-7 Metaphor 93
Metaphor
JOHN R. SEARLE
Formulating theproblem
If you hear somebody say, "Sally is a block of ice," or "Sam is a pig;
you are likely to assume that the speaker does not mean what he saps
literally but that he is speaking metaphorically. Furthermore, you are
not likely to have very much trouble figuring out what he means. If
he says, "Sally is a prime number between 17 and 23,?' or "Bill is a
barn door," you might still assume he is speaking metaphorically, but
it is much harder to figure out what he means. The existence of such
utterances -utterances in which the speaker means metaphorically
something different from what the sentence means literally - poses a
series of questions for any theory of language and communication:
What is metaphor, and how does it differ from both literal and other
forms of figurative utterances? 'M%y do we use expressions metaphorically
instead of saying exactly and literally what we mean? How do
metaphorical utterances work, that is, how is it possible for speakers to
communicate to hearers when speaking metaphorically inasmuch as
they do not say what they mean? And why do some metaphors work
and others not?
In my discussion, I propose to tackle this latter set of questionsthose
centering around the problem of how metaphors work-both
because of its intrinsic interest, and because it does not seem to me that
we shall get an answer to the others until this fundamental question
has been answered. Before we can begin to understand it, however, we
need to formulate the question more precisely.
The problem of explaining how metaelors work is a special case of
the general problem of explaining how speaker meaning and sentence
or word meaning come apart. It is a special case, that is, of the problem
of how it is possible to say one thing and mean something else,
where one succeeds in communicating what one means even though
both the speaker and the hearer know that the meanings of the words
uttered by the speaker do not exactly and literally express what the
1 speaker meant. Some other instances of the break between speaker9s
1 utterance meaning and literal sentence meaning are irony and indirct
speech acts. In each of these cases, what the speaker means is not identI
ical with what the sentence means, and yet what he means is in various
lvays dependent on what the sentence means.
~tis essential to emphasize at the very beginning that the problem of
concerns the relations between word and sentence meaning,
on the one hand, and speaker's meaning or utterance meaning, on the
I
other. Many writers on the subject try to locate the metaphorical ele-
,
ment of a metaphorical utterance in the sentence or expressions
uttered. They think there are two kinds of sentence meaning, literall
and metaphorical. However, sentences and words have only the meaning
that they have. Strictly speaking, whenever we talk about the metaphorical
meaning of a word, expression, or sentence, we are talking
I
about what a speaker might utter it to mean, in a way that departs
from what the word, expression, or sentence actually means. We are,
I
therefore, talking about possible speaker's intentions. Even when we
discuss how a nonsense sentence, such as Chomsky's example, "Color,
less green ideas sleep furiously," could be given a metaphorical interpretation,
what we are talking about is how a speaker could utter the
sentence and mean something by it metaphorically, even though it is
literally nonsensical. T o have a brief way of distinguishing what a
speaker means by uttering words, sentences, and expressions, on the one
hand, and what the words, sentences, and expressions mean, on the ,
other, I shall call the former speaker's utterance meaning, and the
I latter, word, or sentence meaning. Metaphorical meaning is always
I speaker's utterance meaning.
I
In order that the speaker can communicate using metaphorical
utterances, ironical utterances, and indirect speech acts, there must be
1 some principles according to which he is able to mean more than, or
'
something different from, what he says, whereby the hearer, using his
knowledge of them, can understand what the speaker means. The
relation between the sentence meaning and the metaphorical utterance
I meaning is systematicrather than random or ad hoc. Our task in constructing
a theory of metaphor is to try to state the principles which
relate literal sentence meaning to metaphorical utterance meaning.
Because the knowledge that enables people to use and understand metaphorical
utterances goes beyond their knowledge of the literal meanings
of words and sentences, the principles we seek are not included, or
at least not entirely included, within a theory of semantic competence
as traditionally conceived. From the point of view of the hearer, the
problem of a theory of metaphor is to explain how he can understand
the speaker's utterance meaning given that all he hears is a sente
ncewith its word and sentence meaning. From the point of view of the
speaker, the problem is to explain how he can mean something dikp
ent from the word and sentence meaning of the sentence he utters. In
light of these reflections, our original question, How do metaphors
work? can be recast as follows: What are the principles that enable
speakers to formulate, and hearers to understand, metaphorical utterances?
and How can we state these principles in a way that makes it
clear how metaphorical utterances differfrom other sorts of utterances
in which speaker meaning does not coincide with literal meaning?
Because part of our task is to explain how metaphorical utterances
differ from literal utterances, to start with we must arrive at a cIlaracter.
ization of literal utterances. Most -indeed all -of the authors I have
read on the subject of metaphor assume that we know how literal
utterances work; they do not think that the problem of literal utterances
is worth discussing in their account of metaphor. The price they
pay for this is that their accounts often describe metaphorical utterances
in ways that fail to distinguish them from literal ones.
In fact, to give an accurate account of literal predication is an
extremely difficult, complex, and subtle problem. I shall not attempt
anything like a thorougll summary of the principles of literal utterance
but shall remark on only those features which are essential for a comparison
of literal utterance with metaphorical utterance. Also, for the
sake of simplicity, I shall confine most of my discussion of both literal
and metaphorical utterance to very simple cases, and to sentences used
for the speech act of assertion.
Imagine that a speaker makes a literal utterance of a sentence such
as
( 1 ) Sally is tall
(2) The cat is on the mat
(3) It's getting hot in here.
Now notice that in each of these cases, the literal meaning of the sentence
determines, at least in part, a set of truth conditions; and bemuse
the only illocutionary force indicating devices (see Searle, 1969) in the
sentences are assertive, the literal and serious utterance of one Of
sentences will commit the speaker to the existence of the Jet of
conditions determined by the meaningof that sentence, togetllm withthe
other determinants of truth conditions. Notice, furthermore, that in eac!i
case the sentence only determines a definite set of truth conditions
tive to a particular context That is because each of these examples h3s
some indexical element, such as the present tense, or the demonsuative
-,ere;. or the occurrenceof contextually dependent definite descriptions,
as "be cat'' and "the mat"
In these the contextually dependent elements of the sen.?
realized in the semantic structure of the sentence:tence are expliatl,
one an see and hear the indexical expressions. But these sentences,
like most sentences, only determine a set of truth conditions against a
backgound of assumptions that are not explicitly realized in the
Semanticstnlcture of the sentence. This is most obvious for ( 1 ) and
(g), because they contain the relative terms "tall" and "hot." These
are
grammarians called "attributive" terms, and
hey only determine a definite set of truth conditions against a backgrnund
of factual assumptions about the sort of things referred to by
hespeaker in the rest of the sentence. Moreover, these assumptions are
not explicitlyrealized in the semantic structure of the sentence. Thus,
a woman can be correctly described as "tall" even though she is shorter
*$ana giraffe that could correctly be described as "short."
Though this dependence of the application of the literal meaning of
re sentence on certain factual background assumptions that are not
art of the literal meaning is most obvious for sentences containing
attributive terms, the phenomenon is quite general. Sentence (2) only
determinesa definite set of truth conditions given certain assumptions
about cats, mats, and the relation of being on. However, these assump
dons are not part of the semantic content of the sentence. Suppose, for
example, that the cat and mat are in the usual cat-on-mat spatial configuration,only
both cat and mat are in outer space, outside any gravitational
field relative to which one could be said to be '"hove" or .
"over" the other. Is the cat still on the mat? Without some further
assumptions, the sentence does not determine a definite set of truth
conditions in this context. Or suppose all cats suddenly became lighter
than air, and the cat went flying about with the mat stuck to its belly.
1sthe cat still on the mat?
We know without hesitation what are the truth conditions of, "The
is on the ceiling," but not of, "The cat is on the ceiling." and this
diRerenceis not a matter of meaning, but a matter of how our factual
back@o.nd information enables us to apply the meanings of sentences.
Ingeneral, one can say that in most cases a sentence only determines a
Set
truth conditions relative to a set of assumptions that are not realized
in the semantic content of the sentence. Thus, even in literal
where speaker's meaning coincides with sentence meaning,
the 'paker n ~ ~ s tcontribute more to the literal utterance than just the
"mantic of the sentence, because that semantic content only
determinesa set of truth conditions relative to a set of assumptions
made by the speaker, and if communication is to be successful, his
assumptions must be shared by the hearer (for further discussion of
this point, see Searle, 1978).
