Published in Midwest Studies in Philosophy, 25, 2001, pp. 156-186.
1
Metaphor and What is Said: A Defense of a Direct Expression View of Metaphor
Anne Bezuidenhout
1. Introduction
According to one widely held view of metaphor, metaphors are cases in which the speaker
(literally) says one thing but means something else instead. I wish to challenge this idea. I will
argue that when one utters a sentence in some context intending it to be understood
metaphorically, one directly expresses a proposition, which can potentially be evaluated as either
true or false. This proposition is what is said by the utterance of the sentence in that context. We
don’t convey metaphorical meanings indirectly by directly saying something else. One
consequence is that, contrary to what Searle (1993: 110) suggests, we do not arrive at the
metaphorical meaning that the speaker intended via a literal interpretation of the sentence the
speaker utters. The defense of this view depends on articulating a conception of what is said that
is more generous than that allowed for by Searle (1993) and others such as Bach (2001). I hope to
motivate this broadened conception of what is said (what I call a contextualist conception of what
is said), and to show some of the benefits of adopting a direct expression view of metaphor.
Despite the fact that philosophers such as Searle have defended the view that metaphorical
interpretations are indirectly conveyed via the expression of another proposition, the direct
expression view has a lot of initial intuitive appeal. Consider the following example, inspired by
one used by Carston (1997: 113). Suppose that a group of people is trying to decide whom to vote
for as the next chair of their department:
A: How about Bill?
B: Bill’s a bulldozer.
A: That’s true. But isn’t that a good thing in this case? We want someone who’ll stand up to
the administration and get things for our department.
C: I disagree that he’s a bulldozer; that exterior hides someone who’s basically insecure.
But, either way, Bill wouldn’t make a good chair.
In this dialogue A and C are agreeing and disagreeing respectively with the metaphorical
content of B’s utterance. The most natural construal to put on such a dialogue is that B says
something, that B says it with assertoric force, and that A and C are either agreeing or disagreeing
with what B says. Stern (2000: 24) writes: “the ordinary appearance is that utterances of
sentences that contain metaphors are truth-valued, express propositions, and can be used to make
assertions (or other speech acts that presuppose assertion). The burden of the argument, therefore,
falls on those who deny that this appearance is reality.” I am partially in agreement with Stern. I
2
do not agree that utterances of sentences containing metaphors are always truth-valued, since
such sentences can just as readily be used to ask questions and make requests. E.g., ‘Will the
bulldozer make a good chair?’ or ‘Don’t mention the bulldozer again!’ However, I do agree that
all metaphorically used sentences directly express propositions. Assertions, questions and
requests containing metaphors all express propositions, although only in the case of assertions is a
claim made that it is appropriate to evaluate as either true or false.
Even though I sympathize with Stern’s point of view, I believe the direct expression view
needs defense, because the view that metaphorical contents can only be indirectly communicated
is so well entrenched (especially amongst philosophers of language). I will try to provide such a
defense. In sections 2 and 3 I set the stage for my account, which is laid out in section 4. In
section 5 I compare my view with Stern’s view of metaphors as demonstratives, and give some
reasons for preferring my account. In section 6 I anticipate a couple of objections that might be
raised against my direct expression view of metaphor.
Before turning to the task of this paper, I want to say that I agree with White (1996) that
philosophical discussions of metaphor tend to suffer from an impoverished diet of examples. How
often does one see examples such as ‘Sally is a block of ice’, ‘He is a gorilla’, ‘Man is a wolf’,
‘Juliet is the sun’? White thinks that such simple predicative metaphors are rare in actual use,
especially in literary use. In his own book he tries to avoid such trite examples and focuses
instead on highly poetic metaphors. I disagree however that only poetic metaphors should be used
in constructing a view about the nature of metaphor. I agree with Stern (2000) that it is equally
legitimate to focus on examples from ordinary discourse, and that highly poetic metaphors of the
sort discussed by literary theorists may not make the best starting point for theorizing about
metaphors. I certainly have no pretensions to be able to take on literary theorists on their own
turf, so unlike White (1996) I will not in general illustrate my points with literary metaphors.
However, I will try to stay away from the stock examples, and to use examples other than simple
predicative ones.
My own view is that the problem with a lot of philosophical discussions of metaphor is not
that they use hackneyed examples, but that they use examples out of context. So long as we
supply enough of a context to make our examples seem somewhat naturalistic, the use of nonliterary
metaphors does no harm. I think that the reason people use hackneyed examples is
because it means they don’t have to spend too much time developing a plausible context. They
hope to rely on the fact that their readers have seen these examples discussed so often before that
a ready-made context comes along with the example, obviating the need to waste space on
3
supplying such a context. This is what I believe is the mistake. Context is all-important, and
shouldn’t be neglected, even though it is tempting to do that. What one wants to avoid are “dead”
metaphors; ones that have become lexicalized. But even dead and dying metaphors can have new
life breathed into them by imaging a novel context.
2. What is said versus what is meant
Grice (1989: 88) held that a speaker has said that p just in case she intends to communicate
that p and the sentence that she utters in an attempt to communicate this means that p. In other
words, to say something requires that the speaker’s intended meaning coincides with the
conventional meaning of the sentence she utters. When this correspondence fails, as Grice
thought it must in the case of the metaphorical use of language, then nothing is said. At best in
such cases we “make as if” to say something. Many have found this consequence of Grice’s
account of saying unsatisfactory and have argued that we need to divorce what is said from
speaker meaning. For instance, Bach (2001) defends a syntactically constrained notion of what is
said. According to Bach, what is said is a projection of the syntax of the sentence as used in a
given context, in the sense that the constituents of what is said correspond to elements of the
sentence. Others, such as Searle (1993), operate with a similarly constrained notion of what is
said, tying what is said closely to sentence meaning and distinguishing it from speaker meaning
(or what Searle calls utterance meaning).
Once one separates what a speaker says from what the speaker means in this way, this opens
up at least the following four possibilities:
(1) One can say one thing and mean something else in addition. This according to Searle
(1993) is what happens in cases of indirect speech acts. E.g., I utter the sentence ‘It is
cold in this room’ in an appropriate context. What I have said is that it is cold in the
demonstrated room, but indirectly I may also be requesting that my hearer close the
window in that room. Something similar can be claimed about standard cases of
Gricean conversational implicature. E.g., my husband asks me whether I’d like to go
to the movies tonight and I reply by uttering ‘I’ve got a headache’. I’ve said one thing
(and meant it) but, given certain mutually manifest background assumptions shared
by my husband and me, I have in addition conversationally implied that I do not want
to go to the movies tonight.
(2) One can say one thing and mean something else instead. This according to Searle
(1993) is what happens in figurative uses of language, such as metaphorical and
4
ironical uses of language. E.g., I utter the sentence ‘The Israeli-Palestinian peace
process has been mortally wounded’ and I say something false (since the I-P peace
process cannot literally be wounded, given that it is not a living creature). This false
claim is not what I mean to convey. Rather, I intend to convey something that may
well be true, namely that the I-P peace process is unlikely to continue much longer.
In this second category one would also have to include cases of slips of tongue and
other such speech errors. In such cases one intends to say one thing but uses an
expression that (conventionally) means something different from what one intended,
so that what is said and what is meant come apart.
(3) One can say something and mean exactly that. This according to the Bach/Searle
view is what happens when we speak literally. How often this happens and in what
sorts of cases will depend very much on how one defines what is said. For instance,
on Bach’s very restrictive syntactic projection view of what is said there are many
fairly ordinary cases, which some might think involve the literal use of language, that
he would classify as cases of sentence nonliterality. E.g., ‘I am finished’ is
semantically underdetermined, according to Bach. What it is that I have finished
must be recovered from the context. However, there is no syntactic slot in the
sentence ‘I am finished’ that signals the need for this contextual element (unlike the
indexical ‘I’, which does signal the need to search the context for its semantic value).
What is said by means of such a semantically underdetermined sentence cannot be
propositional. Since what is meant is always something propositional, what is said
and what is meant can never coincide in such cases of semantic underdetermination.
Thus they will always involve an element of nonliterality.
(4) One can say something and mean nothing at all. This according to Bach (2001) is
what happens when someone recites lines in a play, or in other cases of pretend
communication. There are also cases in which someone utters words without
comprehension (e.g., when in a trance). Arguably something is said in such a
situation, although nothing is meant. (One might reasonably deny that in such cases
anything is said, for by parity of reasoning one would also have to claim that parrots
and other non-human mimics are saying things, which doesn’t seem plausible).
Perhaps here we also need to recognize the opposite possibility, namely cases in
which the speaker means something but says nothing at all. This might be the case
5
with someone heavily drugged. He tries to say something but what comes out is
something garbled and unintelligible.
Each of these claims about the relation of what is said to what is meant can be fruitfully
explored. For instance, there are many ways in which to question Searle’s views about indirect
speech acts. Bertolet (1994) challenges the very idea of an indirect speech act. Holdcroft (1994)
challenges the idea that a directive interpretation of a sentence such as ‘You will be there’ is any
less direct than an assertive interpretation. Bach (1995) introduces a notion of standardization,
which may go some way towards explaining Holdcroft’s observations. Bach argues that what is
(literally) said in cases that Searle treats as instances of indirection may be bypassed due to
familiarity with a certain sort of use. This sort of streamlining in processing means that the socalled
indirect meaning can be directly communicated. Bach in effect proposes a three-fold
distinction between what is said, what is directly communicated and what is indirectly
communicated.
However, in this paper I will focus exclusively on claim (2), and even more particularly on
the claim that in metaphorical uses of language what is said and what is meant come apart. I will
not be offering an account of ironical and other figurative uses of language. What I want to try to
make a case for is the view that in metaphorical uses of language the metaphorical interpretation
is directly expressed. We don’t convey metaphorical meanings indirectly by directly saying
something else.
