METAPHOR - ENGLISH

Metaphor

The Central Trope

The metaphor is probably the most fertile power possessed by man.
Jose Ortega y Gasset, 1948

The word metaphor derives from the Greek metapherein, transfer, as

META

pherein, to bear (Oxford English Dictionary, 1996). From this



deceptively simple root, metaphor has come to mean different things to
different people, so much so that specialists in the area are often tempo-
rarily confounded when asked for a de?nition of metaphor. I vividly recall
(with some amusement) an incident in which an esteemed colleague and
occasional theoretical adversary, Ray Gibbs,1 was being interviewed by Is-
raeli security personnel during check-in for a ?ight to Tel Aviv. When
asked why he was ?ying to Tel Aviv, Gibbs replied that he had been invited
to a conference on metaphor. The interviewer asked, “What’s a metaphor?”
When Gibbs hesitated, momentarily at a loss for words, the interviewer
asked sharply, “You’re going to a metaphor conference and you don’t even
know what a metaphor is?” A consternated Professor Gibbs was thereupon
hustled away by security guards and interrogated for almost an hour before
one of the conference hosts, an Israeli professor from the University of
Tel Aviv, intervened and vouched for Ray’s legitimacy, if not for his quick-
ness of tongue.
Metaphor challenges de?nition for at least two reasons.2 First, the term
is used in several different, albeit related, senses. Second, both within and
between its different senses, de?nitions vary to re?ect sharply different
theoretical agendas and assumptions. Sometimes the theoretical boundaries
coincide with scholarly disciplines; thus, philosophers; linguists, and psy-
chologists might each de?ne metaphor in their own terms. But there are


3



4




UNDERSTANDING FIGURATIVE LANGUAGE


differences even within disciplines that re?ect different views of metaphor,
as well as different views of the nature of language itself.
Dictionary entries for the term metaphor provide illustrative examples
of how metaphor can be variously de?ned. The two major senses of the
term are captured in the Oxford English Dictionary (1996). The ?rst sense
identi?es metaphor as a type of language: “A ?gure of speech in which a
name or descriptive word or phrase is transferred to an object or action
different from, but analogous to, that to which it is literally applicable; an
instance of this [is] a metaphorical expression.” The second sense identi?es
metaphor as a form of conceptual representation: “A thing considered as
representative of some other (usually abstract) thing: A symbol.”
A particular instance of metaphor use can illustrate both of these senses
simultaneously, as when crime is referred to in terms of disease: “Crime
in our city has become an epidemic that will soon infect even our ?nest
neighborhoods.” In this instance, one thing, crime, is considered as a rep-
resentative instance of some other thing, disease. The concept disease is
thus used as a metaphor for the concept crime. If we can conceptualize
crime as an instance of disease, then crime can have (at least some of) the
properties of diseases: it can be infectious, it can be endemic, it can be an
epidemic, we might try to “cure” it, there might even be a crime virus.
And, if crime can have properties of diseases, then we can use the vocab-
ulary of disease to talk about crime, as in He’s been cured of his thieving
ways, or Corruption and bribery are a cancer in the body politic. Lehrer (1978)
provides a detailed analysis of how a conceptual relationship is ?rst ex-
pressed in terms of a root metaphor—for example, Personalities are like
textures—and then expanded via novel instantiations of this metaphor, as
in George is rough around the edges. Theories of metaphor in philosophy,
linguistics, and psychology, as well as in anthropology and sociology, ad-
dress one or more aspects of these two senses of metaphor: metaphor as a
form of linguistic expression and communication and metaphor as a form
of conceptual representation and symbolization. Analogously, literary the-
ory and criticism also address the issues of metaphor in these two senses,
metaphor as literary or poetic device and metaphor as symbol. Most often,
the two senses are treated as if they were independent of each other, with
Lakoff and his colleagues (Lakoff & Johnson, 1980; Lakoff & Turner, 1989)
a notable exception. We return to this issue in chapter 6. For now, we
brie?y consider metaphor as linguistic expression.


