our real-world actions.
Problematically, in swimming the butterfly, it is no use
trying to lift the shoulders outright. The hips fall, the body
drags, the arms pull less effectively, and, for all that, the
shoulders remain too low in the water, as they rise only as
much as the swimmer’s natural buoyancy permits. The
image of (9) reorients the butterfly recovery as a dynamic
process. It suggests to the swimmer that she can lift the
shoulders for a brief moment in the cycle of the stroke by
harnessing the strength of the abdomen and the propulsion
of the feet, then goes on to hint how she can maintain an
efficient and fluid rhythm in the stroke by accepting and
relaxing into the downward reaction in the head and
shoulders that inevitably follows. She can use the image of
a wave to guide the orchestration of her movements. Such
an imaginative effort can result in swimming the butterfly,
even when the swimmer herself cannot conceive of a literal
description of the movements that constitute that action.
When a swimmer knows her coach demands both her
imaginative and her real-world effort, there seems no difficulty
for her to play along with (9) as an instruction, and
swim in response.
More generally, we think, objections to a Davidsonian
account of metaphor on pragmatic grounds miss the extent
to which non-serious, imaginative and indirect language
often gathers a life of its own, in the collaborative action of
interlocutors, across extended patterns of interaction.
Humor shows this most clearly. Examples (10) and (11) are
edited transcripts from The Colbert Report, omitting disfluencies
and side conversations that are irrelevant to our
point.
10. Stephen Colbert and Penn Gillette, on cold hard
reality
C: But why would you break the hearts of children
who want to believe there is magic in the world?
G: [to the camera] There is no Santa Claus [bleeped].
The tooth fairy is your mother and your father
[bleeped]. There is no Easter bunny [bleeped],
and, no God. Sorry
C: Do we have a puppy for Mr Gillette to punch,
before we leave here today?
G: I don’t like to touch them. I like to hit them with a
hammer
In (10), Stephen Colbert teases magician Penn Gillette, of
the performance team Penn and Teller, about his strategy of
giving away magicians’ secrets. Colbert offers humorous
imagery to describe the pain Gillette might inflict by
revealing the information he does. Gillette ‘‘plays along’’ by
jokingly offering his own mock demonstrations of cruelty.
With his final comment, Colbert compares telling children
the truth to punching puppies. Gillette’s response endorses
this cruel image, and takes it further. He gets no visceral
satisfaction, he jokes: the cruelty itself sustains him.
In (11), Colbert’s banter with documentary filmmaker
Morgan Spurlock hints at the subversiveness of imagery
that allows people to understand one another better.
11. Stephen Colbert and Morgan Spurlock, on empathy
S: We kind of live in a world where we don’t really
know what other people’s lives are like, and by
doing this [Spurlock’s TV show, 30 days] you kind
of immerse yourself in an environment where you
learn a little bit along the way
C: But don’t we live our own lives so we don’t have
to know what other people’s lives are like? I
mean, that’s why I’ve got power windows, so I
E. Lepore, M. Stone
123
can roll up the window when I go through the
neighborhoods I don’t live in
S: I think we need one in between the seats too, so
you can’t see the person next to you
C: That’d be fantastic. Maybe our own oxygen system
In response to Spurlock’s description of his new show,
Colbert takes up his character of conservative talk-show
host to playful extremes with his imagery of disengagement.
Spurlock follows along, pushing Colbert to a
reductio which pictures everyone living alone in a bubble.
We think such examples are commonplace, not just for
jokes, but for metaphor, hints, and other forms of imaginative
and indirect discourse. We emphasize both the
ordinariness and the playfulness of interlocutors’ language.