3. Some Disclaimers
I should make it clear at the outset that I agree with Davidson (1978) that expressions used
metaphorically generally do not have metaphorical meanings that are lexicalized alongside their
literal meanings (although there may be some so-called “frozen” metaphors that have become
lexicalized). So I am not advocating a semantic ambiguity thesis with regard to sentences that are
used metaphorically. For instance, I do not believe that a sentence such as ‘He is a butcher’ –
sorry, this is a trite example – is semantically ambiguous between a literal and a metaphorical
meaning. I can use this sentence literally, to express the claim that a certain contextually indicated
male is a member of the profession that prepares cuts of meat to be sold at market. I can also use
it metaphorically to express the claim that a certain contextually indicated male, who is known by
my hearer and me to be a surgeon, performs his surgery in a manner that seems brutal and
6
unfeeling. He treats his patients’ bodies as though they were animal carcasses, without the dignity
that should be accorded to a fellow human being.1
The metaphorical interpretation that I have imagined involves pragmatically determined
aspects of content, and the hearer cannot retrieve it simply by a process of sense selection. Some
sort of sense creation must be involved. (Gerrig (1989) uses the terms ‘sense selection’ and ‘sense
creation’ to describe the difference between cases in which context is used to select between two
or more lexicalized meanings of an expression and cases in which context must be used to arrive
at a novel – a “nonce” or one-off – interpretation of a phrase.) This process of sense creation is
basically a pragmatic one, even though, as we will see, it is semantically constrained. But
although the intended metaphorical interpretation of ‘He is a butcher’ is pragmatic in nature, it is
not accessed by first considering and rejecting the literal interpretation of that sentence, in the
way suggested by Searle (1993: 110) or Grice (1989: 34). My alternative pragmatic account will
be described in section 4 below.
Given that I accept that metaphorical interpretations are pragmatic in nature, I reject attempts
to argue that metaphor is a semantic, not a pragmatic phenomenon. For instance, Tsohatzidis
(1994) reasons that if metaphorical meanings are pragmatic, they ought to be distinguishable from
other sorts of pragmatic meanings, such as cases of irony and indirection. But, he argues, Searle’s
way of distinguishing these types of speaker meaning fails. According to Searle what
differentiates metaphor from irony is that in the case of irony the opposite of what is said is what
is meant, whereas in the case of metaphor something other than what is said is meant, though it
isn’t simply the opposite of what is said. On the other hand, both irony and metaphor differ from
cases of indirection, since in the case of indirection the speaker means what she says but means
something else in addition, whereas in the case of figurative uses of language the speaker means
something other than what she says.
Tsohatzidis argues against Searle’s way of differentiating metaphor from irony and both of
these in turn from indirection. He believes that there are examples of these phenomena that would
violate Searle’s classification scheme. For instance, consider examples of the following sorts (I
1
In my dictionary ‘butcher’ n. is listed as having as one of its meanings ‘a person guilty of brutal
murder’ and ‘butcher’ v. has as one of its meanings ‘to botch or bungle something’ – as in ‘She really
butchered that job’. Thus one might want to argue that what I am calling the metaphorical meaning is
lexicalized. However, I would claim that the metaphorical interpretation I have imagined doesn’t quite
correspond to the ‘brutal killer’ meaning of ‘butcher’, and the intended metaphorical claim is not
simply that the surgeon botches or bungles his surgery. In fact, one can imagine that this particular
surgeon happens to be quite effective at his job. More important are the inferences the hearer is
intended to draw about the kind of attitude the surgeon has towards his patients and so on.
7
have slightly altered Tsohatzidis’ examples, but the points they are meant to illustrate are the
same):
(5) Our piglet is getting dirty.
(6) She’s the Taj Mahal.
Imagine that (5) is uttered by a husband to his wife in reference to their toddler who has wandered
off and is splashing around in a puddle of mud. (5) is being used metaphorically, but it could also
be intended as an indirect request by the husband to the wife to do something to stop the toddler
from getting any dirtier. In the case of (6), imagine that two men are talking about a woman and a
question arises as to whether one of the men finds the woman attractive. He utters (6), which is
clearly being used metaphorically, but he also intends it ironically (or sarcastically) to mean that
he finds the woman unattractive.
Tsohatzidis thinks that examples such as (5) and (6) present a problem for Searle. For
instance, in the case of (5), which is a case of indirection, Searle has to say that something is said
and meant, and that something else is meant in addition. But there doesn’t seem to be any
candidate for what is said (in Searle’s sense) which is also something that is meant. The husband
does not mean to be referring to a real piglet. In the case of (6), which is a case of irony, Searle
has to say that the ironical meaning is the opposite of what is said (in Searle’s sense – henceforth
what-is-saidS). But this would seem to give the wrong result, as the ironical meaning is not that
the woman in question is not the Taj Mahal.
Examples (5) and (6) do show that Searle’s simple scheme for differentiating between
metaphor, irony and indirection will not work. But perhaps a more complex scheme would work.
Searle has to grant that indirection and irony can be at one remove from that allowed for by his
scheme. There are cases where the hearer must first detect the speaker’s metaphorical meaning
and then only the irony or the indirection. Cases of once-removed indirection look more like what
Searle classifies as cases of metaphor, since no aspect of what is meant corresponds to what-issaidS.
Something similar holds for cases of once-removed irony. They too look more like what
Searle classifies as cases of metaphor, since the speaker’s intended meaning doesn’t correspond
to what-is-saidS, but nor does it correspond to the opposite of what-is-saidS. So in non-removed
cases Searle can retain his simple scheme, but cases of once-removed indirection and irony would
have to be lumped together with cases of metaphor. We could then distinguish within this class
between straightforward metaphors, metaphorical-irony and metaphorical-indirection. These
labels signal the fact that cases of the latter two sorts arise out of a prior metaphorical
8
interpretation. In these cases the hearer is invited to elaborate on the prior metaphorical
interpretation to discover additional (in the case of metaphorical-indirection) or different (in the
case of metaphorical-irony) meanings.
Even if one doesn’t think the more complicated scheme suggested in the previous paragraph
saves Searle, one should not grant Tsohatzidis’ further conclusion. Tsohatzidis claims that the
failure of Searle’s classification scheme means we must give up the idea that metaphor is a
pragmatic phenomenon and embrace the view that it is a semantic phenomenon. But this follows
only if Searle’s pragmatic account of metaphor is the only pragmatic account that is available.
But it isn’t. There are other ways of treating metaphor as a pragmatic phenomenon, including the
direct expression view that I defend in the following section.
Tsohatzidis has another objection to pragmatic accounts that is more worrying. He argues that
if metaphorical interpretations are pragmatic, they ought to pass Grice’s cancellability test. That
is, one ought to be able without anomaly to utter a sentence that could be metaphorically
interpreted, yet follow it up with a disclaimer that cancels the potential metaphorical meaning.
But this doesn’t seem to be possible with cases of metaphor where what-is-saidS involves a
category mistake, such as (6) above. A woman cannot literally have the property of being a
building, so one couldn’t without anomaly mean what-is-saidS by (6).
Metaphorical interpretations do appear to be cancelable in cases where no category mistake
prevents the sentence uttered from being interpreted literally. For example, in my surgeon/butcher
case I could without anomaly utter ‘He is a butcher, and I don’t mean with respect to the way he
behaves in the operating room. He moonlights down at the abattoir.’ But consider an utterance
such as ‘She is the Taj Mahal, and I don’t mean to comment on her remarkable beauty. She …’.
There doesn’t seem to be any content that could be inserted in the ellipsis that would make it
plausible to construe my remark ‘She is the Taj Mahal’ literally (assuming that ‘she’ refers to a
human female and ‘the Taj Mahal’ denotes a building in India). I will come back to this objection
in section 6, as it is a potential problem for my account too, given that I, like Searle, treat
metaphor as a pragmatic phenomenon.
Tsohatzidis discusses cases like (5) and (6), which are cases of once-removed indirection and
irony. It is interesting to ask whether there could be cases of once-removed metaphor. In other
words, we’ve seen that metaphors can be vehicles for irony or indirection. Could irony or
indirection be vehicles for metaphors? That is, could one utter something intending it to be taken
literally, but also intending one’s hearer to indirectly retrieve a further meaning, which in turn one
intended one’s hearer to interpret metaphorically? E.g., can I utter something like ‘It’s cold in
9
here’ intending it literally, but also intending my hearer to indirectly retrieve the idea that I want
my hearer to close the window, which in turn I intend my hearer to interpret metaphorically? And
are there cases in which one utters something that has a literal interpretation, but where one
intends one’s hearer to retrieve a meaning opposite to this literal meaning, which in turn one
intends one’s hearer to interpret metaphorically? E.g., can I utter ‘It’s cold in here’ intending my
hearer to retrieve the idea that it is hot in our immediate vicinity, which in turn I expect my hearer
to interpret metaphorically? Neither of these appears to be possible.
One can of course utter ‘It’s cold in here’ intending it literally and also simultaneously
intending one’s hearer to understand that one thinks the “emotional climate” between the two of
you is chilly. But this is something like a case of double entendre. ‘It’s cold in here’ is intended to
be simultaneously literally and metaphorically interpreted. It is not a case of a metaphor at one
remove.
Also, one can utter ‘It’s cold in here’ and intend one’s hearer to interpret this metaphorically
to mean that emotional relations between the two of you are strained. One might then, at a stretch,
expect one’s hearer to get the idea that one is indirectly implying that there is some “emotional
window” that is causing the strained relations and that one desires the hearer to close this
“window”. But if this is possible it is because the hearer has already interpreted one’s utterance
‘It’s cold in here’ metaphorically, and is able to play along with the idea of a “chilly emotional
climate”. Just as having an open window can cause chilliness in the physical environment, so can
some “rupture” in one’s relations cause chilliness in the emotional environment. And just as
closing the window can end the physical chilliness, so too can closing the emotional rupture end
the emotional chilliness. In other words, this would not be a case of metaphor at one remove. In
fact, this case is analogous to (5) above, where a metaphor is used to communicate a further
meaning or meanings. It would be a case of indirectness at one remove.
Similar sorts of considerations count against the idea that irony can be used as a springboard
for cases of once-removed metaphor. Thus we can say that metaphors (though not what might be
called metaphorical extensions or elaborations) must be launched from an utterance, and cannot
be launched from an interpretation that is itself reached only via the pragmatic interpretation of a
speaker’s utterance.
Stern (2000: 235-237) discusses a case that is a potential counterexample to my claim that
irony cannot be used as a springboard for cases of once-removed metaphor. Consider:
(7) He is a towering figure.
10
It seems possible that a speaker could use (7) intending his hearer to understand it ironically.
Stern assumes that the ironical interpretation of (7) is simply the contrary of (7), namely:
(8) He is a minute figure.