Metaphor as Substitution

In the Poetics (chapter 21), Aristotle proposed four types of metaphors:
genus for genus, genus for species, species for genus, and analogy. The
?rst three types share a common characteristic: the substitution of one
word for another.3 Genus-for-genus metaphors have received the most
attention in contemporary treatments of metaphor, where they are usually



METAPHOR: THE CENTRAL TROPE




5


referred to as nominal metaphors and predicative metaphors. Nominal
metaphors substitute, in Aristotle’s terms, one noun for another, as in some
lawyers are sharks. The metaphor vehicle, sharks, is used instead of a word
that belongs to the same genus—that is, category or semantic domain—
as the metaphor topic, some lawyers.4 A serious dif?culty with the substi-
tution notion is immediately apparent. The noun shark presumably sub-
stitutes for some other noun that is in the same semantic domain as “some
lawyers,” but what is substituted for what is unclear: what noun might
that be? Similarly, in predicative metaphors, verbs are said to be substituted
for one another, as in the guard dog ?ew across the backyard to challenge the
intruder. Dogs cannot literally ?y, but the verb ?ew substitutes for some
other verb that could literally denote an action that dogs can perform.
However, as in the nominal case, what might that other verb be?
Substitution is not as problematic in the two types of metonymy in
which genus substitutes for species or species substitutes for genus.5 In
these expressions, the two terms are drawn from the same semantic do-
main, and the substitution involves level of speci?city, rather than semantic
domain.6 Using a more general term such as insect to refer to cockroaches,
as in the insects scurried when she switched on the kitchen light, would tech-
nically be considered a genus-for-species metaphor. In expressions that
substitute genus for species or species for genus, what substitutes for what
is clear. A more general term, such as insect, can substitute for a more
speci?c term, such as tick, which in turn can substitute for a still more
speci?c term, deer tick. However, it may be inappropriate to refer to such
expressions as metaphors, or even as instances of substitution per se.
Whenever we make reference, we choose a level of speci?city that is ap-
propriate in context. What principles govern such choices?
At the most general level, I assume that people follow Grice’s (1975)
cooperative principle. When speaking, people implicitly cooperate with one
another in order to further the purposes of their conversation (see also
Clark, 1996, for an extended analysis of conversation as a collaborative
activity). In order to be cooperative, people try to be relevant, truthful,
clear and informative. A speaker’s choice of level of speci?city should be
guided by this consideration. For example, if I need change for a soda
machine, I would not ask my companion for a 1989 quarter, nor would I
simply ask for a coin. The former is overly speci?c, the latter too general.
The appropriate level of speci?city in the context would identify the coin
or coins that I need for the soda machine—a quarter if that’s what I need,
a dime or a nickel if these are the coins that I need.
What governs the default level of speci?city? The default level of re-
ferring expressions is usually at the basic level of categorization. Anything
can be referred to at various levels of speci?city. A piano can be referred
to at the superordinate level, musical instrument, at the subordinate level,
grand piano, or at an intermediate level, piano. The intermediate level is
commonly known as the basic level (Rosch, 1973, 1978), a level that usually
suf?ces for conversational reference that is neither overly general nor over-