Gillette, in the final utterance of (10), for example, presents
an utterance that purports, in the ordinary way, to contest a
presupposition of Colbert’s prior utterance: that Gillette
expects to punch puppies. Gillette avers that he will smash,
but not punch. This is an entertaining development of the
interlocutors’ theme—in fact, their metaphor—of truth as
cruelty. It should not be taken to indicate that Colbert really
did presuppose that Gillette would expect to punch a
puppy, nor that Gillette really does like to hit puppies with
hammers, nor that Colbert and Gillette have a real dispute
over what Gillette likes to do with puppies.
Likewise, Colbert and Spurlock in (11) present utterances
that purport, in the ordinary way, to develop a solution to a
collaborative planning problem: how best to avoid contact
with people unlike ourselves. But Spurlock’s point,
clearly—and perhaps Colbert’s as well—is not to solve this
problem, but to ridicule it. In no way is it a goal worth
solving.
In (10) and (11), interlocutors use the language of
debate, but in fact what they are doing is developing
imagery and cementing a shared understanding of their
situation, actions and values. They do not contest propositions.
In much the same way, we expect, interlocutors use
their metaphorical discourse not to assert and deny propositions
but to develop imagery and to pursue a shared
understanding. Such practices can account for our interactions
in using metaphor, without appealing to metaphorical
meaning or metaphorical truth. Indeed, they are all
the stronger for dispensing with such notions.
3.3 Metaphor and Understanding
Whether or not we need metaphorical meaning for semantics
or pragmatics, some have argued we still need metaphorical
meaning to describe the knowledge and processes involved
in understanding a metaphor. Metaphorical interpretation
does rely on background knowledge. No one would interpret
an utterance of ‘‘Man is a wolf’’ as men are vegetarians.
Anyone interpreting the metaphor as such would either, as
Black says, not be a member of our speech community or fall
short of our expectations of the common knowledge of those
members; he would ‘‘produce an effect of paradox and
provoke a demand of justification’’ (Black 1955).10
Indeed,
there is something culturally-defined in the ‘‘system of
associated commonplaces’’ which seems to guide listeners’
imaginative efforts in metaphor. To say of John ‘‘He is a
gorilla’’ is to invite comparisons with the gorillas of myth
and stereotype, and not the gorillas of nature.
We accept such observations, but they do distinguish
sharply between the mechanisms of meaning and other
tools we have for sharing information with one another.
Returning to Herod, Salome, and the head of John the
Baptist, for Salome to understand from Herod’s demonstration
that John is dead, she must draw on background
knowledge: severed heads are doomed to death. To fail to
draw this conclusion would fall short of our expectations of
the common knowledge of members of our speech community.
Nevertheless, in light of the distinctions elaborated
in §1, this is not to argue that Herod means by showing
John’s head to Salome that John is dead. Black’s observation
about wolves is thus no argument for metaphorical
meaning either.
Commonplaces of the imagination, of course, are just as
effective in prompting reactions to real-world stimuli as is
knowledge of actual facts. A gun intimidates, in Austin’s
example; it is genuinely dangerous. But threatening, say, to
feed somebody grubs or insects might be perfectly scary or
revolting—though they are perfectly safe as food, even a
delicacy in certain parts of the world. In our way of life, we
imagine them as foul, diseased, poisonous. To appreciate
the power of the imagination in such cases, and to note that
the imagination lives a life of its own, divorced from what
we might know to be the case, is not to locate the force of
the imagination in convention or meaning. It is no mere
convention that offering the grubs revolts you. But we can
say the same of the power of the imagined gorilla, in
Black’s example. It is no argument for metaphorical
meaning.
We often understand and at times misunderstand metaphors.
Critics disagree about which interpretation is more
accurate. The coherence of their disagreement suggests a
‘‘cognitive content’’ an audience might understand or
misunderstand. But how does such data support the existence
of metaphorical meaning? Why isn’t what’s understood
or misunderstood the point of the metaphor—that is,
why isn’t it what the metaphorist wants the interpreter to
notice, to see without communicating it to him. The sense
in which metaphors are ‘‘cognitive’’ is that they can be
10
See Hills (1997) and Bezuidenhout (2001) for further discussion of
this point.