The speaker might intend this in turn to be metaphorically interpreted to mean that the man in
question is of no consequence, or that he is not taken seriously or respected by his peers, or
whatever. As Stern himself points out, this only seems possible because (7) is a “frozen”
metaphor. “Person of great importance” is arguably one of the lexicalized meanings of “towering
figure”. Stern’s final conclusion is in line with my claim that metaphors must be launched from
utterances. He says that metaphorical interpretations are operations on sentences, whereas ironical
interpretations are operations on propositional contents. (2000: 237).
One might dispute the claim that (7) is a “frozen” metaphor and argue instead that (7) and (8)
are possible linguistic manifestations of an underlying conceptual metaphor of the sort posited by
Lakoff & Johnson (1980). The conceptual metaphor would be something like SOCIAL STATUS
IS PHYSICAL SIZE. Amongst other possible linguistic manifestations of this conceptual
metaphor are:
(9) He is a giant in the recording industry.
(10) She is a person of great stature.
(11) He’s no more than a blip on her radar screen.
But even if one took this line, it would not lend support to the idea that there are once-removed
cases of metaphor. One may be able to understand (7) ironically as (8). But this is because one is
already in a metaphorical space where SOCIAL STATUS IS PHYSICAL SIZE. So this would
not be a case in which one understands the irony prior to understanding the metaphor.
I take the fact that there are no cases of once-removed metaphor (i.e., no cases of metaphors
launched from ironically or indirectly intended propositions) as support for the view that
metaphorical interpretations are different from other sorts of pragmatic interpretations. We are
thus able to satisfy Tsohatzidis’ demand that we specify the ways in which metaphors differ from
other pragmatic phenomena. Metaphorical interpretations have a kind of directness that isn’t
shared by other pragmatically derived interpretations, such as cases of irony and indirect speech
acts. Metaphors must be (initially) launched from utterances, although of course once one is
launched into metaphorical space, one can extend or elaborate on the metaphor – consider for
example the case of the “chilly emotional environment” discussed above. This does not yet
11
establish that metaphors directly express propositions. Something more needs to be said to
establish this stronger claim.
4. Metaphor and Directness
My direct expression view of metaphor depends on defending a notion of what is said that is
not as constrained as the syntactic projection view defended by Bach (2000) nor as constrained as
Searle’s view that ties what is said to literal sentence meaning. I call my conception of what is
said a contextualist conception, because it allows that some aspects of what is said may be
entirely pragmatically determined. That is, there may be no syntactic or semantic rule that
generates these aspects of what is said. Instead, the speaker intends the hearer to use the syntactic
and semantic clues provided, along with non-linguistic information available in their mutual
cognitive environment in order to recover what is said. The syntactic and semantic clues by
themselves will underdetermine what is said. Henceforth I will call this contextualist conception
what-is-saidC.
It will be useful to illustrate the contrast between what-is-saidC and Bach’s notion of what is
said – what-is-saidB – with an example. What-is-saidB allows for elements of content that must be
contextually determined, because it includes elements of content that correspond to indexicals
and to ambiguous expressions. But it excludes from what is said any content that does not
correspond to some syntactic slot in the sentence uttered. E.g., when I utter the sentence ‘I’ve
been to the bank’ in a suitable context, what-is-saidB is that Anne has been to the monetary
depository (i.e., the indexical has been assigned a referent, and the word ‘bank’ has been
disambiguated). Any suggestion as to when specifically this event occurred that you the hearer
might have been able to gather from the context will be excluded from what-is-saidB. This
additional aspect of content (assuming that I intended my hearers to access it) can at best be
implicit in what-is-saidB. Bach calls this pragmatically determined aspect of content an
impliciture to make it clear that it is not a part of what-is-saidB.
What-is-saidC on the other hand includes this purely pragmatically determined aspect of
content. To make this plausible one needs to consider the utterance in context rather than in
isolation. Suppose for instance that I work in a store and it is my job to take the cash and checks
from the previous day’s sales to the bank (i.e., monetary depository) every morning at around
9:00am. The manager comes in late one morning and wants to know what tasks have been done
so far that morning. I utter the sentence:
(12) I’ve been to the bank.
12
What-is-saidC by uttering (12) in this context is:
(13) Anne has been to the monetary depository on the morning of utterance.
On the other hand, suppose that I regularly help out an old lady who lives in a retirement home by
doing various errands for her. Every Friday afternoon I go to the bank (i.e., monetary depository)
for her and withdraw $50.00 from her account. Then when I visit her on Sundays I hand over the
money. One Sunday I come for my usual visit and, as I hand over her $50.00, I utter (12). Whatis-saidC
in this context is:
(14) Anne has been to the monetary depository on the Friday prior to the time of utterance.
In both the cases illustrated in the previous paragraph the sentence is pragmatically enriched
to restrict the event described to some particular period of time in the past. Of course, it is also
possible for a speaker to utter (12) intending simply to express the proposition that the speaker
has been to the monetary depository at some time before the time of her utterance. For instance,
suppose a group of villagers is discussing the amenities available in the next town. There is a
bank and a credit union in the town and for some reason a question arises as to whether anyone
has any knowledge of either of these institutions. Someone utters (12). What the speaker says is
that she has been to the bank sometime before the time of her utterance, but she doesn’t intend to
single out any particular time in the past as the time at which her visit occurred.2
In this situation,
the speaker expresses a complete proposition. It is an existential proposition that quantifies over
times, and it could be represented in first-order predicate logic as follows:
(15) (∃t) (t is a period of time earlier than τ and α visits β at t)
where ‘τ’ is the time of context of utterance, ‘α’ is the speaker in the context and ‘β’ is the
monetary depository denoted by the description ‘the bank’ in the context.
Note that the proposition represented by (15) is not what the sentence (12) expresses, not
even given an assignment of references to the indexical and the definite description in (12). It is
what-is-saidC by (12) in the particular context of utterance we have imagined. The past tense
construction of the verb in (12) simply points to the fact that the time of the event described in the
sentence is past, but it does not (either implicitly or explicitly) quantify over past moments of
time. In other words, what is semantically and syntactically encoded in (12) underdetermines the
propositions it can be used to express in particular contexts. It can be used to express (13), (14),
(15), and indefinitely many other propositions.
2
In this context she may also conversationally imply that she has not been to the credit union.
13
Some people may be tempted to say that (15) deserves to be singled out as a literal
interpretation of (12), whereas (13) and (14) are non-literal interpretations that contain conceptual
material extracted from background assumptions available in the context of utterance. Hence only
(15) has a legitimate claim to be identified with what is said by an utterance of (12). However,
each of these propositions is reached via a process of pragmatic enrichment, although one might
say that the pragmatic enrichment in the case of (15) is the minimal needed to express a complete
proposition. I prefer to say that all the interpretations (including the minimal one) are candidates
for what-is-saidC.3
If one singles out (15) as the only legitimate contender to the title of what is said, then one
will have to say that in the other situations that I imagined there must have been some minimal
existential proposition that was really what was said. On this view (13) and (14) are at best
implicit in what was said. This is in fact the position adopted by Bach (1994, 1999). However, the
idea that interpretations such as (13) and (14) are indirectly conveyed via the direct expression of
the minimal interpretation (15) lacks psychological plausibility. Even if (15) deserves to be called
a literal interpretation, there is a great deal of psychological evidence that literal interpretations
are not privileged from the processing point of view. Many reading time experiments have found
that nonliteral interpretations are accessed as quickly as literal ones, casting doubt on the view
that pragmatically enriched interpretations are indirectly conveyed. Gibbs (1994) summarizes
much of this psychological work, especially as it applies to the processing of metaphors, ironical
utterances, indirect speech acts, idioms and other figurative uses of language, such as cases of
metonymy, hyperbole, litotes, oxymoron and so on. Bezuidenhout & Cutting (2001) discuss the
problem as it applies particularly to cases such (15) compared with (13) and (14).
Bach (2001) invokes his notion of standardization to account for such psychological findings.
It may well be that minimal propositions are bypassed in processing because enriched uses such
as (13) and (14) are so familiar. Familiarity with certain sorts of uses can set up a default
interpretative pathway that obviates the need to first consider the minimal interpretation and only
then move to an enriched interpretation such as (13) or (14). Nevertheless it may still be that the
minimal interpretation is dispositionally available to the processing system. Besides, Bach
complains, the psychological evidence about how hearers process utterances is irrelevant to the
issue as to what the speaker has said. Thus an appeal to psychological findings should not be
allowed to influence our philosophical account of the notion of what is said.
3
Notice that in those cases in which (12) is interpreted as a quantificational claim, what sort of
quantifier is expressed can vary depending on the context. What-is-saidC could be that the speaker has
14
Bezuidenhout (2001) addresses both these claims. Regarding the latter point, I argue that
Bach’s identification of what is said with a minimal proposition is psychologically implausible
both from the point of view of language production (the speaker’s point of view) and from the
point of view of language comprehension (the hearer’s point of view). If this is so, it is not as
easy to dismiss psychological evidence as irrelevant to philosophical concerns. Regarding the
former point, merely to claim that minimal propositions are dispositionally available to the
processing system doesn’t set Bach’s view apart from the contextualist view. Contextualists can
accept the claim that minimal propositions are dispositionally available. There clearly are
contexts in which the minimal proposition is what-is-saidC, and in those contexts the hearer ought
to be able to access that interpretation. Thus unless Bach can make some empirical predictions
about the psychological role of minimal propositions that differ from the predictions made by
contextualists, Bach’s notion of what-is-saidB will be empirically empty. It is unsatisfactory to
retreat to the claim that what-is-saidB is relevant to a philosophical account of semantic and
pragmatic phenomena even if it has no psychological validity. Our account of natural language
semantics and pragmatics ought to be constrained by what it is psychologically plausible to
attribute to speakers and hearers of natural language.
This debate between contextualists and their critics cannot be settled here. I hope to have at
least made it clear what the contextualist view is. So far I’ve discussed cases in which what is
semantically encoded in the sentence uttered underdetermines what is said and in which a
pragmatic process of enrichment adds elements of content to the encoded content. But there is
also a pragmatic process of the opposite sort that eliminates elements of content. This is a process
of loosening in which elements of content that are semantically encoded in the sentence uttered
are discarded with the result that what is expressed is a proposition that is broader than what is
encoded. For example, consider:
(16) It is silent here.
(17) This meat they’ve served me is raw.