Notice finally that the notion of similarity plays a crucial role in
any account of literal utterance. Tllis is because the literal meaning
any general term, by determining a set of truth conditions, also determines
a criterion of similarity between objects. TOknow that a general
term is true of a set of objects is to know that they are similar with
respect to the property specifiedby that term. All tall women are similar
with respect to being tall, all hot rooms similar with respect to
being hot, all square objects similar with respect to being square, and
SO on.
T o summarize this brief discussion of some aspects of literal utter.
ance, there are three features we shall need to keep in mind in our
account of metaphorical utterance. First, in literal utterance the
speaker means what he says; that is, literal sentence meaning and
speaker's utterance meaning are the same; second, in general the literal
meaning of a sentence only determines a set of truth conditions relative
to a set of background assumptions which are not part of the
semantic content of the sentence; and third, the notion of similarity
plays an essential role in any account of literal predication.
When we turn to cases where utterance meaning and sentencemeaning
are different, we find them quite various. Thus, for example, (3)
could be uttered not only to tell somebody that it is getting hot in the
place of utterance (literal utterance), but it could also be used to
request somebody to open a window (indirect speech act), to complain
about how cold it is (ironical utterance), or to remark on the increasing
vituperation of an argument that is in progress (metaphorial
utterance). In our account of metaphorical utterance, we shall need to
distinguish it not only from literal utterance, but also from these other
forms in which literal utterance is departed from, or exceeded, in some
way.
Because in metaphorical utterances what the speaker means cliffen
from what he sap (in one sense of "say"), in general we shall need
two sentences for our examples of metaphor- first the sentence uttered
metaphorically, and second a sentence that expresses literally what
speaker means when he uttm the first sentence and means it
phorically. Thus (3). the metaphor (m):
(3) ME^) It's getting hot in here.
corresponds to (3), the paraphrase (PAR):
(3) (PAR)The argument that is going on is becoming more
perative
and similarlywith the pairs:
7
(om)Sally is a block of ice
(q) (FAR)Sally is an extremely unemotional and unresponsive
person
(5) (Mm)1have climbed to the top of the greasy pole (Disraeli)
I have after great difficulty become prime minister
(6) ( M ~ ~ )Richard is a gorilla
(6)
~ i c h a r dis fierce, nasty, and prone to violence.
;\roticethat in each case we feel that the paraphrase is somehow inad-
e'l
that something is lost. One of our tasks will be to explain this
sense of dissatisfaction that we have with paraphrases of even feeble
metapll~rs.Still, in some sense, the paraphrases must approximate
,,,hatthe speaker meant, because in each case the speaker's metaphoricalassertion
will be true if, and only if, the corresponding assertion
tile PAR sentence is true. When we get to more elaborate examples,
our sense of the inadequacy of the paraphrase becomes more
acute. HOW would we paraphrase
(7) (MET) My Life had stood-a Loaded Gun-
In Comers-till a Day
The Owner passes -identified
And carriedMe away- (Emily Dickinson)?
Clearlya good deal is lost by
(T) (PAR)My life was one of unrealized but readily reaIizable
potential (a loaded gun) in mediocre surroundings (comers)
until such time (a day) when my destined lover (the owner)
came (passed), recognized my potential (identified), and took
(carried)me away.
Yet. even in this case, the paraphrase or something like it must express
a large part of speaker's utterance meaning, because the truth conditions
are the same.
Sometimes we feel that we know exactly what the metaphor means
and Yet would not be able to formulate a literal PAR sentence because
are no literal expressions that convey what it means. Thus even
forsuch a simple a s e as
(8) (MET) The ship ploughed the sea,
we ma). not be able to construct a simple paraphrase sentence even
!haugh there is no obscurity in the metaphorical utterance. And
Indeedmetaphors often serve to plug such semantic gaps as this. In
Othercases,there may be an indefinite range of paraphrases. For example,l'hen
Romeo says:
I
(9) ( ~ I F T )Juliet is the sun,
there be a range of things he might mean. But while lamenting
Ule Inadequacyof paraphrases, let us also recall thatparaphrase is sym'Peuical
relation T o say that the paraphrase is a poor paraphrase
of the metaphor is also to say that the metaphor is a poor paraphraseof
its paraphrase. Furthermore, we should not feel apologetic about the
fact that some of our examples are trite or dead metaphors. Dead melaphors
are especially interesting for our study, because, to speak oxvmo.
ronically, dead metaphors have lived on. They have become iead
through continual use, but their continual use is a clue that they sat.
isfy some semantic need.
Confining ourselves to the simplest subject-predicate cases, we can
say that the general form of the metaphorical utterance is that a
speaker utters a sentence of the form "S is P' and means metapllorically
that S is R. In analyzing metaphorical predication, we need to
distinguish, therefore, between three sets of elements. Firstly, there is
the subject expression "S" and the object or objects it is used to refer
to. Secondly, there is the predicate expression "P" that is uttered and
the literal meaning of that expression with its corresponding truth conditions,
plus the denotation if there is any. And thirdly, there is the
speaker's utterance meaning "S is R" and the truth conditions determined
by that meaning. In its simplest form, the problem of metaphor
is to try to get a characterization of the relations between the three
sets, S, P, and R1, together with a specification of other information
and principles used by speakers and hearers, so as to explain how it is
possible to utter "S is P" and mean "S is R," and how it is possible to
communicate that meaning from speaker to hearer. Now, obviously.
that is not all there is to understand about metaphorical utterances:
the speaker does more than just assert that S is R,and the peculiar effectiveness
of metaphor will have to be explained in terms of how he does
more than just assert that S is R and why he should choose this roundabout
way of asserting that s is R in the first place. But at this <tape we
are starting at the beginning. At the very minimum, a theory
phor must explain how it is possible to utter "S is P' and b(
and communicate that Sis R.
We can now state one of the differences between literal and rnetaphorical
utterances as applied to these simple examples: In the case
literal utterance, speaker's meaning and sentence meaning are the
same; therefore the assertion made about the object referred to
be
true if and only if it satisfies the truth conditions determined by thebackmeaning
of the general term as applied against a set of shared
ground assumptions. In order to understand the utterance, the
does not require any extra knowledge beyond his know led^^
the
rules of language, his awareness of the conditions of utterance*
and a
metaset
of shared background assumptions. But, in the case of the
phorical utterance, the truth conditions of the assertion are not detep
--
of meta)th
mean
minedby the truth conditions of the sentence and its genera1 term. In
order to understand the metaphorical utterance, the hearer requires
something more than his knowledge of the language, his awareness of
rnnditions of the utterance, and background assumptions that he
'~iththe speaker. He must have some other principles, or some
factual information, or some combination of principles and
infomation that enables him to figureout that when the speaker says,
is p," he means "S is I?." What is this extra element?
I believe that at the most general level, the question has a fairly
answer, but it will take me much of the rest of this discussion to
work it out in any detail. The basic principle on which all metaphor
worksis that the utterance of an expression with its literal meaning
and corresponding truth conditions can, in various ways that are specificto
metaphor, call to mind another meaning and corresponding set
of truth conditions. The hard probIem of the theory of metaphor is to
explain what exactly are the principles according to which the utterance
of an expression can metaphorically call to mind a different set of
truth conditions from the one determined by its literal meaning, 2nd
to state those principIes precisely and without using metaphorical
expressions like "call to mind."
Some commonmistakes about metaphor
Before attempting to sketch a theory of metaphor, I want in this section
and the next to backtrack a bit and examine some existing theories.
Roughly speaking, theories of metaphor from Aristotle to the present
can be divided into two types.2 Comparison theories assert that
metaphorical utterances involve a comparison or similarity between
two or more objects (e.g., Aristotle, 1gj2a. 1952b; Henle, 1965)~and
Semantic interaction theories claim that metaphor involves a verbal
Of-Yosition (Beardsley, 1962) or interaction (Black, 1962b) between
hOSem~nti~contents, that of the expression used metaphorically, and
&at of the surrounding literal context. 1 think that both of these theoria*if
one tries to take them quite literally, are in various ways inadequate:nonetheless,
they are both trying to say something true, and we
Oughtto t q to extract what is true in them. But first I want to show
w?e the common mistakes they contain and some further common
made in discussions of metaphor. My aim here is not polemia'rather.