Against Metaphorical Meaning
123
understood or misunderstood; but this phenomenon is only
‘‘derivative,’’ as it were, they are misunderstood only in the
sense that their point is missed.
Again, it’s certainly possible to mistake someone’s hint,
or to fail to get their humor. Nobody would take such cases
of misunderstanding as evidence that jokes, for example,
carry indirect meanings. We insist on the parallel with
metaphors.
3.4 Dead Metaphors
Some authors have thought that dead metaphors pose a
problem for any theory that denies metaphorical meaning.
Everyone agrees that when a metaphor dies, it acquires a
new literal meaning; and in normal circumstances this literal
meaning is what a speaker means when he uses the
dead metaphor. It has seemed plausible to many to suppose
that this newly acquired meaning must have been active all
along while the metaphor was alive. Without making some
such supposition, the argument goes, how could one possibly
account for the cognitive content’s current association
with the dead metaphor? But then, isn’t it plausible to
suppose the newly acquired literal meaning is the ‘‘cognitive
content’’ of the once living metaphor that Davidson
denies exists?
Why is this supposed to be a problem for Davidson?
Suppose that when a metaphor dies, it thereby acquires a
second literal meaning. But even if dead metaphors—when
living—were used to introduce content, how does this help
the advocate of the argument from dead metaphors unless
he can establish that the new literal meaning is what we
speaker meant by uses of the then alive metaphor?
Davidson, in turn, asks us to look at those countless
metaphors that are not destined to die. Why don’t they die?
His response is that such metaphors do not die because they
cannot; and they cannot because they are not used to
speaker mean anything propositional. They are constitutionally
incapable of dying. Thus, while some metaphors
(those destined to die) might well be used to introduce
some content, the same cannot be claimed for all of them.
Here’s Davidson’s own response:
If metaphor involved a second meaning, as ambiguity
does, we might expect to be able to specify the special
meaning of a word in a metaphorical setting by
waiting until the metaphor dies. The figurative
meaning of the living metaphor should be immortalized
in the literal meaning of the dead. But
although some philosophers have suggested this idea,
it seems plainly wrong. ‘‘He was burned up’’ is
genuinely ambiguous (since it may be true in one
sense and false in another), but although the slangish
idiom is no doubt the corpse of a metaphor, ‘‘He was
burned up’’ now suggests no more than that he was
very angry. When the metaphor was active, we would
have pictured fire in the eyes or smoke coming out of
the ears. (p. 419)
If metaphors involved ‘‘second meanings,’’ these would be
the literal meanings acquired by metaphors upon death. But
they are not, and so, metaphors lack them. One might
challenge his premises by claiming only the cognitive
aspect of a living metaphor emerges as the literal meaning
of the dead metaphor or challenge his claim that the literal
meaning of the dead metaphor is the ‘‘second meaning’’ of
the living one. What distinguishes the living from the dead
metaphor is that the former has non-semantic (perhaps
imagistic) components not possessed by the latter. There is
an important difference between the literal meaning a
metaphor acquires upon its death and the ‘‘intimation’’ of a
living metaphor. But this hardly challenges Davidson; it is
his point! The ‘‘intimation’’ of a living metaphor appears
different in kind from the newly acquired speaker meaning
of a dead metaphor. To claim living metaphors have
special ‘‘cognitive contents’’ distinct from their literal
meanings is to suggest what is false: namely, that literal
meanings and metaphorical ‘‘intimations’’ are alike.
4 Conclusion
In this paper, we have distinguished metaphorical thinking—developing
imagery, seeing one thing as another,
noticing similarities—from merely grasping a proposition,
namely, the one that is speaker meant, brought about
through an intention to present information through coordination
or intention recognition. If we are right, then
researchers who try to assimilate metaphor to other
mechanisms for meaning making in language are on the
wrong track. Relevance theorists (Carston 1991; Bezuidenhout
2001, Recanati 2001), contextualists (Stern 2000),
and traditional pragmatic theorists (Grice 1975, Searle
1979), each in their own way, develop an account of
metaphor in familiar terms. To the extent that metaphor is a
distinct process with a distinct character and distinct consequences,
as we have argued, the uniformity of these
accounts is a weakness, not a virtue.