I will discuss only example (16). I leave it to the reader to construct a comparable story about
example (17). The word ‘silent’, as applied to a state or condition of the world as opposed to a
person, encodes the idea of a state in which nothing is audible, a state in which all sound or noise
is absent. But one does not always use the word so strictly. Thus suppose my companion and I
enter a house and stand in the hallway. There appear to be no people around making the usual
sorts of domestic noises that people make. But if I really strain my ears I can catch the sound of
visited the bank many (a few, a couple, several, a dozen, etc.) times.
15
the refrigerator humming. I can hear a faint scratching that I think might be a squirrel or a mouse
up in the eaves of the house. Every now and then there seems to be a plopping sound that could
be water dripping from a tap somewhere in the house. I turn to my companion and utter (16). I
intend my hearer to understand that my use of ‘silent’ is a loose use, which excludes only sounds
that are clearly audible, but doesn’t exclude sounds that are barely audible, or only just above the
threshold of audibility, such as a dripping tap or humming refrigerator.
What-is-saidC by my utterance in such a context can be represented as:
(18) It is silent* in the house.
Silent* is what Carston (199?) calls an ad hoc concept. This ad hoc concept is a constituent of the
proposition I express by my utterance of (16). It is the concept I intend my hearer to recover by a
pragmatic process of loosening. The hearer must use the semantically encoded content of the
word ‘silent’ as one clue, along with other assumptions that are mutually manifest to speaker and
hearer in the context, in order to understand what-is-saidC. The concept silent* is not one of the
lexicalized meanings of the English word ‘silent’ and there is no other word in English whose
lexical meaning corresponds to this concept. We could decide to introduce a new term into our
language whose meaning corresponded to this concept. But there isn’t much point to doing this,
as it is easy enough to express the ad hoc concept simply by using the lexicalized concept as a
springboard and relying on context to fine tune one’s meaning. Moreover, if we introduce a term
for silent*, then why not for all the other possible ad hoc concepts that can be conveyed by (16)
by suitably altering the context? Since there are indefinitely many such concepts, this would be an
exercise in futility.
It is worth making a few remarks here that will have a bearing when we come to discuss
metaphorical uses of expressions. Firstly, the expressibility of such ad hoc concepts does not
depend on the possibility of putting such concepts into words, even in principle. An ad hoc
concept is expressible even in situations in which there is no way in principle of fully
linguistically characterizing the concept, so long as it is possible to convey part of the content by
extra-linguistic means. It may be that part of the individuation conditions of the concept I wish to
express is some feature in my perceptual environment that can only be demonstratively picked
out. In other words, my concept may be a de re one. Such a concept can never be fully
linguistically characterized, in the way that a purely descriptive concept can be, at least in
principle. Nevertheless I can express this de re concept and my hearer can understand me to have
16
expressed this concept so long as the perceptual feature is part of my hearer’s and my mutual
cognitive environment.4
Secondly, as already mentioned, there are many possible ad hoc concepts corresponding to
‘silent’ that can be expressed by uttering (16) in some appropriate context. For instance, different
contexts may set different audibility thresholds, and to each such threshold would correspond a
different ad hoc concept. But it needn’t be only audibility thresholds that vary from context to
context. Another feature that might vary from context to context is what sorts of sounds are
relevant. For instance, imagine there has just been a heated battle in our city. There have been the
sounds of shells and mortars exploding, bursts of gunfire and so on. We emerge from our
underground shelter and listen. No more battle sounds can be heard, although there is an audible
sound of flowing water, the result of a ruptured water main. In such a situation I might utter (16).
I intend my hearer to understand that my use of ‘silent’ is a loose use, which excludes only
clearly audible battle sounds, but doesn’t exclude other sorts of audible sounds, such as the sound
of a gushing water main. (Compare: ‘All’s quiet on the Western Front’).
Thirdly, the talk of ad hoc concepts is just as applicable to cases in which what-is-saidC is
arrived at by a process of enrichment. When I utter (12) and thereby express (13) one could say
that I’ve expressed the ad hoc concept of having been to the bank*, which is the concept of
having been to the monetary depository on a particular Monday morning at a particular time. And
here too there are indefinitely many such ad hoc concepts expressible by the utterance of (12) in a
suitable context, and some of these concepts may be in principle inexpressible in words.
Fourthly, even though I have spoken of the processes of enrichment and loosening as though
they operate separately, it may be that in many situations they operate together. In other words, to
understand the ad hoc concept I intend to express, it may be necessary to both loosen along some
dimensions and enrich along others. Carston (1997: 113-116) gives several examples to illustrate
this possibility. In future, when I talk of loose use I mean to include cases in which loosening and
enrichment operate together.
Fifthly, besides the pragmatic processes of enrichment and loosening, contextualists have
discussed various other sorts of pragmatic processes that might play a role in generating aspects
of what-is-saidC. For instance, Recanati (1993) describes a process that he calls transfer. More
4
When I talk of a person’s cognitive environment I mean any information of which that person can
become aware, even if generally such information is not attended to. Perceptual states encode
information of which we can become aware, so that the contents of such states belong to our cognitive
17
will be said about this in the next section in which I consider the ways in which my direct
expression view of metaphor differs from Stern’s (2000) view of metaphors as demonstratives.
Before moving on to explain how this discussion of pragmatic enrichment and loosening
bears on the case of metaphors, it is worth dispensing with one possible objection to the
contextualist conception of saying I’ve been trying to characterize. This is an objection that has
been pressed by Bach against those, such as Carston (1991), Recanati (1991, 1993) and Travis
(1985), who defend contextualist conceptions of saying. According to Bach (1999) these people
have confused saying, which is a locutionary act, with stating, which is an illocutionary act.
However, I am happy to agree that saying is to be identified with a locutionary act. Where I differ
from Bach is on how to define a locutionary act.
Bach accepts Austin’s (1975) definition of a locutionary act. According to Austin (1975: 94),
when a speaker utters a sentence in a certain context, this counts as a locutionary act only if these
words are used with a certain sense and reference. It is plausible to think that Austin was
operating with the still widely shared assumption that once we’ve used the context to determine
sense (by disambiguation) and reference (by reference assignment) we have determined the
proposition that is expressed by the speaker’s utterance of the sentence in that context. In other
words, Austin thinks that locutionary acts are individuated by their propositional contents, which
in turn he thinks are fixed simply by disambiguation and reference assignment.
However, this latter assumption fails to accommodate cases in which the syntactic and
semantic information encoded in a sentence underdetermines the proposition expressed by the
utterance of that sentence in some context. In such cases, a process of pragmatic enrichment is
called for in addition to disambiguation and reference assignment. There are also cases in which
what is semantically encoded is too narrow or too broad, in which case loosening or enrichment
(which is a type of narrowing) are called for respectively. We need to amend Austin’s definition
of a locutionary act as follows: A locutionary act is the uttering of certain words with a certain
sense and reference and pragmatically enriched/loosened in a certain way. Once we have done
this it is a simple matter to agree that to say something is to engage in a locutionary act.
How do the notions of what-is-saidC and of pragmatic enrichment and loosening apply to
cases of metaphor? Sperber & Wilson (1986) suggest that loose use and metaphor are phenomena
of essentially the same sort. Mundane cases of loose use such as (16) and (17) are at one end of a
continuum with highly poetic metaphors at the other end. When Sperber & Wilson first proposed
environments just as much as the contents of our beliefs and other propositional attitudes. ‘Cognitive’
is not being used synonymously with ‘conceptual’.
18
the idea that the metaphorical use of a sentence can be profitably thought of as a certain sort of
loose use of that sentence, their conception of loose use was more in line with neo-Gricean views
about the nature of metaphor. Sperber & Wilson (1986) held that in cases of loose use there is a
proposition expressed by the sentence the speaker uses (what neo-Griceans would say is literally
expressed by that sentence) but that this expressed proposition is not identical to the loose
proposition that the speaker intends to communicate. The loose interpretation is not directly
expressed but is merely a Gricean conversational implicature.
Carston (1997) points out that this means treating the pragmatic processes of enrichment and
loosening in asymmetrical ways. She argues that it is more in line with Sperber & Wilson’s
theory of relevance to regard enrichment and loosening as symmetrical processes. One should
think of them as processes that move in opposite directions away from what is semantically
encoded in the sentence uttered. Enrichment narrows the encoded content, loosening broadens the
encoded content, but both processes should be thought of as resulting in a pragmatically derived
proposition which is directly expressed by the sentence as uttered in some context. There is no
need to invoke Gricean conversational implicatures. In my discussion of loosening, I have been
following Carston’s (1997) conception of the process, and hence in thinking about metaphor as
loose use I will follow Carston (1997) too.
In what sense can metaphor be thought of as loose use? Consider the following passage,
which describes the emotions Duncan Cambus experiences after learning that his wife Jean is
leaving him for the second time to return to her lover David Crimond:
(19)He [Duncan] groaned, feeling, smelling, as it came bubbling to the surface, all that old
murderous jealousy and hate which had been packed away, a dangerous atomic capsule,
submerged for so long in the darkest sea caverns of his mind. (Iris Murdoch, The Book
and the Brotherhood. Penguin Books, London, 1988, pg. 76-77).
Duncan’s emotions of hate and jealousy are here compared to dangerous atomic weapons. Atomic
weapons can lay waste to whole cities (as they did in Hiroshima and Nagasaki), and the damage
they do is indiscriminate. Innocent civilians and their homes and businesses are destroyed along
with military targets and enemy soldiers. We therefore form an ad hoc concept of things that are
hugely destructive and that are not discriminating about what they damage. It is this ad hoc
concept that is applied to the emotions of jealousy and hate. Note that the emotions here are being
used metonymically to stand for the person who has those emotions. Presumably it is people who
do damage because they are jealous, rather than the mental states of jealousy themselves that do
the damage.
19
There is the further image of jealousy and hate being packed away deep in Duncan’s
unconscious, in the way that atomic weapons may be stored far beneath the Earth’s surface (in
caverns under the sea). Now this hate and jealousy is rising to consciousness like an atomic
capsule “bubbling to the surface” of the sea beneath which it was buried. Two further ad hoc
concepts must be constructed here. Atomic weapons are kept in secure places where they are not
easily accessible, to guard against their being accidentally triggered, because they are so
dangerous and destructive. Thus we form the ad hoc concept of keeping things so inaccessible
that it is no easy matter to retrieve them. Duncan’s feelings of jealousy and hatred for Crimond
have been repressed, and are inaccessible in this way.5
The image of an atomic capsule “bubbling
to the surface” suggests that the capsule is not being raised to the surface in some controlled and
deliberate way, and it is not being shot out of the water, say at the head of a guided missile. Thus
we form an ad hoc concept of something ominous, stealthy, out of control which is applied to
Duncan’s emotions. All these ad hoc concepts are constituents of the proposition that is directly
expressed by (19).