I am trying to clear the ground for the development of a
'6eoq metaphor. One might say the endemic vice of the comparison
Uleories is that they fail to distinguish between the claim that the statemcnt
Of the comparison is part of the meaning, and hence the truth
conditions, of the metaphorical statement, and the claim that the state.
ment of the similarity is the principle of inference, or a step in the
process of comprelzending, on the basis of which speakers produce and
hearers understand metaphor. (More about this distinction later.) The
semantic interaction theories were developed in response to the weaknesses
of the comparison theories, and they have little independent
argument to recommend them other than the weakness of their rivals:
Their endemic vice is the failure to appreciate the distinction between
sentence or word meaning, which is never metaphorical, and speaker or
utterance meaning, which can be metaphorical. They usually try to
locate metaphorical meaning in the sentence or some set of associations
with the sentence. In any event, here are half a dozen mistakes, which I
believe should be noted:
It is often said that in metaphorical utterances there is a change in
meaning of at least one expression. I wish to say that on the contrary,
strictly speaking, in metaphor there is never a change of meaning;
diachronically speaking, metaphors do indeed initiate semantic
changes, but to the extent that there has been a genuine change in
meaning, so that a word or expression no longer means what it previously
did, to precisely that extent the locution is no longer metaphorical.
We are all familiar with the processes whereby an expression
becomes a dead metaphor, and then finally becomes an idiom or
acquires a new meaning different from the original meaning. But in a
genuine metaphorical utterance, it is only because the expressions have
not changed their meaning that there is a metaphorical utterance at
all. The people who make this claim seem to be confusing sentence
meaning with speaker's meaning. The metaphorical utterance does
indeed mean something different from the meaning of the words and
sentences, but that is not because there has been any change in the
meanings of the lexical elements, but because the speaker means something
different by them; speaker meaning does not coincide with sentence
or word meaning. It is essential to see this point, because the
main problem of metaphor is to explain how speaker meaning and sentence
meaning are different and how they are, nevertheless, related.
Such an explanation is impossible if we suppose that sentence or word
meaning has changed in the metaphorical utterance.
The simplest way to show that the crude versions of the comparison
view are false is to show that, in the production and understanding of
metaphorical utterances, there need not be any two objects for comparison.
When I say metaphorically
(4) (MET) Sally is a block of ice,
I am not necessarily quantifying over blocks of ice at all. My utterance
does not entail literally that
hem'on
assertio;
tions ar
phorica
the P tt
suppost
- (lo)
(3x) (x is a block of ice),
and such that I am comparing Sally to x. This point is even more
obvious if we take expressions used as metaphors which have a null
If I say
(11) Sally is a dragon
&at does not entail literally
(12) (ZX)(Xis a dragon).
Or, another way to see the same thing is to note that the negative sentence
is just as metaphoricalas the affirmative. If I say
(13) Sally is not a block of ice,
that, I take it, does not invite the absurd question: Which block of ice
is it that you are comparing Sally with, in order to say that she is not
like it? At its crudest, the comparison theory is just muddled about the
referential character of expressions used metaphorically.
P.~.T-*.~ this might seem a somewhat minor objection to the compari!
orists, but it paves the way for a much more radical objection.
rison theories which are explicit on the point at all, generally
rrear ~11estatement of the comparison as part of the meaning and
hence as part of the truth conditions of the metaphorical statement.
For example, Miller (this volume) is quite explicit in regarding metaphorical
statements as statements of similarity, and indeed for such
theorists the meaning of a metaphorical statement is always given by
an explicit statement of similarity. Thus, in their view, I have not even
formulated the problem correctly. According to me, the problem of
explaining (simple subject-predicate) metaphors is to explain how
the speaker and hearer go from the literal sentence meaning "S is P'to
the metaphorical utterance meaning "S is R." But, according to them,
that is not the utterance meaning; rather the utterance meaning must
be expressibleby an explicit statement of similarity, such as "S is like P
with respect to R," or in Miller's case, the metaphorical statement "S is
P'is to be analyzed as, "There is some property F and some property
G such that S's being F is similar to P's being G." I will have more to
say about this thesis and its exact formulation later, but at present I
want to claim that though similarity often plays a role in the compeof
metaphor, the metaphorical assertion is not necessarily an
n of similarity. The simplest argument that metaphorical asser-enot
always assertions of similarity is that there are true meta-
11assertions for which there are no objects to be designated by
:rm, hence the true metaphorical statement cannot possibly pre?
the existence of an object of comparison. But even where there
are objects of comparison, the metaphorical assertion is not necessarily
an assertion of similarity. Similarity, I shall argue, has to do with the
production and understanding of metaphor, not with its meaning.
A second simple argument to show that metaphorical assertions are
not necessarily assertions of similarity is that often the metaphorical
assertion can remain true even though it turns out that the statement
of similarity on which the inference to the metaphorical meaning is
based is false. Thus, supposeI say,
(6) (MET)Richard is a gorilla
meaning
(6) (PAR)Richard is fierce, nasty, prone to violence, and so forth.
And suppose the hearer's inference to (6 PAR)is based on the belief
that
(14) Gorillas are fierce, nasty, prone to violence, and so forth,
and hence (6 MET)and (14)~on the comparison view, would justify
the inference to
(15) Richard and gorillas are similar in several respects; vir, they
are fierce, nasty, prone to violence, and so forth,
and this in turn would be part of the inference pattern that enabled
the hearer to conclude that when I uttered (6 MET) I meant (6 PAR).
But suppose ethological investigation shows, as I am told it has, that
gorillas are not at all fierce and nasty, but are in fact shy, sensitive creatures,
given to bouts of sentimentality. This would definitely show that
(15) is false, for (15) is as much an assertion about gorillas as about
Richard. But would it show that when I uttered (6 MET), what I said
was false? Clearly not, for what I meant was (6 PAR),and (6 PAR)is
an assertion about Richard. It can remain true regardless of the actual
facts about gorillas; though, of course, what expressions we use to
convey metaphorically certain semantic contents will normally depend
on what we take the facts to be.
T o put it crudely, "Richard is a gorilla," is just about Richard; it is
not literally about gorillas at all. The word "gorilla" here serves to
convey a certain semantic content other than its own meaning by a set
of principles I have yet to state. But (15) is literally about both Richard
and gorillas, and it is true if and only if they both share the properties
it claims they do. Now, it may well be true that the hearer
employs something like (15) as a step in the procedures that get him
from (6 MET)to (6 PAR),but it does not follow from this fact about
his procedures of conaprehension that this is part of the speaker's utterance
meaning of (6 MET);and, indeed, that it is not part of the utterance
meaning is shown by the fact that the metaphorical statement can
be true even if it turns out that gorillas do not have the traits that the
metaphorical occurrence of "gorilla" served to convey. I am not saying
that a metaphorical assertion can never be equivalent in meaning to a
statement of similarity-whether or not it is would depend on the
intentions of the speaker; but I am saying that it is not a necessary feature
of metaphor-and is certainly not the point of having
metaplior -that metaphorical assertions are equivalent in meaning to
statements of similarity. My argument is starkly simple: In many cases
the metaphorical statement and the corresponding similarity statement
cannot be equivalent in meaning because they have different truth conditions.
The difference between the view I am attacking and the one I
shall espouse is this. According to the view I am attacking, (6 MET)
means Richard and gorillas are similar in certain respects. According
to the view I shall espouse, similarity functions as a comprehension
strategy, not as a component of meaning: (6 MET)says that Richard has
certain traits (and to figure out what they are, look for features associated
with gorillas). On my account the P term need not figure literally
in the statement of the truth conditions of the metaphorical statement
at all.
Similar remarks apply incidentally to similes. If I say,
(16) Sam acts like a gorilla
that need not commit me to the truth of
(17) Gorillas are such that their behavior resembles Sam's.
For (16) need not be about gorillas at all, and we might say that
"gorilla" in (16) has a metaphorical occurrence. Perhaps this is one
way we might distinguish between figurative similes and literal statements
of similarity. Figurative similes need not necessarily commit the
speaker to a literal statement of similarity.
The semantic interaction view, it seems to me, is equally defective.