6




UNDERSTANDING FIGURATIVE LANGUAGE


ly speci?c. The basic level can thus be characterized as the level of usual
utility (Brown, 1958a), which usually becomes the default level of speci-
?city. Unless a context indicates otherwise, people use the default (usual)
level of speci?city, which in most cases is the basic level. If a more speci?c
or more general level would be more informative, then speakers should
choose accordingly. For someone in need of a quarter for a soda machine,
the basic level “coin” would be inappropriate: “quarter” would be the
appropriate choice. For a panhandler on a street corner, some change would
be the most appropriate. Money would be too general, nickels, dimes, or
quarters too speci?c. These examples illustrate systematic and contextually
appropriate departures from Brown’s level of “usual” utility. Viewed in
this light, substitution of genus for species and vice versa is a choice of
level of speci?city, not a choice to use ?gurative instead of literal language.
Substitutions of superordinate for subordinate and the reverse, while tech-
nically instances of metonymy, are best characterized as ordinary literal
language, albeit tailored to suit particular situations.
Other types of metonymic substitutions seem more ?gurative, as when
a part of something is used to refer to the whole, for example, wheels to
refer to automobiles, as in she’s really proud of her new wheels. These types
of metonymy do not involve substitutions between levels of speci?city but
instead substitute a term that is associated in one or another way with the
intended referent. In such expressions, places can refer to their occupants,
as in The White House is stonewalling the special prosecutor; an object can
refer to its user, as in the glove at shortstop made two errors in yesterday’s
game; a part can refer to a whole, as in the bigmouth went down in the ?rst
round. As Turner (1987) points out, an expression may be used to refer to
anything that it is conventionally associated with. Thus, people’s names
can be used to refer to their works, as in I read Jane Austen every night
before bedtime or Harry bought another Hockney last week. The constraints
on such referring expressions seem to be primarily cultural and often de-
pend on the relative distinctiveness of the association. Thus, I could say
that I love Sarah Lee to mean that I love the cakes that are sold under
that name. However, I could not felicitously refer to a steak grilled by my
friend Bob as I love Bob (Gibbs, 1993). In addition to culturally shared
associations, speci?c contexts can enable metonymic reference, as when a
waiter in a restaurant says that the lobster will be having white wine tonight
to refer to a customer who had ordered lobster (Nunberg, 1979). But these
latter kinds of usages, as in the levels-of-speci?city substitutions, don’t
seem to be metaphoric or ?gurative at all. What kinds of metonymies
function as metaphors?
One possibility is that metonymic expressions function as metaphors
when they involve transference, in the original Greek sense of the term.
One form of such metonymic transference is a bridging or mapping be-
tween the abstract and the concrete (Gibbs, 1994), as in such substitutions
as bench for the law, car bomb for terrorism, pen for writer, and oval of?ce
for the presidency. Another form of transference conveys or emphasizes a



METAPHOR: THE CENTRAL TROPE




7


salient characteristic of the referent by substituting that characteristic for
the referent itself, as in kitchen for chef, arm for baseball pitcher, skirt for
young woman, and lip for a brash, talkative person. When, however, a re-
ferring expression functions solely to identify an entity and nothing more,
then it is not considered metaphoric. The restaurant practice of referring
to patrons in terms of their orders (e.g., the hamburger wants a Coke)
functions in this way, identifying a referent without characterizing it in
any way.


Some Problems with the Concept of Substitution

The nature of substitutions in metonymic expressions is, as we have seen,
clear and unproblematic. However, the metaphoric status of such expres-
sions is not quite so clear and unproblematic. In the most important of
tropes, genus-for-genus metaphors, metaphoric status is not at issue, but
the nature of the putative substitution is. Following Gibbs (1994), I will
use the term metaphor in its narrow sense to refer to expressions that
involve two conceptual domains and metonymy to those that involve just
one (see note 6).
Metaphors pose a thorny problem for the substitution view. In expres-
sions such as man is a wolf, what word does wolf substitute for? According
to the substitution view, the metaphor resides in replacing a literal ex-
pression with the metaphorical wolf. But there does not seem to be any
way to transform the metaphor into a literal statement by replacing wolf
with a literal equivalent. The obvious reason is that there is no literal
equivalent of wolf in this context, certainly not a single-word equivalent.
If there were, then there would be no reason to use a metaphor rather
than a literal expression.
What, then, is the nature of the substitution in metaphors, particularly
metaphors of the form X is Y, where X and Y are from different conceptual
domains? The substitution seems more an implicit act than a concrete
substitution such as we have seen in metonymic expressions. The substi-
tution argument does not seem to go beyond the claim that a metaphor
vehicle such as wolf substitutes, in some unspeci?ed way, for a term or
terms from the same semantic domain as the topic, man. Alternatively, the
substitution could be taken to refer to any literal paraphrase of the meta-
phor vehicle, so that wolf might substitute for a literal rendering of the
metaphor ground. In the metaphor man is a wolf, wolf substitutes for some-
thing like “a predatory creature, stealthy and vicious, with ?erce loyalties
to the pack (read family, group, country, gang, etc.), etc. . . .” The etc. here
is the rub, as is the open-ended nature of the literal paraphrase of the
ground. There is no single, de?nitive interpretation to be given of this or
any other nontrivial metaphor. Metaphor interpretations are constructed
from the meanings of the two metaphor terms, topic and vehicle, often
with the context of the conversation playing an important role. For ex-