Consider another example, from Shakespeare’s Othello:
(20) Iago: And his [Othello’s] unbookish jealousy must construe / Poor Cassio’s smiles,
gestures and light behaviors / Quite in the wrong. (Othello, IV, i, 9-11)
As in the passage from Murdoch, jealousy is being used metonymically to stand for the person
who has the emotion, in this case Othello. It is Othello who will misconstrue Cassio’s behavior,
not the mental state of jealousy itself. An unbookish person is one who is not devoted to books or
study and who is not familiar with the academic way of life. An unbookish person is likely to
have difficulty with reading and understanding what he reads. Shakespeare frequently refers to
people’s faces and actions metaphorically as texts that must be read and interpreted. Here too
Cassio’s behavior is being treated metaphorically as a text that must be read. From the context we
also know that Othello is a military man, unfamiliar with the mannerisms of the Venetian court.
Thus we form the ad hoc concept of someone who is not good at understanding a certain sort of
courtly behavior. Presumably, being in the grip of the emotion of jealousy, Othello’s interpretive
5
Of course, talk of repressed emotions and of storing things in one’s unconscious is itself a
metaphorical way of talking of emotions. But this sort of Freudian conception of the mental is so much
part of our everyday talk of the mind that it is hardly recognized as metaphorical. It is also possible to
see all these metaphors, including Murdoch’s more colorful ones, as linguistic expressions of a couple
of underlying conceptual metaphors in Lakoff and Johnson’s (1980) sense, namely MENTAL STATES
ARE OBJECTS and THE MIND IS A CONTAINER.
20
task is made even more difficult, as he is unlikely to stop to consider whether his judgments about
Cassio are warranted.
Both (19) and (20) seem to pattern with the example of loose use discussed above. Thus we
have good reason to think of metaphors as directly expressing certain propositions, just as
sentences used loosely directly express propositions. Of course, the more poetic the metaphor, the
more difficult it is to be sure we have correctly represented the content. Metaphors also have an
open-ended quality to them. If a metaphor is a good one, then the more one thinks about it, the
more one seems to discover in it, so that one can never be sure that one has exhausted its content.
This is true, but this observation doesn’t count against the direct expression view of metaphor that
I am attempting to defend.
Firstly, it doesn’t help to retreat to an indirect communication view of metaphor. Those who
defend the view that metaphorical interpretations are indirectly conveyed via literally saying
something else nevertheless want to hold that some proposition is indirectly communicated. Thus
they too must face the problem that it is sometimes hard to say what this proposition is, due to the
open-ended nature of metaphorical interpretation.
But secondly, the open-ended nature of metaphorical interpretation isn’t an embarrassment.
Literal interpretation is just as open-ended. Any difficult claim that requires the mastery of a lot
of background information for its interpretation is likely to have an open-ended quality. The more
time one can spend thinking about the claim and the more background one can bring to bear on it,
the greater one’s understanding is likely to be. As one processes at a deeper level one is likely to
read more into the claim and hence to revise one’s initial interpretation(s).
When discussing the loose use of (16), I remarked that there are indefinitely many
propositions that can be expressed by (16), depending on the context in which it is uttered.
Similar remarks can be made about metaphors. For instance, Shakespeare was fond of
metaphorically comparing his characters to the Sun. But by situating this metaphor in different
contexts, what is expressed is never the same. Consider for example the following:
(21) Romeo: But soft, what light through yonder window breaks? / It is the East, and Juliet is
the sun. / Arise fair sun and kill the envious moon, / Who is already sick and pale with
grief, / That thou her maid art more fair than she. (Romeo and Juliet II, ii, 2-6)
(22) Salisbury: Ah, Richard, with the eyes of heavy mind / I see thy glory like a shooting star/
Fall to the base earth from the firmament. / Thy sun sets weeping in the lowly west, /
21
Witness storms to come, woe and unrest; / Thy friends are fled to wait upon thy foes, /
And crossly to thy good all fortune goes. (Richard II, II, iv, 18-24)
(23) Cleopatra: The crown o’ th’ earth doth melt. (Antony and Cleopatra, IV, xv, 73)
I will not discuss these examples in detail, but just briefly remark on some of their features. The
comparisons in (21) and (22) of Juliet and Richard II to the Sun are very different. Juliet is
compared to the rising sun, whereas Richard is compared to the setting sun. Juliet is a “fair sun”
whereas Richard is a “weeping sun”. The surrounding contexts are very different. Juliet is being
praised by a besotted Romeo who is trying to woo her. Salisbury has been told that Richard is
dead and is lamenting all the evil consequences of this event. These different background
assumptions will give rise to very different ad hoc concepts to be applied to Juliet and Richard
respectively.
White (1996: 176) very beautifully describes example (23). Cleopatra utters (23) just at the
moment Antony dies. White argues that this is a double metaphor. The Sun is compared to a
golden crown and Antony is compared to the sun. As the sun sets its light is diffracted and
spreads out along the horizon like molten gold. Antony’s death is then compared to such a setting
sun and hence to a melting crown. The set of assumptions here is very different again from either
(21) or (22). Both Richard and Antony are compared to setting suns, but the background
assumptions give rise to different ad hoc concepts. Antony’s political greatness is emphasized,
his “preciousness” and there are none of the evil portents (shooting stars falling to Earth) as in the
case of Richard. (I am talking just about the immediate context for (23). In other places in the
play Antony’s death is compared to the falling of a star that halts time. A&C, IV, xiv, 129f).
I want to end this section by discussing a distinction Stern (2000: 27-28) draws between what
he calls extension and transfer. Cases of extension are metaphorical interpretations “that extend
the (literal) meaning of an expression, say, by dropping at least one condition in the (literal)
meaning” (2000: 27. Stern’s emphasis). Cases of transfer on the other hand are metaphorical
interpretations that involve a change of extension; “the resulting interpretation is applicable to a
domain disjoint from its original [literal] one (or sufficiently disjoint for the change to count as
one of transfer).” (2000: 27). Stern gives (24) as an example of extension, and (25) as a case of
transfer:
(24) Quine demolished Carnap’s argument.
(25) Juliet is the sun.
22
Stern grants that both sorts of uses are metaphorical uses. However, he claims that his account of
metaphor is focused on cases of transfer, implying that these are the more interesting cases, and
that an account of metaphor based just on cases of extension would be somehow impoverished.
Since his definition of extension may seem to suggest what I’ve been calling loose use, and since
my direct expression view of metaphor has made the notion of loose use central, we need to ask
whether there is an important distinction here that my account is missing.
Stern says that in cases of extension one or another of the conditions of application of the
extended term is dropped, but that the metaphorical interpretation “properly contains” the original
(literal) one. He says that what condition is to be dropped will vary from context to context, and
so metaphorical extensions are in some sense context-dependent. But he also thinks they are
“context-independent insofar as they do not draw upon extra-linguistic presuppositions for the
content of the metaphorical interpretation.” (2000: 27. His emphasis). How might this work in the
case of (24)? Well, to demolish is to destroy or tear down a physical structure, such as a building
or a bridge or a dam wall. If we drop the condition that the structure torn down be a physical
structure, then the term will be applicable to any kind of structure, including such abstract
structures as arguments or proofs or ideologies. It looks as though we’ve simply broadened the
extension of the predicate, and that all we needed to do was access the lexical entry for the verb
‘to destroy’ and eliminate one of the semantic features associated with the term, viz. [+thing
destroyed is physical].
However, I do not agree that this is how we interpret (24). Of course, Stern provides us with
absolutely no context for the utterance of this sentence, which is one of the things I have already
complained is often missing in philosophical discussions of metaphor. Even without too much in
the way of background I would argue that many extra-linguistic presuppositions enter into the
interpretation of (24). For example, although the verb ‘to destroy’ is not subcategorized for an
obligatory following prepositional phrase, talk of demolishing things will bring to mind various
sorts of instruments and implements used in demolition jobs. We may think of wrecking balls and
dynamite and huge earth moving equipment. What specifically we think of will depend on the
size of the thing to be torn down. A dam wall or a huge building will probably call for large
amounts of dynamite, but if it is simply a rickety old wooden outhouse that needs to be torn
down, then probably a single person wielding an axe can accomplish the task. In the case of (24)
we know that Carnap was a good philosopher whose arguments were generally well constructed.6
6
Note that here again we could invoke an underlying conceptual metaphor of the Lakoff & Johnson
sort. In this case the metaphor would be ARGUMENTS ARE BUILDINGS.
23
So refuting his arguments would be no easy matter. Presumably Quine was up to the task, being
himself no slouch. But Quine and Carnap were also pretty well matched in intellect. For Quine to
demolish Anne’s arguments would be an easy matter, requiring no heavy-duty logical apparatus.
In the case of Carnap, Quine might have to muster argumentative techniques of a more
sophisticated sort.
I would argue that by relying on such non-linguistic background assumptions we arrive at an
ad hoc concept demolish* and it is this concept that is applied to Quine’s arguments against
Carnap. This concept is no mere extension in Stern’s sense of the meaning of ‘demolish’, and it
certainly does not “properly contain” the literal interpretation of the term. This ad hoc concept
would not be applicable to the act of tearing down a building. Thus I dispute Stern’s claim that
cases like (24) are any different from cases like (25), that he calls cases of transfer. In the case of
(25), as I have already indicated in my discussion of Shakespeare’s sun metaphors, we construct
an ad hoc concept sun* that is applicable to Juliet and would not be applicable to the Sun. But
once again, this doesn’t distinguish (24) from (25). Thus I conclude that by appealing to the
notion of loose use I have not restricted myself to the discussion of an uninteresting sub-class of
metaphorical interpretations and precluded myself from the discussion of the truly interesting
cases.7
5. Metaphors as demonstratives
In this section I want to discuss in some detail the account of metaphor offered by Stern
(2000), and to compare it to my own direct expression view. Stern too can be thought of as
offering a direct expression view of metaphor, but whereas I have argued for this conclusion by
assimilating metaphors to cases of loose use, he argues for this conclusion by assimilating
metaphors to demonstratives. Although there are aspects of Stern’s view that I am sympathetic to,
I think it is wrong to think of metaphors as demonstratives. I will briefly characterize Stern’s
view, and then say why I think it is preferable to think of metaphors as cases of loose use.
