Metaphor: Davidson, Searle, and Me
W. Zachary Wolff, 2001

Metaphor poses particularly difficult problems for the philosophy of language. Davidson and Searle have proposed theories of metaphor that stand at odds, each of which I will outline and discuss here. In the end, I hope to give something of my own account of metaphor.

Davidson holds that metaphors have no special 'metaphorical meaning.' Instead, they only mean what their words say. A first argument against this stance is that it denies the existence of metaphorical truth. This is somewhat troubling for artistic metaphor. It seems that Donne said something true or false when he wrote, 'No man is an island,' depending on your view of humankind's nature as inherently social or solitary. Still, perhaps truth-values can be given up in certain artistic metaphors. Donne was neither right nor wrong, he was only prodding us into noticing a certain condition of humankind. I think the error in Davidson's thinking becomes more apparent when everyday uses of metaphor are considered. An adulterous man who tells his wife, 'You are the only star in my sky,' can rightly be called a liar.

Davidson admits the existence of a sort of metaphorical truth, but not nearly the kind I am looking for. Davidson says, 'Metaphor does lead us to notice that which might not otherwise be noticed, and there is no reason, I suppose, not to say these visions, thoughts, and feelings inspired by the metaphor are true or false.' I would like to offer a small correction to Davidson's wording that I do not think changes his view as a whole. It seems inappropriate to say that metaphors make us 'notice' things. For the cheated wife surely does not notice that her husband loves only her. It is impossible to notice things that are not true. Perhaps it would be better amended to, 'Metaphor does lead us to think (or feel) what might not otherwise be thought (or felt).' In any case, this is not the sort of truth that I want to assign to metaphors. It does allow Davidson some escape from the above arguments though. The adulterer is a liar because he has uttered a metaphor that made his wife think something untrue. Or perhaps properly he is not a liar, but only a scoundrel, as is any who seeks to deceive through non-linguistic devices. This allows Davidson to castigate those who use metaphors in ways we may typically think of as lying, while still holding onto his view that metaphors lack truth-values.

A more compelling argument against Davidson stems from the problem of dead metaphors. Dead metaphors certainly have a meaning that differs from their literal meaning. More precisely, dead metaphors have acquired a new literal meaning that is different from their original literal meaning. I think there are two ways of seeing the problem this holds for Davidson's view. The first concerns the existence of the meaning. Where did it come from? Those who hold that metaphors have metaphorical meaning can explain it easily: the metaphorical meaning became literal over time. Davidson has no ready answer. The second way of looking at this problem concerns the history of a piece of language. How can Davidson account for the change in meaning over time? Is there some point in time at which one usage of metaphor fresh and lively, bearing only the literal meaning of its words and acting as a psychological device, while the next is dead and boring, though ambiguous as to its two possible literal meanings?

Davidson must look for his answer in speaker meaning, a la Grice. Although I cannot find it in his writing, I assume that Davidson agrees that a speaker does speaker-mean something when uttering a given metaphor. This seems fairly uncontroversial. This explains where the new meaning comes from. The meaning of a dead metaphor is whatever it was used to speaker-mean when it was living. I still think the moment of meaning change gives Davidson problems. A sudden change seems ludicrous. A gradual change entails equally ludicrous ideas. Does the metaphor 'sort of mean' something for a while? Does it mean 99.9% its literal meaning and .1% its metaphorical meaning (sorry, its speaker meaning) at first and gradually the two balance out? Perhaps Davidson can skirt this issue by spreading the meaning out over a speech community. For any given speaker, the change will be sudden. There will be one use of a metaphor, upon which that speaker decides that what that metaphor means is what it has speaker-meant in its past occurrences. As this wave of change flows across a speech community, the new literal meaning gradually takes hold. Here Davidson runs into a problem in that his view is so counter-intuitive. For a given speaker (at least the large majority of speakers, who are not philosophers) a metaphor means what it speaker-means on its first utterance

Lycan claims that Davidson's view leads to the counter-intuitive notion that metaphors cannot be misinterpreted. Regarding Shakespeare's 'Juliet is the sun,' Lycan says, 'If in response to Romeo's utterance, some eavesdropper had chirped, 'I get it! Juliet depresses him because she's so stupid and smells horrible,' on the Causal Theory this would not have been an incorrect account of Romeo's metaphorical utterance, but only evidence that the eavesdropper's mental architecture was causally different from Romeo's and from ours.' Davidson may attempt something of an escape through the analogy he makes from metaphors to jokes, 'There are no unfunny jokes.' The metaphor has clearly failed here, but the failure is not necessarily the eavesdropper's misinterpretation, nor is it necessarily Romeo (or Shakespeare) who is to blame. I think this attempt by Davidson fails. The artistic merit of a metaphor corresponds to the humor of a joke, but the meaning does not. This statement assumes that metaphors have meaning, which I mean to avoid in order to give Davidson a fair hearing, but I cannot think of another way to put it. Davidson can admit that Romeo speaker-meant something, which the eavesdropper missed. This is leading rapidly to Searle's view, as many of the patches on Davidson's view are. I suppose I should introduce that now.