8




UNDERSTANDING FIGURATIVE LANGUAGE


ample, elephant in Western society can serve either as a symbol of enor-
mous size or as a symbol of prodigious memory. The metaphor David is
an elephant can thus be taken to mean either that David is a very large
man or that he has a prodigious memory. Does the phrase “is an elephant”
substitute for “is a very large man” or “has a prodigious memory,” or does
it instead provide these characterizations as possible attributes of David?
Because of the insoluble problem of specifying exactly what is substi-
tuted for what in metaphor, the strong form of the substitution view has
sunk into oblivion, but not without leaving a signi?cant trace. Its basic
assumptions survive, recast in terms of standard pragmatic theory. These
assumptions include:
1. Literal meaning is basic and has unconditional priority. Implicit in
this assumption is a corollary assumption: that literal meaning is unprob-
lematic and is context-free, that is, the literal meanings of expressions
remain unchanged regardless of context of use.
2. Figurative meaning is derived from the literal and can be discovered
by discovering the nature of the substitution of the metaphorical for the
literal. One implication of this assumption is that metaphoric interpreta-
tions involve recovering the original literal expression for which the meta-
phor substitutes.
3. It follows from assumptions 1 and 2 that metaphor understanding is
more complex and requires more cognitive work than literal understand-
ing. Metaphor understanding also requires the use of contextual infor-
mation, which literal understanding, by de?nition, does not.
For these reasons, among others, many linguists and philosophers of
language take the position that metaphor lies outside theories of meaning
or semantics. Linguists are concerned with the relations between form and
meaning, that is, the meanings of individual words and the meanings of
sentences that can be derived compositionally, that is, by rule, from the
meanings of the sentence constituents. When the meaning of an utterance
cannot be speci?ed in purely linguistic terms, then that kind of meaning
is simply excluded from consideration. Sadock exempli?es this view: “All
nonliteral speech . . . including metaphor, falls outside of the domain of
synchronic linguistics . . .” (Sadock, 1993, p. 42). Philosophers of language
have traditionally been concerned with the truth conditions for expres-
sions. In their view, to know the meaning of an expression is to know the
conditions under which that expression would be true or false. Thus, Da-
vidson (1977), for example, emphatically denies that metaphors have a
metaphorical meaning over and above their literal meaning: “metaphors
mean what the words, in their most literal sense, mean and nothing more”
(p. 246). For Davidson, literal meaning is linguistic meaning: it is inde-
pendent of context, completely systematic and rule governed. Utterances
are understood by interpretation, roughly, by ?rst arriving at the literal
meaning of an utterance and then by inferring what that literal meaning
is used for. If I say it will rain tomorrow, there is only one literal meaning,
and that meaning is the ?rst one to be derived by an interpreter. The



METAPHOR: THE CENTRAL TROPE




9


interpreter can, however, infer (in principle) an in?nite number of alternate
interpretations. I may use the utterance to convey my belief that a partic-
ular weather forecast is wrong, or that we should not plan on going to the
beach, or that the crops won’t fail after all because the drought will end
tomorrow, ad in?nitum. These and all other possible interpretations of my
use of the utterance it will rain tomorrow are not alternative meanings but,
rather, alternative uses.
The distinctions among form, meaning, and use are motivated by a
commitment to language as a logical system, and that meaning resides in
truth conditions. To know the meaning of a sentence, on this classical
view, is to know the circumstances under which it would be true or false
(Miller & Glucksberg, 1988). Thus, to know the meaning of “rain is dry”
requires not a belief that the statement is true or false but simply an
understanding of the conditions, in all possible worlds, under which the
statement would be true or false.7 More generally, language is a system of
learned conventions and regularities that enable literal meanings to be de-
rived independent of context or occasion of utterance. Once literal mean-
ings are derived, then the work of interpretation can begin.8
The distinction between meaning and use in truth-conditional semantics
is analogous to the distinction between sentence meaning and utterance
meaning in standard pragmatic theory. As initially proposed by Grice
(1975), there are at least two kinds of logic involved in discourse compre-
hension: the logic of language and the logic of conversation. The logic of
language applies to literal or linguistic meanings. The logic of conversation
applies to the rules that people use to infer what a speaker intends to
convey, beginning with the literal meaning of an utterance and ending with
an utterance meaning (also known as speaker meaning, intended meaning,
or conveyed meaning). Grice’s maxims of conversation follow from his
cooperative principle: listeners assume that speakers will be truthful, rel-
evant, clear, and informative. Normally, when a speaker intends the literal
meaning of an utterance, these rules do not come into explicit play. How-
ever, these rules are systematically invoked whenever a speaker appears to
violate a conversational maxim. Listeners assume that speakers will be
cooperative. Therefore, whenever a maxim appears to be ?outed, it will
function as a conversational implicature—a signal that the speaker intends
something other than the literal meaning of what has been said. If, for
example, someone replies to an invitation to go to the movies by saying,
“I have an exam tomorrow morning,” then that reply will be taken as a
rejection of the invitation. On the face of it, saying that one has an exam
tomorrow is not a relevant response to an invitation. However, if the reply
is assumed to be relevant, then it can be taken to imply that the person
can’t spare the time to go to the movies because she has to study for an
impending exam.
Analogously, when someone says something that is literally false, then
this too should function as a conversational implicature, that is, as a signal
to search for a meaning that is not the literal meaning. When a proud