Stern and I are in agreement that expressions used metaphorically (whether these be words,
phrases or whole sentences) are context-dependent expressions.8
Thus we wish to think of
7
The view that metaphors involve ad hoc concepts is similar to the view of metaphor defended in
Glucksberg & Keysar (1993). They argue that simple predicative metaphors like ‘My job is a jail’ or
‘Cigarettes are time bombs’ are class inclusion statements. A class inclusion statement is one in which
an object is included in a category or a category is included within a superordinate category. They
argue that when novel metaphors are used new categories are created.
8
Stern (2000) focuses on what he calls predicative metaphors (such as ‘Juliet is the sun’) and briefly
discusses what he calls nominative metaphors (such as ‘The sun is furious’). He thinks the latter can be
24
metaphors on the model of other context-dependent expressions. Stern appears to think that we
have only two choices. Metaphors are either like ambiguous expressions or like demonstrative
and indexical expressions. Ambiguous expressions are context-dependent, in the sense that
context is needed to select the speaker’s intended meaning. Stern and I agree that semantic
ambiguity is not a good model for metaphor. As I noted at the beginning of section 3,
constructing a metaphorical interpretation is not a case of sense selection but of sense creation. So
Stern is left with indexicals and demonstratives as the only model for metaphors.
On the other hand, I have tried to argue that disambiguation and reference assignment are in
many cases not sufficient to determine what is said. We must grant that pragmatic processes of
enrichment and loosening are (at least sometimes) needed to determine propositional constituents.
Thus it is open to me to think of metaphors as context-dependent expressions that directly express
propositions, but that are nevertheless not indexical or demonstrative expressions.9
The context
dependence may be due to the fact that some propositional constituents are fixed via the
pragmatic processes of enrichment and/or loosening. These aspects of content go beyond
anything that is syntactically or semantically encoded. This is the sense in which metaphor is a
pragmatic phenomenon. So, contrary to Stern’s claim, metaphor is not something to be handled
within indexical semantics.
Stern’s account of metaphor is modeled on Kaplan’s (1989a) discussion of indexicals and
demonstratives. In particular, it makes use of Kaplan’s distinction between character and content,
of Kaplan’s notion of a ‘Dthat’ operator, and of Kaplan’s idea that pure indexicals are parametric
expressions whose values are fixed relative to a context via a semantic rule. Stern argues that
there is a general operator ‘Mthat’ which when applied to a predicate expression ‘Φ’ forms a
metaphorical expression ‘Mthat[Φ]’. This is a context-dependent expression that has a certain
metaphorical character ‘{Mthat[Φ]}’. This character is a function from context to content.
reduced to special cases of the former. But he has to grant that expressions of all types, including noun
phrases, verb phrases, prepositional phrases, adjectives and adverbs, can be used metaphorically. Also,
whole sentences may be used metaphorically (e.g., ‘The crown o’ th’ earth doth melt.’).
9
Note that the issue here is what the best model for metaphor is. In denying that the best model is an
indexical model I am not denying that indexicals and demonstratives can sometimes be used
metaphorically. I think that all kinds of expressions can be used metaphorically, including
demonstratives. For example, I can utter ‘That old gin-soaked rag needs to be wrung out’, intending to
use the demonstrative expression ‘that old gin-soaked rag’ metaphorically to refer to one of my
colleagues. I would say that the use of the expression ‘gin-soaked rag’ is a loose one, and in context
you will retrieve an ad hoc concept gin-soaked rag* that will help you to determine my reference.
Another way of characterizing the difference between Stern’s view and mine is to note that for him
even a non-indexical expression ‘Φ’ is turned into an indexical one by the application of his ‘Mthat’
operator.
25
Relative to some context of interpretation, this function determines a set of properties P, which is
the interpretation or content of the metaphorical expression in that context. The set P =
{Mthat[Φ]}(c), that is the value in context c of the metaphorical character that is associated with
the metaphorical expression ‘Mthat[Φ]’. These properties are constituents of the proposition that
is expressed by the sentence containing that metaphorical expression. This proposition can in turn
be evaluated as either true or false. (As previously noted, Stern’s account is in the first instance an
account of predicative metaphors. He later tries to extend the account to nominative metaphors,
by reducing these to a special case of predicative metaphors).
The set of properties assigned by the metaphorical character of an expression ‘Mthat[Φ]’ as
the value of that expression in a context c is the set of properties presupposed to be m-associated
with ‘Φ’ in c. Which properties are m-associated with ‘Φ’ in c is not something that is determined
by the semantics of the language. It is an entirely pragmatic affair, and Stern is willing to grant
that in different kinds of cases there may be different sorts of pragmatic mechanisms at work. He
mentions three such possible pragmatic mechanisms, one involving what he calls exemplification
schemas, one based on thematic frameworks, and one based on the idea of inductive/associative
networks. He devotes the most space to explaining the mechanism of exemplification, which he
claims is involved in simple predicative metaphors of the ‘Juliet is the sun’ sort.
I cannot go into any detail here, but in rough outline the idea of exemplification is that in
using an expression ‘Φ’ metaphorically the speaker is trying to focus the hearer’s attention on a
set of properties. She does this by getting the hearer to recognize (by relying on certain shared
presuppositions) that the object picked out by the literal vehicle ‘Φ’ exemplifies that set of
properties. The exemplified properties will be the properties m-associated with ‘Φ’ in the relevant
context.
A non-linguistic case of exemplification is when one presents someone with a sample (say a
swatch of material in a tailor’s booklet). In presenting the sample one exemplifies (makes salient)
a set of properties. Each swatch works as a sample not in isolation but as part of a whole schema.
So what properties of the swatch will be salient will depend on the other members of the schema.
For instance, whether it is the color or the texture of the sample that one is to focus on will
depend on what other samples belong to the exemplification schema.
Something analogous is supposed to be true in the case of the metaphorical use of an
expression. Thus consider again:
(25) Juliet is the sun.
26
The speaker is trying to make salient a set of properties that is exemplified (relative to the
presupposed context) by the object that belongs to the extension of the literal vehicle ‘is the sun’,
viz. the properties exemplified in the relevant context by the Sun. What properties these are will
depend on the exemplification schema to which the Sun belongs. To determine this we need to
look at the literary context of (25). Some of the context in which the metaphor is presented is
given above in (21), where Juliet and Rosaline are compared respectively to the Sun and the
Moon. Several lines later Romeo says:
(26) Two of the fairest stars in all the heaven, / Having some business, do entreat her
[Juliet’s] eyes / To twinkle in their spheres till they return. / What if her eyes were there,
they in her head? / The brightness of her cheek would shame those stars, / As daylight
doth a lamp; her eyes in heaven / Would through the airy region stream so bright, / That
birds would sing, and think it were not night. (Romeo and Juliet, II, ii, 15-22)
Stern therefore suggests that the exemplification scheme in this case is composed of the sun, the
moon, the two stars, daylight, the lamp. The multiple contrasts between these objects will
determine what it is that the Sun exemplifies. He also suggests that there is a second stage of the
interpretation process in which certain filter-presuppositions will be invoked to select and extract
relevant features from the ones generated by the sets of contrasts between the members of the
exemplification scheme. These selected features will be features appropriate to humans, and in
particular appropriate to Juliet. Here we will rely on what we know from the rest of the play about
Juliet’s personality (or more accurately, on what we know the story says that Romeo knows or
believes about Juliet’s personality).
One other aspect of Stern’s account that I’ll briefly mention is his claim that metaphorical
characters share some of their features with those of pure indexicals (‘I’, ‘here’, ‘now’ etc.) and
some of their features with the characters of Dthat-descriptions. Metaphorical characters are like
those of pure indexicals in that they determine their semantic values parametrically. That is, their
semantic values are determined in context by a semantic rule, and not denotationally, as Stern
assumes is the case with Dthat-descriptions.10
For instance, the rule in the case of the indexical ‘I’
is to assign as its value in a context c the agent of c. The rule in the case of a metaphorical
10
Kaplan’s original view was that the reference of a Dthat-description is fixed in context as what is
denoted by the description in that context. However, he later argued that the reference of Dthatdescriptions
is determined by the speaker’s intentions, and that the description was a mere
externalization of the intention, functioning as an aid to communication, like speaking more slowly or
loudly, but having no semantic significance (1989b: 582). The idea that there are pure indexicals
whose reference is determined parametrically by a semantic rule, and without reference to a speaker’s
intentions, has been challenged by Smith (1989) and Nunberg (1993).
27
expression ‘Mthat[Φ]’ is to assign as its value in a context c the set of properties m-associated
with ‘Φ’ in c. On the other hand “what the speaker knows when she knows how to interpret a
metaphor Φ is not simply something about the single expression Φ. Her knowledge of metaphor
is closer to knowledge of an operator, of an interpretive operation she can perform on any
(literally interpreted) expression. And in this respect the character of a metaphor is more like that
of the demonstrative interpretation operator ‘Dthat’ than of that of any individual indexical (type)
such as ‘I’ or ‘here’.” (Stern 2000: 198-199). In other words, with indexicals one has to learn a
separate rule for each expression type, whereas with metaphors we simply learn a single general
operation ‘Mthat’ that we can apply to any expression that can be used metaphorically.
One problematic consequence of Stern’s view is that he has to deny that one and the same
expression type can be both literally and metaphorically interpreted. Consider again an example
that was used in one of the footnotes:
(27) That old gin-soaked rag needs to be wrung out.
On the face of it there looks to be one sentence type that can, depending on context, be used
either literally (say to refer to a cloth that I’ve used to mop up a spill) or metaphorically (say to
refer to an alcoholic who is in need of detoxification). But Stern thinks that the metaphorical
interpretation of an expression ‘Φ’ requires the application of a metaphorical character in context.
This metaphorical character is not the same as the literal character of ‘Φ’. But if we say that the
sentence-type (27) has both a literal and a metaphorical character, then we are committed to its
being ambiguous, and Stern does not want to explain metaphor by appeal to the notion of
semantic ambiguity. So instead he denies that there is a single expression type that has both a
literal and a metaphorical character. What has the metaphorical character is the metaphorical
expression-type:
(28) Mthat[‘that old gin-soaked rag’] needs to be Mthat[‘wrung out’].