One of the assumptions behind the view that metaphorical meaning is
a result of an interaction between an expression used metaphorically
and other expressions used literally, is that all metaphorical uses of
expressions must occur in sentences containing literal uses of expressions,
and that assumption seems to me plainly false. It is, incidentally,
the assumption behind the terminology of many of the contemporary .
discussions of metaphor. We are told, for example, that every metaphorical
sentence contains a "tenor" and a "vehicle" (Richards,
1936a) or a "frame" and a "focus" (Black, ig6zb). But it is not the
case that every metaphorical use of an expression is surrounded by literal
uses of other expressions. Consider again our example (4): In
uttering, "Sally is a block of ice," we referred to Sally using her proper
name literally, but we need not have. Suppose, to use a mixed metaphor,
we refer to Sally as "the bad news." We would then say, using a
mixed metaphor
(18) The bad news is a block of ice.
If you insist that the "is" is still literal, it is easy enough to construct
examples of a dramatic diange on Sally's part where we would be inclined,
in another mixed metaphor, to say
(19) The bad news congealed into a block of ice.
Mixed metaphors may be stylistically objectionable, but I cannot see
that they are necessarily logically incoherent. Of course, most metaphors
do occur in contexts of expressions used literally. It would be
very hard to understand them if they did not. But it is not a logical
necessity that every metaphorical use of an expression occurs surrounded
by literal occurrences of other expressions and, indeed, many
famous examples of metaphor are not. Thus Russell's example of a
completely nonsensical sentence, "Quadrilaterality drinks procrastination,"
is often given a metaphorical interpretation as a description of
any postwar four-power disarmament conference, but none of the
words, so interpreted, has a literal occurrence; that is, for every word
the speaker's utterance meaning differsfrom the literal word meaning.
However, the most serious objection to the semantic interaction view
is not that it falsely presupposes that all metaphorical occurrences of
words must be surrounded by literal occurrence of other words, but
rather, that even where the metaphorical occurrence is within the context
of literal occurrences, it is not in general the case that the metaphorical
speaker's meaning is a result of any interaction among the elements
of the sentence in any literal sense of "interaction." Consider
again our example (4). In its metaphorical utterances, there is no
question of any interaction between the meaning of the "principal subject"
("Sally") and the "subsidiary subject" ("block of ice"). "Sally" is
a proper name; it does not have a meaning in quite the way in which
"block of ice" has a meaning. Indeed, other expressions could have
been used to produce the same metaphorical predication. Thus,
(20) Miss Jones is a block of ice
or
(21) That girl over there in the corner is a block of ice
could have been uttered with the same metaphorical utterance mean-
ing.
I conclude that, as general theories, both the object comparison view
and the semantic interaction view are inadequate. If we were to diagnose
their failure in-Fregean terms, we might say that the comparison
view tries to explain metaphor as a relation between references, and
the interaction view tries to explain it as a relation between senses and
beliefs associated with references. The proponents of the interaction
view see correctly that the mental processes and the semantic processes
involved in producing and understanding metaphorical utterances
cannot involve references themselves, but must be at the level of intentionality,
that is, they must involve relations at the level of beliefs,
meanings, associations, and so on. However, they then say incorrectly
that the relations in question must be some unexplained, but metaphorically
described, relations of "interaction"3 between a literal
frame and a metaphorical focus.
Two final mistakes I wish to note are not cases of saying something
false about metaphors but of saying something true which fails to distinguish
metaphor from literal utterance. Thus it is sometimes said
that the notion of similarity plays a aucial role in the analysis of metaphor,
or that metaphorical utterances are dependent on the context for
their interpretation. But, as we saw earlier, both of these features are
true of literal utterances as well. An analysis of metaphor must show
how similarity and context play a role in metaphor different from their
role in literal utterance.
A further examination of the comparison the0ry
One way to work up to a theory of metaphor would be to examine the
strengths and weaknesses of one of the existing theories. The obvious
candidate for this role of stalking horse is a version of the comparison
theory that goes back to Aristotle and can, indeed, probably be considered
the common-sense view- the theory that says all metaphor is
really literal simile with the "like" or "as" deleted and the respect of
the similarity left unspecified. Thus, according to this view, the metaphorical
utterance, "Man is a wolf," means "hian is like a wolf in certain
unspecified ways"; the utterance, "You are my sunshine," means
"You are like sunshine to me in certain respects," and "Sally is a block
of ice," means "Sally is like a block of ice in certain but so far unspecified
ways."
The principles on which metaphors function, then, according to this
theory are the same as those for literal statements of similarity together
with the principle of ellipsis. We under~tandthe inetaphor as a shortened
version of the literal simile.4 Since literal simile requires no special
extralinguistic knowledge for its comprehension, most of the
knowledge necessary for the comprehension of metaphor is already
contained in the speaker's and hearer's semantic competence, together
with the general background knowledge of the world that makes literal
meaning comprehensible.
\Ve have already seen certain defects of this view, most notably that
metaphorical statements cannot be equivalent in meaning to literal
statements of similarity because the truth conditions of the two sorts of
statements are frequently different. Furthermore, we must emphasize
that even as a theory of metaphorical comprehension- as opposed to
a theory of metaphorical meaning-it is important for the simile
theory that the alleged underlying similes be literal statements of similarity.
If the simile statements which are supposed to explain metaphor
are themselves metaphorical or otherwise figurative, our explanation
will be circular.
Still, treated as theory of comprehension, there does seem to be a
large number of cases where for the metaphorical utterance we can
construct a simile sentence that does seem in some way to explain how
its metaphorical meaning is comprehended. And, indeed, the fact that
the specification of the values of R is left vague by the simile statement
may, in fact, be an advantage of the theory, inasmuch as metaphorical
utterances are often vague in precisely that way: it is not made exactly
clear what the R is supposed to be when we say that S is P, meaning
metaphorically that S is R. Thus, for example, in analyzing Romeo's
metaphorical statement, "Juliet is the sun," Cavell (1976, pp. 78-9)
gives as part of its explanation that Romeo means that his day begins
with Juliet. Now, apart from the special context of the play, that reading
would never occur to me. I would look for other properties of the
sun to fill in the values of R in the formula. In saying this I am not
objecting to either Shakespeare or Cavell, because the metaphor in
question, like most metaphors, is open-ended in precisely that way.
Nonetheless, the simile theory, in spite of its attractiveness, has serious
difficulties. First, the theory does more-or rather, less-than fail
to tell us how to compute the value of R exactly: So far it fails to tell
us how to compute it at all. That is, the theory still has almost no
explanatory power, because the task of a theory of metaphor is to
explain how the speaker and hearer are able to go from "S is P" to "S
is R," and it does not explain that process to tell us that they go from
"S is I"' to "S is R" by first going through the stage "S is like P with
respect to R" because we are not told how we are supposed to figure
out which values to assign to R. Similarity is a vacuous predicate: any
two things are similar in some respect or other. .Saying that the metaphorical
"S is P' implies the literal "S is like R" does not solve our
problem. It only pushes it back a step. The problem of understanding
literal similes with the respect of the similarity left unspecified is only
a part of the problem of understanding metaphor. How are we supposed
to know, for example, that the utterance, "Juliet is the sun,"
does not mean "Juliet is for the most part gaseous," or "Juliet is go
million miles from the earth," both of which properties are salient and
well-knownfeatures of the sun.
Yet another objection is this: It is crucial to the simile thesis that the
simile be taken literally; yet there seem to be a great many metaphorical
utterances where there is no relevant literal corresponding similarity
between S and P. If we insist that there are always such similes, it
looks as if we would have to interpret them metaphorically, and thus
our account would be circular. Consider our example (4) "Sally is a
block of ice." If we were to enumerate quite literally the various distinctive
qualities of blocks of ice, none of them would be true of Sally.
Even if we were to throw in the various beliefs that people have about
blocks of ice, they still would not be literally true of Sally. There
simply is no class of predicates, R,such that Sally is literally like a
block of ice with respect to R where R is what we intended to predicate
metaphorically of Sally when we said she was a block of ice. Being
unemotional is not a feature of blocks of ice because blocks of ice are
not in that line of business at all, and if one wants to insist that blocks
of ice are literally unresponsive, then we need only point out that that
feature is still insufficient to explain the metaphorical utterance meaning
of (4), because in that sense bonfires are "unresponsive" as well.
but
(2s) Sally is a bonfire
has a quite different metaphorical utterance meaning from (4). Furthemore,
there are many similes that are not intended literally. For
example, an utterance of "My love is like a red, red rose" does not
mean that there is a class of literal predicates that are true both of my
love and red, red roses and that express what the speaker was driving
at when he said his love was like a red, red rose.,.
