Poetic and Prosaic Metaphors 
Geoffiey Nunberg 
Xerox PARC and CSL\[, Stanford 
Metaphor (some people say) is "the dreamwork of language," where ideas are 
combined in ways that are "special, odd, or startling." Metaphor is a "mutation" of 
language, a phenomenon that relies crucially on our recognition of its "deviance" or 
"nonsensicality." Metaphor, if it means at all (it may only "suggest" or "evoke"), does not 
mean as literal language does. Metaphor is the stuff of "playful prose and high poetic m't." 
Metaphor (say others) "covers the entire semantic, semiological and performative 
field of language" and "is natural and widespread in our speech, not having any apparent 
special status." 
Well, but (we say) there are ways and ways of being a metaphor. What some people 
are talking about are the "fresh" and "vivid" figures that puilulate in poetry, play, and the 
rudimentary stages of theory-building. For them, metaphor is essentially an rhetorical 
category, identified by its affective import. Whereas other people take metaphor to be 
essentially a semantic phenomenon, which crucially involves a perception of transfer of 
extension of meaning or application. 
In the standard story, moreover, there is no real conflict between these views. 
Metaphors begin their lives as novel poetic creations with marked rhetorical effects, whose 
comprehension requires a special imaginative leap. As time goes by, they become a part of 
general usage, now "stale," "worn," and "idiomatic"; their comprehension becomes more 
automatic, and their rhetorical effect is dulled. Finally, as "dead" or "frozen" metaphors, 
they are not psychologically distinct from literally-used terms. In point of fact, "most 
lexical items prove to be dead metaphors that ~vere alive and kicking at some time in the 
past." The natural conclusion is that there is a difference in conversational status between 
the "poetic metaphors" that strike our fancy and the "prosaic metaphors" that pass by 
unnoticed in our ordinary discourse. The former arise out of conversational implicatures, 
indirect speech acts, or whatever, while the latter are matters of literal or conventional 
meaning. Not surprisingly, this leads in turn to the assumption that we ought to be able to 
explain the rhetorical effect of poetic metaphors in virtue of their conversational properties, 
as something that grows directly out of the process of inference required in non-literal 
comprehension. 
The standard story does explain some of the differences in our intuitive reactions we 
have to metaphorical utterances. It is beyond dispute that metaphors do work their way 
into literality, and may lose some of their luster along the way. If you say in English "he 
answered in the wink of an eye" you occasion no special rhetorical effect; but if you say in 
French "I1 a repondu en un clin d'oeil," you may get credit for a slightly clever turn of 
phrase, ~imply because the French have not made this expression a commonplace. 
But there is more to the distinction between "poetic" and "prosaic" figures than 
that; ultimately, it transcends dimensions of relative novelty and conventionality. On the 
one hand, there are extensions of meaning that are so natural, so obvious, that they must 
surely be have been counted prosaic on first utterance--indeed, they must have been born 
along with the literal meaning of a term. For example, I imagine that the discovery that the 
word chicken could be used to refer to chicken meat could not have taken too long once it 
had been established that the bird was edible. Nor did it want a vast leap of imagination to 
realize that the word sun could be used to refer to the volume directly illuminated by the 
sun (as in "we were standing in the sun") or (to take an example from Sam Levin) that the 
198 
word tremble, once applied to the shaking or quivering of an animal, would do as well to 
describe the oscillations of the earth or of leaves in the wind. In many cases like these, in 
fact, we may be reluctant to speak of"figures of speech" at all, and say rather that the words 
chicken, sun and tremble are "polysemous," and have several "related literal meanings." 
Now if by this we mean only that the extended usages are prosaic and common, and that 
hearers probably do not work them out afresh on every hearing, then \[ have no objection to 
calling these "literal meanings." But it is a mistake to try to argue from our intuition for the 
unremarkableness of these usages to the conclusion that they are prescribed by 
language-specific rules or convention, or that the figuration that underlies them is in any 
sense "dead" or "idiomatic." The uses are unremarkable precisely because they are 
obvious, and so can take hold in the general practice without the introduction of any special 
rules. (If you are just learning Italian, say, and know that i/sole is the word for the sun, it 
wouldn't occur to you to check before using the same word to refer to an volume 
illuminated by the sun. You'd figure---rightly--that the Italians would make the same 
obvious moves that we do.) In any event, it is clear that the figuration involved in these 
cases has never had and could never have had the power to startle us. 