When trying to decide how to interpret a particular token expression whose surface realization is
That ∩old ∩gin-soaked ∩rag ∩needs ∩to ∩be ∩wrung ∩out, we have to decide what type it
belongs to. We have to assign it either to type (27) or type (28). Once we’ve decided this,
different semantic procedures will apply. Interpretation will proceed by applying either a literal
character or a metaphorical character to the context to yield its content.11
11
Stern (2000: 133) writes as though the literal character of an expression ‘Φ’ is always what Kaplan
(1989a: 506) called a fixed character, namely one that determines the same content in every context,
whereas the metaphorical character ‘{Mthat[Φ]}’ is what Kaplan calls a context-sensitive character,
28
This is to make an invidious distinction between literal and metaphorical interpretation that
doesn’t seems to exist. I agree with Rumelhart (1993: 76) that any account rich enough to give an
account of literal interpretation is already rich enough to give an account of metaphorical
interpretations. I take it to be a virtue of my account that metaphorical uses of language do not
require interpretative mechanisms any different from those used in the interpretation of mundane
loose uses.
Stern (2000: 317) argues that literal interpretation is relatively independent of context,
whereas metaphorical interpretation is context-dependent. He wants to reject a continuum
between the literal and the metaphorical. However, it seems that here Stern fails to take seriously
enough his own earlier claim that we must distinguish between literal word (and sentence)
meaning and the literal interpretation or understanding of a word or sentence. (2000: 64) Literal
meaning is what is semantically encoded in an expression, but interpretation is sensitive not only
to linguistically encoded information but also to information available in the context of utterance,
including non-linguistic background information. All interpretation is context dependent, and all
interpretation is constrained by literal word meaning. There is simply a difference in degree to
which interpretations depart from what is semantically encoded in the sentence uttered. Where the
departure is minimal the interpretation will seem more literal than in cases where the departure is
substantial. Thus cases of the sort Stern (2000: 27) calls transfer will seem more metaphorical
than cases of the sort he calls extension.
It also seems wrong to say that generally speaking sentences encode (either explicitly or
implicitly) whether they are to be interpreted literally or metaphorically. In the case of a sentence
that could be interpreted either way (i.e. one where no category mistake arises from interpreting it
strictly and literally) the sentence does not encode how it is to be interpreted. Instead, the
speaker’s intentions play a crucial role here, and provided the hearer focuses on the relevant
background assumptions, ordinary interpretive processes, including enrichment and loosening,
will enable the hearer to figure out the proposition expressed by the speaker’s utterance in that
particular context. Stern on the other hand, is committed to the idea that even if on the surface
sentences do not encode how they are to be interpreted, at deep structure sentences do encode
this, because at deep structure we have two different sentence types, literal and metaphorical.
However, Stern doesn’t give convincing arguments to support his claim that when sentences
are interpreted metaphorically there is an underlying deep structure different from literal deep
namely one that determines different contents in different contexts. However, the literal character of
the demonstrative expression ‘that old gin-soaked rag’ is not a fixed character.
29
structure, and that this deep structure is associated with a metaphorical character different from
literal character. The phenomena he points to in support of his claims are more parsimoniously
explained simply by distinguishing literal meaning from literal interpretation in the way indicated
in the previous two paragraphs. I will briefly examine two of the arguments Stern (2000: 69-71,
215-217) offers in support of his demonstrative model of metaphor. These arguments are meant
to establish that we need to posit metaphorical characters in addition to literal characters.
One argument depends on cases of verb phrase anaphora in which certain “crossed”
interpretations are ruled out. Consider the following;
(29) The largest blob of gases in the solar system is the sun.
(30) Juliet is the sun.
(31) Antony is the sun.
*(32) The largest blob of gases in the solar system is the sun and Juliet is too.
?(33) Juliet is the sun and Antony is too.
(29) is to be understood literally, and (30) and (31) are to be understood metaphorically, but let us
assume that their metaphorical contents differ in the ways already indicated when these examples
were discussed in section 4. Given these interpretations, the cases of verb phrase anaphora in (32)
and (33) are unacceptable, and (32) seems more so than (33). Stern (2000: 216) claims that the
best explanation of these interpretive constraints is that ‘is the sun’ in (29) has a literal character
different from the metaphorical character that ‘is the sun’ – or more strictly that ‘Mthat[‘is the
sun’]’ – has in (30) and (31). So the predicate expressions in (29) and (30) have both different
characters and different contents. On the other hand, the predicate expressions in (30) and (31)
have the same character and differ only in content. This explains the fact that (33) is not quite as
unacceptable as (32).
However, it is by no means clear that the unacceptability of such crossed interpretations is to
be explained in this way. Firstly, it is by no means clear that ‘is’ is being used univocally in (29) –
(31). In (29) we appear to have the ‘is’ of identity, as is clear from the fact that we can reverse the
order of the terms on either side of ‘is’. (29) says the same thing as is said by ‘The sun is the
largest blob of gases in the solar system’. However, the ‘is’ in (30) and (31) is the ‘is’ of
predication, because these do not say the same things as ‘The sun is Juliet’ and ‘The sun is
Antony’ respectively. This equivocation would be enough to explain why (32) is more
unacceptable than (33).
30
Secondly, even if we grant for the sake of argument that no equivocation is going on, a
similar pattern of results about the unacceptability of crossed interpretations holds in cases of
loose use. Consider the following:
(34) It is silent in this specially constructed experimental chamber.
(35) It is silent in this house.
(36) It is silent on this street.
*(37) It is silent in this specially constructed experimental chamber and in this house too.
?(38) It is silent in this house and on this street too.
Suppose that ‘silent’ in (34) is interpreted strictly to mean the absence of all sounds, whereas in
(35) and (36) it is interpreted loosely, although in the different ways that I imagined when
discussing example (16) in section 4. (35) is to be interpreted in a way that allows there to be
barely audible sounds such as humming refrigerators and dripping taps but not to include louder
sounds of any sort. (36) on the other hand is to be interpreted in a way that allows audible sounds
of water gushing from a water main and disallows only loud battle sounds, such as machine gun
fire, exploding mortars and so on. Given these interpretations, (37) and (38) are unacceptable,
although arguably (37) is more unacceptable. This cannot be explained by saying that the
expression ‘is silent’ has a different character in (34) from the one it has in (35) and (36). In all
three sentences the phrase has the same literal meaning. What differs is whether an ad hoc
concept is intended and if so which one it is. Thus the explanation must be that mixing strict and
loose uses is more unacceptable than mixing loose uses.
This suggests that a similar explanation will be forthcoming in the metaphorical case, given
that metaphor and loose use are closely related phenomena. We do not need to invoke the notion
of a metaphorical character to explain why certain crossed interpretations are ruled out and why
some seem worse than others.
A second argument that Stern gives in favor of his demonstrative account of metaphor is to
argue that metaphors behave like demonstratives in modal contexts. The truth-conditional content
of a demonstrative or indexical is always determined relative to its context of utterance, never
relative to the counterfactual or alternative possible circumstances in which we evaluate its truth.
According to Stern (2000: 71) the same thing is true of metaphors. Suppose Paris disagrees with
Romeo’s utterance of ‘Juliet is the sun’, but he is willing to concede the following:
(39) Juliet might have been the sun.
31
Stern claims that this requires that ‘is the sun’ be interpreted metaphorically in the same way in
Romeo’s and Paris’ utterances. This in turn means that the predicate expression in (39) must be
assigned the set of properties that are m-associated with ‘is the sun’ in the context of utterance. It
cannot be that (39) is true in virtue of the fact that Juliet has the set of properties that ‘is the sun’
would be assigned if it were interpreted metaphorically in some other possible circumstance.
Stern claims that only if we posit metaphorical characters that behave like the characters of
demonstratives and indexicals can we explain this constraint on acceptable interpretations of (39).
Once again, this does not seem to be true. The most obvious explanation for the constraint on
the interpretation of (39) is that Romeo and Paris, if they are to be genuinely disagreeing, must be
disagreeing about the same proposition. Thus we cannot interpret ‘is the sun’ in (39) in a different
way (i.e., interpret it as involving a different ad hoc concept), because then Romeo and Paris
would be talking at cross-purposes.
But secondly, it simply seems false that metaphors are always to be interpreted relative to
their actual context of utterance and not at the counterfactual circumstances at which they are
being evaluated. Consider:
(40) The shrinking violet might have been more gregarious.
Suppose the definite description is being used metaphorically to denote a person, not a plant.
There seem to be two possible interpretations here, depending on whether the definite description
or the modal operator takes wide scope. On one possible reading the description takes wide
scope, and then its interpretation is the one it has in the actual context. Suppose it denotes Susie.
Then (40) is true in some possible circumstance of evaluation just in case Susie is gregarious in
that circumstance. On the other hand, if the modal operator takes wide scope, then the description
‘the shrinking violet’ will be interpreted relative to the circumstance of evaluation, and (40) will
be true just in case whoever is picked out in that circumstance is gregarious in that circumstance.
I conclude that Stern’s reasons for assimilating metaphors to demonstratives are by no means
compelling. The sorts of constraints on interpretation that he argues push us towards the
demonstrative model of metaphor can be accounted for without invoking metaphorical characters
in addition to literal characters. All we need is a distinction between literal sentence meaning
(what is semantically encoded) and (literal or metaphorical) interpretation, as well as the
interpretive mechanisms of pragmatic enrichment and loosening.
32
6. Some Objections Considered
In this final section I consider a couple of objections that might be raised against my direct
expression view. The first is Tsohatzidis’ (1994) objection that metaphorical interpretations are
not always cancellable. He takes this to count against pragmatic accounts of metaphor. Since my
direct expression view treats metaphor as a pragmatic phenomenon of the same kind as the
phenomenon of loose use, Tsohatzidis’ objection is a potential problem for my account.