The defender of the simile thesis, however, need not give up so
easily. He might say that many metaphors are also examples of other
figures as well. Thus, "Sally is a block of ice" is not only an example
of metaphor, but of hyperbole as well.VThe metaphorical utterance
meaning is indeed derived from the simile, "Sally is like a block of
ice," but then both the metaphor and the simile are cases of hyperbole;
they are exaggerations, and indeed, many metaphors are exaggerations.
According to this reply, if we interpret both the metaphor and the
simile hyperbolically, they are equivalent.
Furthermore, the defender of the simile thesis might add that it is
not an objection to the simile account to say that some of the respects
in which Sally is like a block of ice will be specified metaphorically,
because for each of these metaphorical similes we can specify another
underlying simile until eventually we reach the rock bottom of literal
similes on which the whole edifice rests. Thus "Sally is a block of ice"
means "Sally is like a block of ice," which means "She shares certain
traits with a block of ice, in particular she is very cold." But since
"cold" in "Sally is very cold" is also metaphorical, there must be an
underlying similarity in which Sally's emotional state is like coldness,
and when we finally specify these respects, the metaphor will be completely
analyzed.
There are really two stages to this reply: First, it points out that
other figures such as hyperbole sometimes combine with metaphor,
and, secondly, it concedes that some of the similes that we can offer as
translations of the metaphor are still metaphorical, but insists that
some recursive procedure of analyzing metaphorical similes will eventually
lead us to literal similes.
Is this reply really adequate? I think not. The trouble is that there
do not seem to be any literal similarities between objects which are
cold and people who are unemotional that would justify the view that
when we say metaphorically that someone is cold what we mean is that
he or she is unemotional. In what respects exactly are unemotional
people like cold objects? Well, there are some things that one can say
in answer to this, but they all leave us feeling somewhat dissatisfied.
We can say, for example, that when someone is physically cold it
places severe restrictions on their emotions. Rut even if that is true, it
is not what we meant by the metaphorical utterance. I think the only
answer to the question, "What is the relation between cold things and
unemotional people?" that would justify the use of "cold" as a metaphor
for lack of emotion is simply that as a matter of perceptions, sensibilities,
and linguistic practices, people find the notion of coldness
associated in their minds with lack of emotion. The notion of being
cold just is associated with being unemotional.
There is some evidence, incidentally, that this metaphor works across
several different cultures: It is not confined to English speakers (cf.
Asch, 1958). Aforeover, it is even becoming, or has become, a dead metaphor.
Some dictionaries (for example, the O.E.D.) list lack of emotion
as one of the meanings of "cold." Temperature metaphors for
emotional and personal traits are in fact quite common and they are
not derived from any literal underlying similarities. Thus we speak of
a "heated argument," "a warm welcome," "a lukewarm friendship,"
and "sexual frigidity." Such metaphors are fatal for the simile thesis,
unless the defenders can produce a literal R which S and P have in
common, and which is sufficient to explain the precise metaphorical
meaning which is conveyed.
Because this point is bound to be contested, it is well to emphasize
exactly what is at stake. In claiming that there are no sufficient similarities
to explain utterance meaning, I am making a negative existential
claim, and thus not one which is demonstrable from an examination of
a finite number of instances. The onus is rather on the similarity theorist
to state the similarities and show how they exhaust utterance meaning.
But it is not at all easy to see how he could do that in a way that
would satisfy the constraints of his own theory.
Of course, one can think of lots of ways in which any S is like any P,
e.g., ways in which Sally is like a block of ice, and one can think of lots
of F s and G's such that Sally's being F is like a block of ice's being G.
But that is not enough. Such similarities as one can name do not
exhaust utterance meaning and if there are others that do, they are certainly
not obvious.
But suppose with some ingenuity one could think up a similarity
that would exhaust utterance meaning. The very fact that it takes so
much ingenuity to think it up makes it unlikely that it is the underlying
principle of the metaphorical interpretation, inasmuch as the metaphor
is obvious: There is no difficulty for any native speaker to explain
what it means. In "Sam is a pig," both utterance meaning and similarities
are obvious, but in "Sally is a block of ice," only the utterance
meaning is obvious. The simpler I~ypotl~esis,then, is that this metaphor,
'
like several others I shall now discuss, functions on principles other
than similarity.
Once we start looking for them, this class of metaphors turns out to
be quite large. For example, the numerous spatial metaphors for temporal
duration are not based on literal similarities. In "time flies," or
"the hours crawled by," what is it that time does and the hours did
which is literally like flying or crawling? We are tempted to say they
went rapidly or slowly respectively, but of course "went rapidly" and
"went slowly" are further spatial metaphors. Similarly, taste metaphors
I for personal traits are not based on properties in common. IVe speak ofl
a "sweet disposition" or a "bitter person," without implying that the
sweet disposition and the bitter person have literal traits in common
with sweet and bitter tastes which exhaust the utterance meaning of
I
the metaphorical utterance. Of course, sweet dispositions and sweet
things are both pleasant, but much more is conveyed by the metaphor
1 than mere pleasantness.
I So deeply embedded in our whole mode of sensibility are certain
metaphorical associations that we tend to think there must be a similarity,
or even chat the association itself is a form of similarity. Thus,
1 we feel inclined to say that the passage of time just is like spatial movement,
but when we say this we forget that "passage" is only yet another
spatial metaphor for time and that the bald assertion of similarity,
1 with no specification of the respect of similarity, is without content.
I The most sophisticated version of the simile thesis I have seen is by
George Miller (this volume), and I shall digress briefly to consider
some of its special features. P.liller, like other simile theorists, believes
that the meanings of metaphorical statements can be expressed as statements
of similarity, but he offers a special kind of similarity statement
(rather like one of Aristotle's formulations, by the way) as the form of
"reconstruction" of metaphorical statements. According to Miller, metaphors
of the form "S is P",where both S and P are noun phrases, are
equivalent to sentencesof the form
(23) (3F) (3G) {sIMCF (S),G (P)l).
Thus, for example, "hlan is a wolf," according to Miller would be analyzed
as
(24) There is some property F and some property G such that
man's being F is similar to a wolf's being G.
And when we have metaphors where a verb or predicate adjective I; is
used metaphorically in a sentence of the form "x is F' or "xFs", the
analysisis of the form
(25) P G ) (ZY){sIM [G (x)J @)I 1.
Thus, for example, "The problem is thorny" would be analyzed as
(26) There is some property G and some object y such that the
problem's being G is similar to y's being thorny.
I believe this account has all the difficulties of the other simile
theories-namely, it mistakenly supposes that the use of a metaphorical
predicate commits the speaker to the existence of objects of which
that predicate is literally true; it confuses the truth conditions of the
metaphorical statement with the principles under which it is comprehended;
it fails to tell us how to compute the values of the variables
(Miller is aware of this problem, he calls it the problem of "interpretation"
and sees it as different from the problem of "reconstruction");
to compute no less than three sets of predicates. Inasmuch as similarity
is a vacuous predicate, we need to be told in which respect two things
are similar for tile statement that they are similar to have any informative
content. His formalization of the above metaphorical utterance is
(29) (XI'>(3G) {SIM [F (man), G (wolf)]).
In order to complete this formula in a way that would specify the
respect of the similarity we would have to rewrite it as
(go) (31;)(XG)(3H) {SIMH[F (man), G (wolf)]).
But both the reformulation (30), and Miller's original kg),contain
:oo many predicate variables: When I say, "Man is a wolf," I am not
iaying that there are some differentsets of properties that men have
irom those that wolves have, I am saying they have the same set of
properties (at least on a sympathetic construal of the simile thesis, that
is what I am saying). But according to Miller's account, I am saying
that man has one set of properties I;: wolves have a different set of
properties G, and man's having F is similar to wolves having G with
respect to some other properties H. I argue that this "reconstruction"
is (a) counterintuitive, (b) unmotivated, and (c) assigns an impossible
computing task to the speaker and hearer. What are these Fs, G's
and H's supposed to be? and, How is the hearer supposed to figure
them out? It is not surprising that his treatment of the interpretation
problem is very sketchy. Similar objections apply to his accounts of
other syntactical forms of metaphorical utterances.