10




UNDERSTANDING FIGURATIVE LANGUAGE


father says, “My daughter is an angel,” no one believes that she has wings.
But a metaphor need not be literally false. The opposite assertion—that
one’s daughter is no angel—is literally true; she does not have wings. Yet
this is not likely to be the speaker’s intended meaning, nor is it likely to
be a hearer’s interpretation. In each of these two cases, hearers must go
beyond the literal meaning to arrive at the speaker’s intention—what the
hearer is intended to understand. Searle put the issue clearly: “Where the
utterance is defective if taken literally, look for an utterance meaning that
differs from sentence meaning” (1979, p. 114).
This straightforward dictum leads directly to the standard three-stage
model of nonliteral language comprehension proposed in linguistics (Ly-
ons, 1977), as well as in philosophy (Grice, 1975; Searle, 1979) and in
psychology (Clark & Lucy, 1975; Janus & Bever, 1985):

1. Derive the literal meaning of an utterance.
2. Test the derived literal meaning against the context of the utterance.
3. If the literal meaning makes sense, accept that meaning as the utterance
meaning, that is, the speaker’s intended meaning. If it does not make sense,
then seek an alternative, nonliteral meaning that does make sense in the
context.

The application of this general model to metaphor is straightforward.
Metaphors are usually (although not always) false. They are therefore de-
fective, in Searle’s sense. When the hearer recognizes a metaphor as being
false (or otherwise defective), she therefore implicitly transforms it from
its false categorical form to its correspondingly true simile form. For ex-
ample, the utterance my lawyer is a shark is literally false. By transforming
it into the simile My lawyer is like a shark, the listener reframes it in such
a way that it becomes true and therefore can be understood in the same
way that any comparison assertion might be understood. Metaphors, on
this view, are essentially implicit similes and so pose no further problems
of interpretation.
There are four testable implications of this model. The ?rst and most
obvious is that literal meanings are unproblematic and context-free. The
second is that literal meanings have unconditional priority. Because they
have unconditional priority, they will always be derived ?rst, before any
nonliteral meanings are even attempted. Hence, they will always require
less time and effort than nonliteral meanings. Third, because nonliteral
meanings are sought if and only when literal meaning is “defective,” there
is an important difference between literal and nonliteral meanings. Literal
meanings are derived automatically, but nonliteral meanings are derived
only optionally. The term automatic here does not mean that literal un-
derstanding is not effortful or does not require complex computation. In-
stead, it means that a ?uent speaker of a language has no voluntary control
over whether or not an utterance such as “rocks are hard” will be under-
stood. The language comprehender is data driven in the sense that given
a linguistic input, it will process that input and generate a literal interpre-



METAPHOR: THE CENTRAL TROPE




11


tation willy-nilly, whether or not it intends to (cf. Fodor, 1983). As Miller
and Johnson-Laird put it, understanding “occurs automatically without
conscious control by the listener: he (sic) cannot refuse to understand . . .
loss of control over one’s [language comprehension system] may corre-
spond to knowing a language ?uently” (1976, p. 166). According to the
three-stage model I have outlined, such automaticity does not apply to
nonliteral understanding. Nonliteral understanding must be triggered by
the failure of a literal meaning to make sense in context. The fourth im-
plication is that, because metaphors are (usually) literally false, they are
implicitly transformed into true comparison statements and interpreted via
a comparison process. I argue that each of these four claims is wrong. In
this chapter, I try to show that literal language understanding is not context
independent and unproblematic. In chapter 2, I examine the evidence
showing that (a) nonliteral understanding is not in principle more effortful
or more complex than literal understanding; (b) nonliteral understanding
can be as data driven (i.e., automatic and nonoptional) as literal under-
standing; and (c) metaphors are not implicit comparisons and so are not
understood via a comparison process.