I begin with three observations. Firstly, cancellability was a test that Grice (1989: 44) devised
to determine whether some content p is semantically encoded in a sentence or merely
conversationally implied by the utterance of that sentence in some context. However, the test can
be applied more generally to any content that is pragmatically determined and that goes beyond
what is syntactically and semantically encoded. Thus it should apply to content determined via
pragmatic enrichment and loosening. Secondly, as I noted in section 3, metaphorical
interpretations are cancellable in favor of a literal interpretation in cases in which the sentence
uttered is capable of a literal interpretation (E.g., ‘He is a butcher, and I don’t mean because of
the way he carries out his surgery. He moonlights down at the abattoir’). Thirdly, the sorts of
cases in which the metaphorical interpretation is not cancelable in favor of a literal interpretation
are cases in which a literal interpretation of the sentence uttered would involve some sort of
category mistake. And here we need to distinguish two sub-cases. Sometimes the literal
interpretation is ruled out even at the level of sentence meaning, because the encoded meaning is
semantically anomalous. But sometimes literal sentence meaning can be non-anomalous, yet
given certain interpretive decisions with respect to some of the component expressions, a literal
interpretation of the whole sentence is ruled out.
Thus it does not seem possible, keeping the references of ‘she’ and ‘the Taj Mahal’ fixed, to
cancel the metaphorical interpretation of the following in favor of a literal interpretation:
(41) She is the Taj Mahal.
The literal meaning of (41) by itself does not rule out a literal interpretation. After all, we
sometimes use ‘she’ to refer to inanimate things such as ships, and it could be that (41) is to be
interpreted as being about a ship named ‘the Taj Mahal’. However, given that ‘she’ refers to a
woman and ‘the Taj Mahal’ refers to a certain building in India, we are constrained to find a
metaphorical interpretation of (41). This metaphorical interpretation cannot be cancelled in favor
of a literal one, given these interpretive constraints. This does not mean that it cannot be cancelled
at all. It can be cancelled in favor of another metaphorical interpretation:
33
(42) She is the Taj Mahal, and I don’t mean that she is beautiful. Rather, she is someone her
countrymen are proud of.
(41) does not lend support to the claim that the sentence a speaker utters semantically encodes its
metaphorical interpretation, nor to the claim that metaphor is a semantic rather than pragmatic
phenomenon.
It might be thought that examples of the following sort are more problematic for anyone who
wants to maintain that metaphor is a pragmatic phenomenon:
(43) Humans are spiders.12
The literal meaning of the sentence is semantically anomalous, on the assumption that the lexical
entries for ‘human’ and ‘spider’ contain semantic features that are incompatible with one another
(e.g., [+belongs to phylum Chordata] versus [+belongs to phylum Arthropoda] respectively). One
might say that the information semantically encoded in the sentence gives a strong clue that the
sentence is to be interpreted metaphorically (if one assumes that the speaker is semantically
competent and isn’t simply misusing words). Does this mean that we have to give up the idea that
metaphorical interpretation is a pragmatic affair and grant instead that it is purely semantic? No,
all it means is that what sort of interpretation is required is sometimes over-determined.
Sometimes semantically encoded information by itself is neutral as to whether the sentence
uttered is to be interpreted literally or metaphorically, and then information about the speaker’s
intentions will be decisive. But sometimes semantically encoded information can point strongly
in the direction of a metaphorical interpretation. This does nothing to support the claim that any
specific metaphorical interpretation is semantically encoded. Moreover, even though a specific
metaphorical interpretation is not cancellable in favor of a literal interpretation, it will be
cancellable in favor of another metaphorical interpretation:
(44) Humans are spiders, and not because they are disposed to spin “webs of deceit”, but
because they are capable of trapping and killing things larger than they are.
The final objection that I will consider is that my direct expression view, by denying that
there are metaphorical characters in addition to metaphorical contents, must deny that there are
aspects of metaphorical meaning that are not part of the proposition expressed. Yet it seems as
though there are aspects of metaphorical meaning that go beyond metaphorical content. Consider
the following example, based on one used by Stern (2000: 259). A young woman Marie, who is
12
I was going to use ‘Humans are insects’, but my dictionary lists ‘contemptible person’ as one of the
meanings of ‘insect’.
34
in psychotherapy because she is suffering from anorexia nervosa, tells her therapist that her
mother has forbidden her to see her boyfriend. Referring to her mother’s injunction, Marie utters:
(45) I won’t swallow that.
Here ‘swallow’ is being used metaphorically, and Stern suggests that the content of Marie’s
utterance (the proposition she has expressed) can be paraphrased as:
(45) Marie won’t accept her mother’s injunction.
Given her eating disorder, it seems significant that Marie chose to frame her comment about her
mother’s injunction by using the word ‘swallow’. But once we’ve accessed the metaphorical
interpretation it seems that we’ve lost these echoes of meaning that might connect what she is
saying to her eating disorder and hence to any problems that she might be having with her mother
connected to this disorder. Stern claims that only if we admit that there is a metaphorical
character – a metaphorical “mode of presentation” of the proposition (45) – can we capture these
dimensions of meaning which are not parts of the proposition expressed.
However, that these echoes of meaning are lost from the proposition expressed is due to the
way Stern proposes to represent the proposition expressed by Marie’s utterance of (45). If instead
one represents it as containing an ad hoc concept swallow* that is arrived at via a pragmatic
process of loosening (and possibly also enrichment) from the lexical concept swallow, then one
can see that the proposition expressed is not somehow impoverished:
(46) Marie won’t swallow* her mother’s injunction.
Thus a character/content distinction is not needed to account for the ways in which what Marie
has said by her utterance of (45) is connected to her eating disorder. My direct expression view of
metaphor does not have to ignore relevant dimensions of metaphorical meaning.
Much more would have to be said to provide a full-fledged defense of the view according to
which metaphors directly express propositions with constituents that are pragmatically
determined by loosening (or loosening and enrichment). I hope at least to have motivated the
view, and to have shown how it falls out naturally from a certain contextualist conception of
saying. I also hope to have shown that if one wishes to assimilate metaphor to context-dependent
expressions, an account that identifies metaphors with the phenomenon of loose use is more
parsimonious than one that assimilates metaphors to demonstratives.
35
References:
Austin, J. L. 1975. How To Do Things With Words. 2nd
edition, ed. J. O. Urmson and M. Sbisà.
Cambridge, MA Harvard University Press.
Bach, K. 1994. “Conversational impliciture.” Mind & Language 9: 124-162.
Bach, K. 1995. “Standardization vs. conventionalization.” Linguistics and Philosophy 18: 677-
686.
Bach, K. 1999. “The semantics-pragmatics distinction: What it is and why it matters.” In K.
Turner ed., The Semantics/Pragmatics Interface from Different Points of View. Oxford,
Elsevier: 65-84.
Bach, K. 2001. “You don't say.” Synthese 127: 11-31.
Bertolet, R. 1994. “Are there indirect speech acts?” In S. L. Tsohatzidis ed., Foundations of
Speech Act Theory: Philosophical and Linguistic Perspectives. London, Routledge: 335-349.
Bezuidenhout, A. 2001. “Radical Pragmatics.” In M. O'Rourke, D. Shier and J. Campbell eds.,
Topics in Contemporary Philosophy: Meaning and Truth. New York, Seven Bridges Press.
Bezuidenhout, A. L. and J. C. Cutting 2001. “Literal meaning, minimal propositions and
pragmatic processing.” Journal of Pragmatics. Forthcoming.
Carston, R. 1991. “Implicature, explicature, and truth-theoretic semantics.” In S. Davis ed.,
Pragmatics: A Reader. New York, Oxford University Press: 33-51.
Carston, R. 1997. “Enrichment and loosening: Complementary processes in deriving the
proposition expressed?” Linguistische Berichte: 103-127.
Davidson, D. 1978. “What metaphors mean.” In S. Sacks ed., On Metaphor. Chicago, University
of Chicago Press: 29-46.
Gerrig, R. 1989. “The time course of sense creation.” Memory and Cognition 172: 194-207.
Gibbs, R. W. 1994. The Poetics of Mind: Figurative Thought, Language and Understanding.
Cambridge, Cambridge University Press.
Glucksberg, S. and B. Keysar 1993. “How metaphors work.” In A. Ortony ed., Metaphor and
Thought. Cambridge, Cambridge University Press: 401-424.
Grice, P. 1989. Studies in the Way of Words. Cambridge, MA, Harvard University Press.
Holdcroft, D. 1994. “Indirect speech acts and propositional content.” In S. L. Tsohatzidis ed.,
Foundations of Speech Act Theory: Philosophical and Linguistic Perspectives. London,
London: 350-364.
Kaplan, D. 1989. “Afterthoughts.” In J. Almog, J. Perry and H. Wettstein eds., Themes From
Kaplan. New York, Oxford University Press: 565-614.
36
Kaplan, D. 1989. “Demonstratives: An essay on the semantics, logic, metaphysics, and
epistemology of demonstratives and other indexicals.” In J. Almog, J. Perry and H. Wettstein
eds., Themes From Kaplan. New York, Oxford University Press: 481-563.
Lakoff, G. and M. Johnson 1980. Metaphors We Live By. Chicago, University of Chicago Press.
Nunberg, G. 1993. “Indexicality and deixis.” Linguistics and Philosophy 16: 1-43.
Recanati, F. 1991. “The pragmatics of what is said.” In S. Davis ed., Pragmatics: A Reader. New
York, Oxford University Press: 97-120.
Recanati, F. 1993. Direct Reference: From Language to Thought. Oxford, Blackwell.
Rumelhart, D. 1993. "Some problems with the notion of literal meanings." In A. Ortony ed.,
Metaphor and Thought. Cambridge, Cambridge University Press: 71-82.
Searle, J. R. 1993. “Metaphor.” A. Ortony ed., Metaphor and Thought. Cambridge, Cambridge
University Press: 83-111.
Smith, Q. 1989. “The multiple uses of indexicals.” Synthese 78: 167-191.
Sperber, D. and D. Wilson 1986. Relevance: Communication and Cognition. Oxford, Blackwell.
Stern, J. 2000. Metaphor in Context. Cambridge, MA, MIT Press.
Travis, C. 1985. “On what is strictly speaking true.” Canadian Journal of Philosophy 152: 187-
229.
Tsohatzidis, S. L. 1994. “Speaker meaning, sentence meaning and metaphor.” In S. L.
Tsohatzidis ed., Foundations of Speech Act Theory: Philosophical and Linguistic
Perspectives. London, Routledge: 365-373.
White, R. M. 1996. The Structure of Metaphor: The Way the Language of Metaphor Works.
Oxford, Blackwell.