There is a dass of metaphors, that I shall call "relational metaphors,"
for which something like his analysis might be more appropriate.
Thus, if I say
(8) The ship ploughed the sea
--01and
it is refuted by the fact that not all metaphors have literal statements
of similarity underlying them. But it has some additional prob- (31) Washington is the father of his country,
lems of its own. In my view, the most serious weakness of Miller's
these might be interpreted using something like his forms. We might
account is that according to it the semantic contents of most metaphop treat (8) as equivalent to
ical utterances have too many predicates, and, in fact, rather few meta- (32) There is some relation R which the ship has to the sea and
phors really satisfy the formal structure he provides us with. Consider,
which is similar to the relation that ploughs have to fields
for example, "Man is a wolf." On what I believe is the most plausible when they plough fields;
version of the simile thesis, it means something of the form (31)as
(27) Man is like a wolf in certain respects R. (33) There is some relation R which WTashingtonhas to his country
We could represent this as and which is like the relation that fathers have to their
(28) SIMR (man, wolf). offspring.
The hearer is required to compute only one set of predicates, the
And (32) and (33) are fairly easily formalized 4 la B~filler.However,
values for R. But according to hliller's account, the hearer is required even these analyses seem to me to concede too much to his approach:
(8) makes no reference either implicitly or explicitly to fields and
(y 1) makes no reference to offspring. On the simplest and most plausible
version of the simile thesis (8) and (31)are equivalent to:
(34) The ship does something to the sea which is like ploughing
and
(35) Washington stands in a relation to his country which is like
the relation of being a father.
And the hearer's task is simply to compute the intended relations in
the two cases. By my account, which I shall develop in the next section,
similarity does not, in general, function as part of the truth conditions
either in hliller's manner or in the simpler version; rather, when it
functions, it functions as a strategy for interpretation. Thus, very
crudely, the way that similarity figures in the interpretation of (8) and
(31) is given by
(36) The ship does something to the sea (to figure out what it is,
find a relationship like ploughing)
and
(37) Washington stands in a certain relationship to his country (to
figure out what it is, find a relationship like that of being a
father).
But the hearer does not have to compute any respects in which these
relations are similar, inasmuch as that is not what is being asserted.
Rather, what is being asserted is that the ship is doing something to
the sea and that IVashington stands in a certain set of relations to his
country, and the hearer is to figure out what it is that the ship does
and what the relations are that Washington stands in by looking for
relations similar to ploughing and being a father of.
T o conclude this section: The problem of metaphor is either very
difficult or very easy. If the simile theory were true, it would be very
easy, because there would be no separate semantic category of
metaphors-only a category of elliptical utterances where "like" or
"as" had been deleted from the uttered sentence. But alas, the simile
theory is not right, and the problem of metaphor remains very difficult.
I hope our rather lengthy discussion of the simile theory has been illuminating
in at least these respects. First, there are many metaphors in
which there is no underlying literal similarity adequate to explain the
metaphorical utterance meaning. Second, even where there is a correlated
literal statement of similarity, the truth conditions, and hence the
meaning of the metaphorical statement and the similarity statement,
are not, in general, the same. Third, what we should salvage from the
simile theory is a set of strategies for producing and understanding
metaphorical utterances, using similarity. And fourth, even so construed,
that is, construed as a theory of interpretation rather than of
meaning, the simile theory does not tell us how to compute the respects
of similarity or which similarities are metaphorically intended by the
speaker.
The principles of metaphorical interpretation
The time has now come to try to state the principles according to
which metaphors are produced and understood. T o reiterate, in its
I simplest form, the question we are trying to answer is, How is it possi-
1 ble for the speaker to say metaphorically "S is P" and mean "S is R,"
, when P plainly does not mean R; furthermore, How is it possible for
I the hearer who hears the utterance "S is P" to know that the speaker
I means "S is R"? The short and uninformative answer is that the utterl
ance of P calls to mind the meaning and, hence, truth conditions asso-
I
I
ciated with R, in the special ways that metaphorical utterances have of
I calling other things to mind. But that answer remains uninformative
until we know what are the principles according to which the utterance
calls the metaphorical meaning to mind, and until we can state
these principles in a way which does not rely on metaphorical expressions
like "calls to mind." I believe that there is no single principle on
which metaphor works. -'
The question. "How do metaphors work?" is a bit like the question.
"How does one thing remind us of another thing?" There is no single
answer to either question, though similarity obviously plays a major
role in answering both. Two important differences between them are
that metaphors are both restricted and systematic; restricted in the
sense that not every way that one thing can remind us of something
else will provide a basis for metaphor, and systematic in the sense that
metaphors must be communicable from speaker to hearer in virtue of a
shared system of principles.
Let us approach the problem from the hearer's point of view. If we
can figure out the principles according to whicK hearers understand
metaphorical utterances, we shall be a long way toward understanding
how it is possible for speakers to make metaphorical utterances,
because for communication to be possible, speaker and hearer must
share a common set of principles. Suppose a hearer hears an utterance
such as, "Sally is a block of ice," or "Richard is a gorilla," or "Bill is a
barn door." What are the steps he must go through in order to comprehend
the metaphorical meaning of such utterances? Obviously an
answer to that question need not specify a set of steps that he goes
througll consciously; instead it must provide a rational reconstruction
of the inference patterns that underlie our ability to understand such
metaphors. Furtliermore, not all metaphors will be as simple as the
cases we shall be discussing; nonetheless, a model designed to account
for the simple casessliould prove to be of more general application.
I believe that for the simple sorts of cases we have been discussing,
the hearer must go through at least three sets of steps. First, he must
have some strategy for determining whether or not he has to seek a
metaphorical interpretation of the utterance in the first place. Secondly,
when Ile Ilas decided to look for a metaphorical interpretation,
he must have some set of strategies, or principles, for computing possible
values of R, and third, he must have a set of strategies, or prindples,
for restricting the range of R's -for deciding which R's are likely
to be die ones the speakeris assertingof S.
Suppose he hears the utterance, "Sam is a pig." He knows that that
cannot be literally true, that the utterance, if Ile tries to take it literally,
is radically defective. And, indeed, such defectiveness is a feature
of nearly all of tlle examples that .we have considered so far. Tlie
defects which cue the hearer may be obvious falsellood, semantic nonsense,
violations of the rules of speech acts, oersational
principles of communication. This suggests a strategy that
underlies the first step: Where the utterance is defective if taken literally,
look for an utterance meaning that diflers from sentence meaning.
This is not the only strategy on which a hearer can tell that an utterance
probably has a metaphorical meaning, but it is by far the most
common. (It is also common to the interpretation of poetry. If I hear a
figure on a Grecian Urn being addressed as a "still unravish'd bride of
quietness," I know I had better look for alternative meanings.) But it
is certainly not a necessary condition of a metaphorical utterance that
it be in any way defective if construed literally. Disraeli might have
said metaphorically
(5) (MET)I have climbed to the top of the greasy pole,
though he had in fact dimbed to the top of a greasy pole. There are
various other cues that we employ to spot metaphorical utterances. For
example, when reading Romantic poets, we are on the lookout for metaphors,
and some people we know are simply more prone to metaphorical
utterances than others.
Once our hearer has established that he is to look for an alternative
meaning, he has a number of principles by which he can compute passible
values of R. I will give a list of these shortly, but one of them is
this: When you hear "S is P," to find Possible values of R look for ways
in which S might be like P, and to fill in the respect in which S might
be like P, look for salient, well known, and distinctive feature of 1'
things.
In this case, the hearer might invoke his factual knowledge to come
up with such features as that pigs are fat, gluttonous, slovenly, filthy,
and so on. This indefinite range of features provides possible values of
R. However, lots of other features of pigs are equally distinctive and
well known, for example, pips have a distinctive shape and distinctive
bristles. So, in order to understand the utterance, the hearer needs to
go through the third step where he restricts the range of possible R's.