Much the same point can be made about other transfers of application, which are 
licensed only by the circumstances of a particular narrow context, and so could not become 
part of conventional usage. Say you want someone who knows nothing about cars to hand 
you one of those little tools you use to pry open a spark-plug gap, which in this case 
happens to be shaped like a clothespin. "Hand me the clothespin," you say metaphorically. 
Now it's unlikely that your interlocutor has ever heard clothespin used in this way, nor 
could that word ever become the standard means of referring to spark-plug gap adjusters 
(which come in different shapes). But your remark is clearly lacking in the divine emollient 
of poetry. 
On the other hand, there is plenty of poetry in many of the metaphors that have 
been adopted by the general use, even those that have become in some measure 
conventionalized. Indeed, it is precisely because they are evocative that we have made 
idioms of expressions like knock one's socks off lick one's wounds, or steal someone's 
thunder. It is true that there is a difference between the effect of these figures and those 
offered in "real poetry" (which by our lights must own an individual author). But there is 
no missing the spirit of play in these figures, which is absent in the use of clothespin to 
identify a nameless tool, or of chicken to refer to a kind of meat. This pleasure survives 
even after the motive for an idiomatic metaphor has been largely obscured. Who knows 
why we rob Peter to pay Paul, rather than Max to pay Marvin? What is clear is that the 
personification of exigent circumstances gives them a satisfyingly lighthearted cast, and so 
m some measure alleviates them. 
What I'm calling "poetic" metaphors are just those that we ,associate with "play," 
"affect," "evocation," "pleasure," or whatever. (It is not important that we be able to say 
what pleasure is, so long as we know it when we feel it.) What I'm calling "prosaic 
metaphors," on the other hand, are those that strike us ,as involving no-nonsense referential 
business-as-usual. 
Where does the difference between the two types of figures come from? Not, we 
have seen, from differences in the degree of their conventionality, Nor, surely, from the 
nature of the mechanism involved transfer or reapplication of the meaning of an expression. 
The same principles of transfer (based on physical resemblance) that license the prosaic 
application of clothesp#t to a tool result in a poetic (if not exactly sublime) figure when the 
word is applied to a skinny person. It is not poetic to use beerto refer to the interval of time 
199 
over which beers are served, as in I could have toldyou the answer two beers ago; it is poetic 
to use grief in the same way, as in Dylan Thomas's a grief ago. 
What makes all the difference here, rather, is the degree to which a usage is justified 
in terms of a strictly instrumental rationality. I take it as obvious that metaphor and 
figuration are ways ofd0ing whatever we do with expressions: referring, predicating, and so 
forth. That is, figuration is a certain way of identifying objects, properties, and relations. 
For the present purposes, I'll stick to the referring uses of expressionsmthe clothespin, that 
witch, and so on. Let's say that a figurative use of such a expression is instrumentally 
rational in a given context if a rational agent would conclude that this use offers the speaker 
an optimal probability of success in enabling the hearer to identify the thing he is talking 
about, for such purposes as are conversationally relevant, and that it is that consideration 
that motivated the speaker to use the expression in that way. (This definition would require 
various qualifications before it could be airtight, but it will do for here.) 
Take the case in which we want to enable someone to pick out a tool from among an 
array. In this case, the tool has no commonly used name, and even if it did, we could not 
expect that an automotively unsophisticated hearer would be able to identify this tool solely 
on the basis of its appearance, armed only with its name. A description of its function ("the 
gap adjuster") would be of no more use, for the same reasons. We are left with getting the 
hearer to identify it in virtue of its appearance, either by describing it in such terms ("the 
little thing with two prongs"), or by describing it using the name of something elsema 
clothespin, a fork, whatever--that it resembles more than anything else in the array. Now 
there are any number of factors that might make one or the other of these approaches the 
optimal strategy for accomplishing reference. It depends how easily the protrusions on the 
tool are identified as prongs, how much thetool looks in silhouette like a typical clothespin, 
how likely it is in this context that the speaker might have been looking for an actual 
clothespin, or that the hearer might have thought he could have been looking tbr an actual 
clothespin, and so on. Then too, you have to calculate the likelihood--here, pretty good-- 
that the hearer will know that the speaker will be going after the referent in virtue of its 
appearance, rather than its function or something else. And of course you have to run all 
these factors through various strategies for arriving at common belief. But it is clear enough 
that the speaker thought it was an optimal way of accomplishing reference, and that that's 
why he used the term. At the very least, it would be irrational to ascribe to him any ulterior 
motive. 