Other researchers, however, give accounts of metaphorical
meaning that do depend on a distinctive kind of
metaphorical thinking. Hills (1997), for example, following
Walton (1993), sees metaphorical thinking as prop-oriented
make-believe. On Hills’ account, the resemblances we
appreciate in metaphor help us to develop an imagined
world where one thing takes the place of another. Camp
(2003), meanwhile, describes metaphorical thinking as a
kind of analogical reasoning. For Camp, by finding
E. Lepore, M. Stone
123
metaphorical parallels, we can focus on salient properties
of familiar objects and use them to attend anew to the
counterparts we discover for them in a new domain.
In this paper, we have said nothing to undermine such
views of metaphorical thinking. We think addressees might
recognize the point of a metaphor in the ways that Hills or
Camp describe. We counter only the arguments Hills and
Camp advance that the point so recognized is the meaning
of the metaphor.
Why should we care if a point so recognized is a
meaning? What’s in a name? Both meaning and metaphor
sit at the center of complex webs of phenomena, principles
and puzzles. There is much to gain from keeping the issues
separate. In the realm of meaning, the hardest problems
come in regimenting our knowledge of language, and thus
coming to understand how that knowledge is acquired,
structured, and used in speakers’ faculty of language. Our
intuitions about literal meaning place heavy constraints on
key semantic notions: truth, reference, context and logical
form, among others. Progress, we have argued (Cappelen
and Lepore 2005), depends on developing and applying
clear standards that demarcate semantic phenomena narrowly.
If we can locate metaphor elsewhere, it is good
news for meaning.
The study of metaphor likewise benefits from putting
issues of meaning aside. Philosophers have always wanted
to know, not just what metaphor is, but what we do when
we appreciate or craft a metaphor. It prejudges such
questions to class metaphor as meaning: it suggests that we
produce and understand metaphors in more or less the
ordinary way we produce and understand all other language.
Distinguishing the point of metaphor from meaning
lets us settle these questions on their merits.
Since Aristotle’s invocation of the special genius of the
metaphorist, critics have worked to pin down the ingredients
that make metaphor special. Rarely do they invoke the
‘‘prosaic’’ ideas of coordination, convention, or intention
recognition. More often, they locate the power of metaphor
in the real relationships among things that only metaphorical
thinking can bring us to notice. Barfield (1928/
1973) quotes Bacon on metaphor:
Neither are these only similitudes, as men of narrow
observation may conceive them to be, but the same
footsteps of nature, treading or printing upon several
subjects or matters.
Barfield goes on:
This is the answer. It is these ‘footsteps of nature’
whose noise we hear alike in primitive language and
in the finest metaphors of poets. Men do not invent
those mysterious relations between separate external
objects, and between objects and feelings or ideas,
which it is the function of poetry to reveal. These
relations exist independently, not indeed of Thought,
but of any individual thinker. And according to
whether the footsteps are echoed in primitive language
or, later on, in the made metaphors of poets, we
hear them after a different fashion and for different
reasons (Barfield, 1928/1973, p. 86).
Meaning, certainly, does not work this way. Metaphor
might. Divorcing metaphor from meaning, as we advocate
here, opens the door to honor, to refine and to defend such
insights, in the terminology of psychology, action and
information. Where it belongs.
Acknowledgments We are grateful to Sam Cumming, Alex Lascarides
and Dan Sperber for feedback on this work. This work was
supported in part by NSF grants CCF-0541185 and HSD-0624191.
Earlier drafts were presented at the University of Oslo, St Petersburg
State University, and CRAL/CNRS in Paris. We thank these audiences
for their input.