Here again the hearer may employ various strategies for doing that but
the one that is most commonly used is this: Go back to the S term and
see which of the many candidates for the values of R are likely or even
possible properties of S.
Thus, if the hearer is told, "Sam's car is a pig," he will interpret that
metaphor differently from the utterance, "Sam is a pig." The former,
he might take to mean that Sam's car consumes gas the way pigs consume
food, or that Sam's car is shaped like a pig. Though, in one sense,
the metaphor is the same in the two cases, in each case it is restricted
by the S term in a differentway. The hearer has to use his knowledge
of S things and P things to know which of the possible values of R are
plausible candidates for metaphorical predication.
Now, much of tggghe dispute between the interaction theories and the
object comparison theories derives from the fact that they can be construed
as answers to different questions. The object comparison theories
are best construed as attempts to answer the question of stage two:
"How do we compute the possible values of R?" The interaction theories
are best construed as answers to the question of stage three:
"Given a range of possible values of R, how does the relationship
between the S term and the P term restrict that range?" I think it is
misleading to describe these relations as "interactions," but it seems
correct to suppose that the S term must play a role in metaphors of the
sort we have been considering. In order to show that the interaction
theory was also an answer to the question of stage two, we would have
to show that there are values of R that are specifiable, given S and P
together, that are not specifiable given P alone; one would have to
show that S does not restrict the range of R's but in fact, creates new
R's. I do not believe that can be shown, but I shall mention some pos-
sibilitieslater.
I said that there was a variety of principles for computing R, given
P- that is, a variety of principles according to which the utterance of
P can call to mind the meaning R in ways that are peculiar to metaphor.
I am sure I do not know all of the principles that do this, but
here are several (not necessarily independent) for a start.
Principle I
Things which are P are by definition R. Usually. if the metaphor
works, R will be one of the salient defining characteristics of P. Thus,
for example,
(38) (MET)Sam is a giant
will be taken to mean
(38) (PAR)Sam is big,
because giants are by definition big. That is what is special about
them.
Principle 2
Things which are P are contingently R. Again, if the metaphor works,
the property R sllould be a salient or well known property of P things.
(39) (MET)Sam is a pig
will be taken to mean
($9) (PAR)Sam is filthy,gluttonous, sloppy, and so on.
Both principles 1 and 2 correlate metaphorical utterances with literal
similes, "Sam is like a giant," "Sam is like a pig," etc. Notice in connection
with this principle and the next that small variations in the P
term can create big differences in the R terms. Consider the differences
between "Sam is a pig," "Sam is a hog," and "Sam is a swine."
Principle 3'
Things which are P are often said or believed to be R, even though
both speaker and hearer may know that R is false of P. Thus,
(7) (MET)Richard is a gorilla
can be uttered to mean
(7) (PAR)Richard is mean, nasty, prone to violence, and so on,
even though both speaker and hearer know that in fact gorillas are shy,
timid, ancl sensitive creatures, but generations of gorilla mythology
have set up associations that will enable the metaphor to work even
though both speaker and hearer know these beliefs to be false.
Principle q
Things which are P are not R, nor are they like R things, nor are they
believed to be R, nonetheless it is a fact about our sensibility, whether
7
or naturally determined, that we just do perceive a connection,
so that utterance of P is associated in our minds with R proper-
ties.
Thus,
(4) (MET)Sally is a block of ice
(40) (MET)I am in a bIack mood
(41) (MET)Mary is sweet
(42) (MET)John is bitter
(43) (MET)The hours by as we waited for the plane
whizzed
are sentences that could be uttered to mean metaphorically that: Sally
is unemotional; I am angry and depressed; Mary is gentle, kind, pleasant,
and so on; John is resentful; and the hours seemed (of varying
degrees of duration) as we waited for the plane; even though there are
no literal similarities on which these metaphors are based. Notice that
the associations tend to be scalar: degrees of temperature with ranges
of emotion, degrees of speed with temporal duration, and so forth.
Principle 5
P things are not like R things, and are not believed to be like R things,
nonetheless the condition of being P is like the condition of being R.
Thus, I might say to someone who has just received a huge promotion
(44) You have become an aristocrat,
meaning not that he has personally become like an aristocrat, but that
his new status or condition is like that of being an aristocrat.
Principle 6
There are cases where P and R are the same or similar in meaning, but
where one, usually P, is restricted in its application, and does not literally
apply to S. Thus, "addled" is only said literally of eggs, but we can
metaphorically say
(45) This souffle is addled
(46) That parliament was addled
and
(47) His brain is addled.
Principle 7
This is not a separate principle but a way oE applying principles 1
through 6 to simple cases which are not of the form "S is P" but relational
metaphors, ant1 metaphors of other syntactical forms such as
those involving verbs and predicate adjectives. Consider such relational
metaphors as
(48) Sam devours books
(8) The ship ploughs the sea
(31) Washington was the father of his country.
In each case, we have a literal utterance of two noun phrases surrouncling
a metaphorical utterance of a relational term (it can be a transitive
verb, as in (48) and (8) but it need not be, as in (31)). The
hearer's task is not to go from "S is P to "S is R" but to go from "S Prelation
S' " to "S R-relation S' " and the latter task is formally rather
different from the former because, for example, our similarity principles
in the former case will enable him to find a property that S and P
things have in common, namely, R. But in the latter, he cannot find a.
relation in common, instead he has to find a relation R which is different
from relation P but similar to it in some respect. So, as applied to
these cases, principle 1 , for example, would read
P-relations are by definition R-relations.
For example, ploughing is by definition partly a matter of moving a
substance to either side of a pointed object while the object moves forward;
and though this definitional similarity between the P-relation
and the R-relation would provide the principle that enables the hearer
to infer the R-relation, the respect of similarity does not exhaust the
context of the R-relation, as the similarity exhausts the content of the
R term in the simplest of the "S is P" cases. In these cases, the hearer's
job is to find a relation (or property) that is similar to, or otherwise
associated with, the relation or property literally expressed by the
metaphorical expression P; and the principles function to enable him
to select that relation or property by giving him a respect in which the
P-relation and the R-relation might be similar or otherwise associated.
Principle 8
According to my account of metaphor, it becomes a matter of terminology
whether we want to construe metonymy and synecdoche as special
cases of metaphor or as independent tropes. When one says, "S is P,"
and means that "S is R," P and R may be asscciated by such relations
as the part -whole relation, the container-contained relation, or even
the clothing and wearer reIation. In each case, as in metaphor proper,
the semantic content of the P term conveys the semantic content of the
R term by some principle of association. Since the principles of metaphor
are rather various anyway, I am inclined to treat metonymy and
synecdoche as special cases of metaphor and add their principles to my
list of metaphorical principles. I can, for example, refer to the British
monarch as "the Crown," and the executive branch of the U.S. government
as "the White House" by exploiting systematic principles of association.
However, as I said, the claim that these are special cases of
metaphor seems to me purely a matter of terminology, and if purists
insist that the principles of metaphor be kept separate from those of
metonymy and synecdoche, I can have no nontaxonomical objections.
In addition to these eight principles, one might wonder if there is a
ninth. Are there cases where an association between P and R that did
not previously exist can be created by the juxtaposition of S and P in
the original sentence? This, I take it, is the thesis of the interaction
theorists. However, I have never seen any convincing examples, nor
any even halfway clear account, of what "interaction" is supposed to
mean. Let us try to construct some examples. Consider the differences
between
(sandpaper 1
and
(50) Kant's second argument for the transcendental deduction is so
f I
much {;:el 1sandpaper
C
The second set clearly gi<es us different metaphorical meaningsdifferentvalues
for R -than the first trio, and one might argue that
this is due not to the fact that the different S terms restrict the range of
possible R's generated by the P terms, but to the fact that the different
combinations of S and P create new R's. But that explanation seems
implausible. The more plausible explanation is this. One has a set of
associations with the P terms, "mud," "gravel," and "sandpaper." The
principles of these associations are those of principles 1 through 7. The
differentS terms restrict the values of R differently, because different
R's can be true of voices than can be true of arguments for transcendental
deductions. Where is the interaction?