Literal Meaning: Some Problems and Issues

What distinguishes literal from nonliteral meanings? There are two ques-
tions here. The ?rst concerns how people judge whether a given interpre-
tation is literal or not. The second concerns the ways in which literal
meanings and nonliteral meanings are generated: do they rely on the same
or on different sets of language-processing principles and mechanisms?
Consider, ?rst, the issue of recognizing whether or not a given meaning is
literal. How do people recognize that an utterance is literal rather than
nonliteral?
To address this question we need to distinguish between two kinds of
operations: linguistic decoding and linguistic interpretation. Linguistic de-
coding involves only those operations that are theoretically de?ned as lin-
guistic, namely, phonological, lexical, and syntactic operations. I stress
“theoretically de?ned” because the literal meaning of an utterance cannot
be identi?ed apart from the linguistic theory that supplies the mechanism
for analysis. Stern (2000) puts the issue succinctly: the “literal meaning of
a simple expression is whatever our best linguistic theory tells us is its
semantic interpretation . . . [The] literal meaning of a sentence is the rule-
by-rule composition of the literal meanings of its constituents” (p. 23).
On this view, literal meanings are an abstraction, restricted to what
Lyons (1977) refers to as maximally decontextualized system-sentences.9
Linguistic-literal meanings are thus the products of a particular (one hopes,
the “best”) theory of semantics and syntax, a theory that does not pretend
to describe or explain what people actually do when talking and listening.
Does this abstract theoretical construct have any functional utility in help-



12




UNDERSTANDING FIGURATIVE LANGUAGE


ing us understand how people decide whether an interpretation is literal
or metaphorical?
Indirectly, yes. Just as theoretical linguistics de?nes the literal in terms
of particular theories of the language system, so do speakers of a language
de?ne the literal in terms of their folk or intuitive theories of language.10
Dictionaries presumably re?ect common usage, and the Oxford American
Dictionary provides a clue to our folk theory of language in the de?nition
of the word literal: “in accordance with the primary meaning of a word or
the actual words of a phrase, as contrasted with a metaphorical or exag-
gerated meaning.” The entry also notes that “the word literally is some-
times used mistakenly (sic) in statements that are clearly not taken literally,
as in he was literally glued to the TV set every night” (1980, p. 386). In our
folk theory of language, words have primary meanings, and the literal
meaning of a phrase or sentence is one that does not go beyond the primary
meanings of the phrase or sentence constituents. Apart from formal lin-
guistic theories, how are the primary, that is, literal, meanings of words
identi?ed?
One possibility would be to use the criterion of context dependence. A
commonly held view is that literal language is real, true, unambiguous,
and relatively context independent. Literal meaning is context independent
in the sense that the meaning remains the same irrespective of the context
of utterance. For example, we have the intuition that the sentence dogs are
animals literally means the same thing no matter who utters it, when or
where or to whom and under any circumstances. Nonliteral, in contrast,
is felt to be open to alternative interpretations. A literal interpretation of
the utterance dogs are animals would be something to the effect that dogs
belong to the category of animals, as opposed to vegetables, minerals, or
abstract ideas. A nonliteral interpretation could be something to the effect
that dogs behave as they do because they are animals. The particular in-
terpretation depends on the context of utterance. If the utterance is a reply
to the complaint “Rex doesn’t seem able to control his barking at night,”
then the assertion that dogs are animals might be an indirect way of saying
that this behavior is not surprising because animals rarely can control
themselves. In this interpretation, both the literal meaning and the addi-
tional indirect meaning constitute the conveyed meaning.
Under what circumstances can people make a purely literal interpre-
tation—that dogs belong to the category of animals, period? It is dif?cult
to imagine a context in which such a barebones interpretation would be
made. Even if the assertion dogs are animals were intended to inform a
listener of this fact, there would still be some interpretive work to do
beyond the minimalist literal; for example, what alternatives to “animal”
are intended? Are dogs animals as opposed to plants, or are dogs animals
as opposed to rocks? As this example illustrates, even literal interpretations
require contextual information beyond the linguistic meaning. The picture
becomes even messier when we consider words whose meanings cannot be
identi?ed or speci?ed without considering the context of use.