Now suppose you're trying to get me to try on a jacket with wide horizontal stripes, 
and I, ample, say "Not me. Offer it to the clothespin you were just talking to." A poetic 
figure, in our (extremely generous) sense of the term. And it counts as poetic for two 
reasons. First, it's pretty likely that I could' have used another expression--say the skinny 
fellow, though depending on the circumstances it might do as well to say the man with the 
red cap, or Jackson--if I were interested only in optimizing the probability that you could 
pick out the person I had in mind ~ a reasonable customer for the jacket. This, on the 
assumption that it's easier to pick out skinny than to pick out clothespin-resembling, at least 
so far as any array of potential jacket customers is concerned. What's more, you can 
construct an ulterior motive for my saying things the way 1 did. Briefly, I'm behaving in a 
way that wouM be instrumentally rational if the world were different: in particular, if people 
like Jackson were most readily picked out in virtue of their physical resemblance to 
clothespins. (Though in another context, to be sure, I could have been counting on the 
function of clothespins to enable reference to go through.) This would be the case only if 
Jackson looked a great deal more like a clothespin than he did like anything else; indeed, if 
200 
he were more readily identifiable as looking like a clothespin than ,as merely "skinny." Now 
it could be, I suppose, that l am pretending that clothespins are more anthropomorphic 
than they are (and there are metaphors that work this way). But it's a sad fact that we take 
more pleasure in exaggerating the physical peculiarities of people than of household 
implements; it is funnier to imagine Jackson as looking like nothing so much as a clothespin 
than to imagine a clothespin looking like nothing so much as Jackson. 
Now for the use of beer and grief to refer to intervals. Beers do come at publicly 
observable intervals, which are moreover the most relevant stages for chopping up the 
progress of certain evenings. So it is instrumentally rational to use beer to refer to these 
intervals. Whereas the analogous use of grief would be instrumentally rational only in a 
world in which the state was generally accepted as recurring with predictable regularity (and 
probably, in a publicly observable way). To comprehend the usage, then, we have to 
imagine a world in which our inner states were driven by natural cycles quite as the seasons 
are. The reason that poetic metaphors are such an important part of the poet's arsenal is not 
because they say what is false. If they are in fact false-saying (and I would argue they are 
not), they share this property with prosaic metaphors that have no particular rhetorical 
effect. Rather, they force us to presuppose what is false. That is what Shelley had in mind, I 
calculate, when he said (though the italics are mine) that poets are the "unacknowledged 
legislators of the world." 
It is this aspect of play-acting that gives poetic metaphors their affective import, and 
makes them quite literally playful or dramatic, as the case may be. In this way, poetic 
metaphors have a great deal in common with ironic utterances and other tropes that involve 
a pretence (rather than an assertion) that the world is other than it is. Note that it is 
immaterial whether we actually work out these inferences on each occasion of hearing a 
familiar or idiomatic poetic metaphor, so long as we recognize that the usage is in fact 
motivated by such inferences. 
If this much is granted, then most of the stories we tell about the aesthetics of 
metaphor are wrong, or at best misleading. Their impact, when they have one, does not 
come from their "literal falsity," from their "semantic deviance," from their "indirection," 
or from the "compression of two ideas into one." Criteria like these do not distinguish 
poetic from prosaic metaphors. More generally, there is nothing about metaphor qua 
metaphor (or of conveyed-meaning of any sort qua conveyed meaning) that makes it an 
interesting subject for aesthetic investigation. More generally still, it is a mistake to suppose 
that we can derive rhetorical categories from purely conversational principles. 
201 