Because this section contains my account of metapllorical predication,
it may be well to s~~mmarizeits main points. Given that a speaker
and a liearer have sllared linguistic and factual knowledge sufficient to
enable them to communicate literal utterance, the following principles
are individually necessary and collectively sufficient to enable speaker
and hearer to foinl and comprehend utterances of the form "S is P,"
wliere the speaker means metaphorically that S is R (where P +R).
First, there must be some shared strategies on the basis of which the
hearer can recognize that the utterance is not intended literally. The
most common, but not the only strategy, is based on the fact that the
utterance is obviorlsly defective if taken literally.
Second, tliere must be some shared principles that associate the P
term (whether the meaning, the truth conclitions, or the denotation if
tliere is any) with a set of possible values of R. The heart of tlie problem
of metaphor is to state these principles. I have tried to state several
of them, but I feel confident that there must be more.
Third, there must be some shared strategies that enable the speaker
and tlie hearer, given their knowledge of the S term (whether tlle
meaning of the expression, or the nature of the referent, or both), to
restrict the range of possible values of R to the actual value of R. The
basic principle of this step is that only those possible values of R which
determine possible properties of S can be actual values of R.
Metaphor, irony, and indirect speech acts
T o conclude, I wish to compare briefly the principles on which metaphor
works with those on which irony and indirect speech acts work.
Consider first a case of irony. Suppose you have just broken a priceless
K'ang Hsi vase and I say ironically, "That was a brilliant thing to do."
Here, as in metaphor, the speaker's meaning and sentence meaning are
different. What are the principles by which the hearer is able to infer
that the speaker meant, "That was a stupid thing to do," when what
lie heard was the sentence, "That was a brilliant thing to do"? Stated
very crudely, the mechanism by which irony works is that the utterance,
if taken literally, is obviously inappropriate to the situation.
Since it is grossly inappropriate, the hearer is compelled to reinterpret
it in such a way as to render it appropriate, and the most natural way
to interpret it is as meaning the opposite of its literal form.
I am not suggesting that this is by any means the whole story about
irony. Cultures and subcultures vary enormously in the extent and
degree of the linguistic and extralinguistic cues provided for ironical
utterances. In English, in fact, there are certain characteristic intonational
contours that go with ironical utterances. However, it is important
to see that irony, like metaphor, does not require any conventions,
extralinguistic or otherwise. The principles of conversation and the
general rules for performing speech acts are sufficient to provide the
basic principles of irony.
Now consider a case of an indirect speech act. Suppose that in the
usual dinner-table situation, I say to you, "Can you pass the salt?" In
this situation you will normally take that as meaning, "Please pass the
salt." That is, you will take the question about your ability as a
request to perform an action. What are the principles on which this
inference works? There is a radical difference between indirect speech
acts, on the one hand, and irony and metaphor, on the other. In the
indirect speech act, the speaker means what he says. However, in addition,
he means something more. Sentence meaning is part of utterance
meaning, but it does not exhaust utterance meaning. In a very simplified
form (for a more detailed account, see Searle, 1975), the principles
on which the inference works in this case are: First, the hearer must
have some device for recognizing that the utterance might be an indirect
speech act. This requirement is satisfied by the fact that in the
context, a question about the hearer's ability lacks any conversational
point. The hearer, therefore, is led to seek an alternative meaning.
Second, since the hearer knows the rules of speech acts, he knows that
the ability to pass the salt is a preparatory condition on the speech act
of requesting him to do so. Therefore, he is able to infer that the question
about his ability is likely to be a polite request to perform the act.
The differences and similarities between literal utterances, metaphorical
utterances, ironical utterances, and indirect speech acts are illustrated
in Figure 4.
The question of whether all metaphorical utterances can be given a
literal paraphrase is one that must have a trivial answer. Interpreted
one way, the answer is trivially yes; interpreted another way, it is trivially
no. If we interpret the question as, "Is it possible to find or to
invent an expression that will exactly express the intended metaphorical
meaning R, in the sense of the truth conditions of R, for any metaphorical
utterance of 'S is P,' where what is meant is that S is R?" the
answer to that question must surely be yes. It follows trivially from the
Principle of Expressibility (see Searle, 1969) that any meaning whatever
can be given an exact expression in the language.
If the question is interpreted as meaning, "Does every existing language
provide us exact devices for expressing literally whatever we
wish to express in any given metaphor?" then the answer is obviously
no. It is often the case that we use metaphor precisely because there is
Figure 4. A graphical comparison of the relations between sentence meaning
and utterance meaning where the sentence meaning is s is P and the utterance
meaning is S is R, that is, where the speaker utters a sentence that means
literally that the object S falls under the concept P, but where the speaker
means by his utterance that the object S falls under the concept R.
a. Literal Utterance. A speaker says S is P and he means S is P. Thus the
speaker places object S under the concept P, where P = R. Sentence
meaning and utterance meaning coincide.
b. Metaphorical Utterance (simple). A speaker says S is P but means metaphorically
that S is R. Utterance meaning is arrived at by going through
literal sentence meaning.
C. Metaphorical Utterance (open ended). A speaker says S is 3, but means
metaphorically an indefinite range of meanings, S is R,,S is R2,etc. As
in the simple case, utterance meaning is arrived at by going through
literal meaning.
d. Ironical Utterance. A speaker means the opposite of what he says. Utterance
meaning is arrived at by going through sentence meaning and then
doubling back to the opposite of sentence meaning.
e. Dead Metaphor. The original sentence meaning is bypassed and the
sentence acquires a new literal meaning identical with the former metaphorical
utterance meaning. This is a shift from the metaphorical utterance
(simple), b above, to the literal utterance, diagram a.
f. Indirect Speech Act. A speaker means what he says, but he means something
more as well. Thus utterance meaning includes sentence meaning
but extends beyond it.
LI I HAL uI I tnHIlLc
P, R
IRONICAL UTTERANCE
csp
.....
.:;;;ii ..lp
no literal expression that expresses exactly what we mean. Furthermore,
in metaphorical utterances, we do more than just state that S is
R; as figure 4 shows, we state that S is R by way of going through the
meaning of "S is P." It is in this sense that we feel that metaphors
somehow are intrinsically not paraphrasable. They are not paraphrasable,
because without using the metaphorical expression, we will not
reproduce the semantic content which occurred in the hearer's comprehension
of the utterance.
The best we can do in the paraphrase is reproduce the truth conditions
of the metaphorical utterance, but the metaphorical utterance
does more than just convey its truth conditions. It conveys its truth
conditions by way of another semantic content, whose truth conditions
are not part of the truth conditions of the utterance. The expressive
power that we feel is part of good metaphors is largely a matter of two
features. The hearer has to figure out what the speaker means-he has
I to contribute more to the communication than just passive uptakeand
he has to do that by going through another and related semantic
content from the one which is communicated. And that, I take it, is
what Dr. Johnson meant when he said metaphor gives us two ideas for
one.
1 a m
this a]
Paul l
I It is (
fsSENTENCE (-J UTTERANCE
MEANING MEANING
uI I CMHI\ILC
(SIMPLE)
{.g::.... R
DEAD METAPHOR
4Re$l;
thes
mea
timt
n T F-1
U I I ~ K A N G ~
(OPEN ENDED)
RI R2 R 3 R4
y...... (.. .:.:..:.. ..........
INDIRECT SPEECH ACT
R
P
NOTES
indebted to several people for helpful comments on earlier drafts of
rticle, and I especially want to thank Jerry Morgan, Andrew Ortony,
iauber, and Dagmar Searle.
:ssential to avoid any use -mention confusions when talking about
e sets. Sometimes we will be talking about the words, other times about
nings, other times about references and denotations, and still other
:s about truth conditions.
,a LullowBeardsley (1962) in this classification.
3 Even in Black's clarification (this volume) of interaction in terms of
"implication-complexes" there still does not seem to be any precise statement
of the principles on which interaction works. And the actual example
he gives, "Marriage is a zero-sum game," looks distressingly like a comparison
metaphor: "Marriage is like a zero-sum game in that it is an adversary
relationship between two parties in which one side can benefit only at the
expense of the other." It is hard to see what the talk about interaction is
supposed to add to this analysis.
4 By "literal simile," I mean literal statement of similarity. It is arguable that
one should confine "simile" to nonliteral comparisons, but that is not the
usage I follow here.
5 Furthermore, it is at least arguable that "block of ice" functions metonymously
in this example.