METAPHOR: THE CENTRAL TROPE




13


The logical connectives are one class of words whose meanings are
context dependent. In formal logic, if-then has a speci?ed, context-
independent meaning. The assertion if p then q is true when the following
conditions hold: p and q, not p and q, not p and not q. It is false when we
have p and not q. If we substitute concrete events for p and q, then some-
times the logical meanings work, but often they do not. If instead of p and
q we substitute the sun shines tomorrow and we will go to the beach, then we
can consider the following alternative states of affairs:


1.
2.
3.
4.


p and q: The sun shines and we go to the beach.
p and not q: The sun shines and we do not go to the beach.
not p and q: The sun doesn’t shine and we go to the beach.
not p and not q: The sun doesn’t shine and we do not go to the beach.


Recall that if p then q is false only when we have p and not q, (ex. 2).
Alternatives 1 and 4 seem to follow from the assertion if the sun shines we
go to the beach, and so, both logically and pragmatically, they do not falsify
this assertion. The third outcome poses a problem. Even it does not log-
ically falsify the if-then assertion, it does seem to violate conversational
expectations. To compound matters, the interpretation of if-then depends
on particular contexts of use. There is a logical implication in the assertion
If you mow the lawn, then I’ll pay you $5.00. This implies that if you do
not mow the lawn, then I will not pay you $5.00. No such logical impli-
cation appears, however, in the assertion If you are a U.S. senator, then you
are over thirty-?ve years old. This does not imply that if you are not a U.S.
senator, then you are not over thirty-?ve years old. In the latter case, the
if-then expression states a prerequisite condition for being a senator; it does
not state a logical relation.
Other connectives display the same variability in natural language use.
In logic, p AND q is true whenever p is true and q is true. In natural
language, the word AND can be used to express a variety of relations:

1. Do that one more time and I’ll smack you (an if-then relation).
2. Mark is a genius and there are twenty inches in a foot (sarcastic denial of
Mark’s genius status, paraphrasable in the if-then form).
3. Mark is a lawyer and Mark is a lawyer (logically true, conversationally inane).

In natural language, then, the primary meanings of connectives are not
context independent, yet people treat them as literal nonetheless.
Other classes of words whose meanings are explicitly context dependent
include quanti?ers (some, a few, many), deictic terms (here, there, in this
place), adjectives (good, tall, expensive), and pronouns (he, she, them).
One cannot know what these terms refer to outside their contexts of use.
A few people in the kitchen would be taken to mean four or ?ve people; a
few people in the football stadium might mean several thousand (Horman,
1983). Christmas is here refers to a time. The newspaper is here could mean
that it has been delivered, is available at the newsstand, or is in the kitten’s
litter box. If the ?rst of these, then we already have it to read; if the



14




UNDERSTANDING FIGURATIVE LANGUAGE


second, it is potentially available if we go out and buy it; if the third, it is
not something to be read at all. And what are we to make of words such
as good when meanings can vary so enormously (Bierwisch, 1967):

1. He got a good whipping for being late; good
2. Harry Truman was a good president; good

painful?
honest, effective?

3. Hannibal Lecter was more than just a good villain; good

ruthless, vicious,

terrifying?
Are these examples limited to these particular classes of words, or do we
also see context dependence in ordinary nouns and verbs?
The word line couldn’t be more ordinary. It is a word that is used very
frequently, and as a noun its primary or core meaning involves the notion
of extension (Caramazza & Grober, 1976). But, even with this common
semantic feature of extension, different contexts of use induce different
interpretations:

1.
2.
3.
4.
5.
6.
7.