References
Austin JL (1962) How to do things with words. Harvard University
Press, Cambridge
Barfield O (1928/1973) Poetic diction: a study in meaning. First
published by Faber and Gwyer, 1928. Reference to Wesleyan
edition, 1973
Bergmann M (1982) Metaphorical assertions. Philos Rev 41:229–245
Bezuidenhout A (2001) Metaphor and what is said: a defense of a
direct expression view of metaphor. Midwest Stud Philos
25:156–186
Black M (1955) Metaphor. Proc Aristot Soc 55:273–294
Camp E (2003) Saying and seeing-as: the linguistic uses and
cognitive effects of metaphor. PhD Dissertation, UC Berkeley
Cappelen H, Lepore E (2005) Insensitive semantics: a defense of
semantic minimalism and speech act pluralism. Blackwell,
Oxford
Carston R (1991) Implicature, expliciture, and truth-theoretic semantics.
In: Davis S (ed) Pragmatics: a reader. Oxford University
Press, Oxford
Davidson D (1979) What metaphors mean. In: Sacks S (ed) On
metaphor. University of Chicago Press, Chicago, pp 29–45
Devault D (2008) Contribution tracking: participating in task-oriented
dialogue under uncertainty. PhD Dissertation, Rutgers
Grice HP (1957) Meaning. Philos Rev 64:377–388
Grice HP (1969) Utterer’s meaning and intentions. Philos Rev
78:147–177
Grice HP (1975) Logic and conversation. In: Cole P, Morgan J (eds)
Syntax and semantics, vol 3. Academic Press, New York
Grice HP (1989) Studies in the way of words. Harvard University
Press, Cambridge
Hills D (1997) Aptness and truth in verbal metaphor. Philos Top
25(1):117–153
Lampert MD, Ervin-Tripp SM (2006) Risky laughter: teasing and
self-directed joking among male and female friends. J Pragmat
38:51–72
Lepore E, Stone M (2007) Logic and semantic analysis. In: Jacquette
D (ed) Handbook of the philosophy of science, vol 5: philosophy
of logic. Elsevier, Amsterdam, pp 173–204
Against Metaphorical Meaning
123
Lewis D (1969) Convention: a philosophical study. Harvard University
Press, Cambridge
Lewis D (1979) Scorekeeping in a language game. J Philos Log
8:339–359
Moran R (1989) Seeing and believing: metaphor, image, and force.
Crit Inq 16:87–112
Moran R (1997) Metaphor. In: Hale B, Wright C (eds) Companion to
the philosophy of language. Blackwell, Oxford
Recanati F (2001) Literal/nonliteral. Midwest Stud Philos 25:
264–274
Rorty R (1987) Unfamiliar noises I: Hesse and Davidson on
metaphor. In: Proceedings of the Aristotelian society (supp.
volume 61), 283-296
Searle J (1979) Metaphor. In: Expression and meaning: studies in the
theory of speech acts. Cambridge University Press, Cambridge
Stalnaker R (1978) Assertion. In: Cole P (ed) Syntax and semantics,
vol 9. New York Academic Press, New York, pp 315–332
Stalnaker R (1998) On the representation of context. J Log Lang Inf
7:3–18
Steedman M, Stone M (2006) Is semantics computational? Theor
Linguist 32:73–89
Stern J (2000) Metaphor in context. MIT Press, Cambridge
Stone D, Patton B, Heen S (1999) Difficult conversations: how to
discuss what matters most. Penguin, Harmondsworth
Thomason RH (1990) Accommodation, meaning, and implicature:
interdisciplinary foundations for pragmatics. In: Cohen PR,
Morgan JL, Pollack M (eds) Intentions in communication. MIT
Press, Cambridge, pp 325–363
Walton K (1993) Metaphor and prop-oriented make-believe. Eur J
Philos 1:39–57
E. Lepore, M. Stone
123