Sam owned the local bus line.
She said it was a line from Keats.
The rich man was able to line his pocket with money.
Sergeant Jones would bring him into line.
The judge had to draw a line between right and wrong.
I pulled on the line with all my strength.
The shortest distance between two points is a straight line.

Analogously, the word open has different, albeit related, interpretations in
different contexts.

1.
2.
3.
4.
5.
6.
7.

He opened the box of cookies.
She opened the conversation by commenting on the weather.
The surgeon opened the patient’s chest.
He opened his eyes.
She kept an open mind.
He opened Pandora’s box.
He opened her eyes to her husband’s in?delities.

There seems to be a common core of meaning to all of these uses of open,
even when it is used metaphorically, as in 6 and 7 (and perhaps 5?). What
distinguishes the literal uses from the metaphorical ones? It cannot be in
the way that context is used to arrive at an interpretation. It also cannot
be in any differences in conventionality. The expression open mind may
not even qualify as metaphorical, while opening Pandora’s box and opening
someone’s eyes to something border on cliche´.
Perhaps the most useful position is that the concept of literal cannot be
explicitly de?ned except in formal linguistic-theory terms. Within our folk
theory of language, we make a sharp distinction between the literal and
the nonliteral. However, when we make judgments about speci?c examples,
the distinctions become graded, rather than discrete. People can make re-
liable judgments about degrees of metaphoricity, for example, suggesting
that there is a continuum from the literal to the nonliteral (Ortony, 1979).



METAPHOR: THE CENTRAL TROPE




15


In this respect, the concept of literal (or the concept of metaphorical, for
that matter) behaves as do other natural-kind concepts. For natural-kind
concepts, such as fruits, there are clear, prototypical examples, such as
apples, pears and bananas. There are also not so clear, nonprototypical
examples, such as pumpkin, tomato, and olive (McCloskey & Glucksberg,
1978). People are unanimous when asked if an apple is a fruit but disagree
about tomatoes, even though both apples and tomatoes are, technically
(literally?) fruits: they are the fruiting bodies of their plants and have seeds.
Our best de?nition of the concept of literal meaning, then, is analogous
to our best de?nitions of natural-kind concepts. On the one hand, experts
have explicit theories for assigning candidate exemplars to their appropriate
categories. Linguists have a theory of the lexicon, of syntax, and semantics.
Analogously, botanists have a theory of plant life and biological taxonomies.
Within such theories, clear categorical distinctions can often be made. On
the other hand, lay persons have implicit folk theories from which to make
categorical judgments, and, in both the language case and the botanical
case, graded judgments seem to be the rule, even when elements of the
technical theory are, in principle, available. Thus, even though we may
know the technical de?nition of fruit as “the edible product of a tree, shrub
or other plant, consisting of the seed and its envelope” (Oxford English
Dictionary, 1996), we are reluctant to say that a tomato, a pumpkin, or an
olive is a fruit. Our folk theory of plants and foods leads us to assign these
exemplars to different categories on a probabilistic, rather than a deter-
minate, basis. Similarly, even though we may know that the expression
glued to the TV set does not use the primary meaning of the verb to glue,
we still feel that this is a perfectly straightforward, literal-like usage. Per-
haps it is this double awareness—of the technically nonliteral and of the
simultaneously perfectly straightforward usage—that prompts people to
produce what is technically a contradiction: describing metaphorical ex-
pressions as literal, as in he was literally glued to the TV set. Technically,
glued to the TV set is a metaphor, but intuitively it is literal, just as tech-
nically a tomato is a fruit but intuitively it’s a vegetable.
These examples imply that metaphor appreciation and metaphor un-
derstanding may be independent of one another (see Gerrig & Healy,
1983). Much like Moliere’s character M. Jourdain in Le Bourgeois Gentil-
homme, who was amazed and delighted to learn that he spoke in prose
(Moliere, 1675), people may use and understand metaphorical expressions
without being aware that the expressions are metaphorical at all. This
should certainly be the case if metaphor and literal understanding depend
on the same linguistic, cognitive, and pragmatic principles. However, our
intuitive folk theory of language views literal meaning as primary meaning.
Is this merely an intuition about the appreciation of a difference between
literal and metaphorical expressions, or does it also re?ect differences in
modes and ease of processing? We consider the evidence on this issue in
chapter 2.