Searle, like Davidson, holds that metaphors do not have a secondary meaning, unique to them as metaphors, and that therefore there is not any linguistic ambiguity. Searle reduces anything we might call metaphorical meaning to simple speaker meaning. I say simple here because we are already familiar with the concept of speaker meaning, but determining the speaker meaning associated with a given metaphor will prove to be a rather intricate task.

The first tool Searle draws upon is the fact that metaphors generally (perhaps always) violate the cooperative principle. This sends up red flags telling the hearer to look for another meaning, the speaker meaning. Searle labels this step one, of three. In step two, the hearer constructs all possible speaker meanings. In step three, the hearer evaluates these possible meanings to determine which are significant in context. This process can be sketched out very easily, but the second two steps are remarkably difficult to explicate. At this point that is probably unimportant. It will be a monumental task in both psycholinguistics and sociolinguistics to untangle the many factors that come into play. What is important for Searle's view is that the process be rule-governed.

Davidson holds that there are no rules that can do the job. I do not know what views Davidson holds in regards to psychology and philosophy of mind, but it seems that whatever causal mechanisms Davidson resorts to, will be rule based. The rules may differ from person to person, and even from moment to moment, but they must be rules of a sort. I'm of a mind that even the most distributed, parallel processing, connectionist models follow rules. Rules are abstract descriptions. Connectionist models are only instantiations of a mind that still ultimately follows rules, no matter how unwieldy a statement of those rules may be. I am not particularly well versed in this subject, but to me, if metaphors cause us to think and feel things, they must do so in a rule-based way.

Davidson also holds that whatever metaphors cause us to notice (I return to his terminology now because it may have a bearing on this point) is nonpropositional. I do not know what to make of this. What does it mean to notice something that is nonpropositional? I certainly cannot conceive of thinking anything that is nonpropositional. It seems that propositions are simply disembodied thoughts. Perhaps one can feel something nonpropositional. There is a proposition, 'I am experiencing joy,' but I suppose the actual experience of joy does not correspond to a proposition, and metaphors can certainly bring about joy. This does not say anything special about metaphor though. Purely literal speech can cause these nonpropositional feelings, while also clearly carrying propositional content. Take for example the sentence, 'I want to spend the rest of my life with you.' Outside of examples in linguistics and philosophy, I doubt that anyone has ever heard this sentence without experiencing an emotional response, that is 'noticing' something nonpropositional. This does not mean that the sentence lacks propositional meaning.

I agree with Davidson that part of a metaphor's causal power can be to cause one to notice things that are nonpropositional, but I hold that this is a characteristic of all language, and that this fact makes no arguments regarding metaphorical meaning. In fact, it may actually help shore up some weaknesses in ideas that hold that metaphorical meaning does exist. It is often claimed that any propositional paraphrase of a metaphor loses something. If metaphorical meaning is the propositional content of a metaphor, and this is what a paraphrase preserves, then what is lost is the nonpropositional causal power. Surely a paraphrase like 'alpha('I', C) desire to live the remainder of alpha('my life', C) in close companionship with alpha('you', C),' loses the nonpropositional causal powers, but that does not mean that it is an inaccurate paraphrase.

So in my eyes Searle wins this debate. Speakers speaker-mean things when they utter metaphors and hearers compute the propositional speaker meaning through some rule based cognitive system. Unfortunately, I am not done. I disagree with both Davidson and Searle, in that I still think metaphorical meanings exist. I think Searle is on the right track, with his pragmatic efforts, but I think there are flaws that can be avoided by positing actual metaphorical meaning.

Following Davidson's approach a hearer cannot misinterpret a metaphor. It only has a causal effect on the hearer's mind that cannot be right or wrong. Searle's view swings the opposite way. A speaker cannot misuse a metaphor (at least an original one). If I say, 'Sher e Punjab is the heavyweight champion of Indian restaurants in Tucson,' (which it is, on the correct interpretation), but I mean that Sher e Punjab is an awful restaurant, I have misused the metaphor. Under any analysis, I speaker-meant that Sher e Punjab is awful, even if you add the extra clause that to speaker mean something there must be a relationship between what is said and what is speaker-meant. Heavyweight champions tend to smell bad and hurt people. Of course, this is not what the metaphor means. The metaphor means that Sher e Punjab is the best Indian restaurant in Tucson, and I have simply misused the metaphor.

The catch that I feel really gets Searle is that following his steps, any normal English speaker would decide that I had meant that Sher e Punjab is the best Indian restaurant in Tucson. If all a metaphor means is what the speaker speaker-means by it, than any mechanism Searle proposes will fail to find the 'correct' meaning in many cases. Specifically, the mechanism will fail in those cases in which the speaker speaker-means something other than what the mechanisms come up with. Certainly any mechanism that is interesting must come up with a definite meaning, or at least a narrow range of meanings for a given utterance, and certainly a sufficiently incompetent speaker could utter those words, all the while speaker-meaning something entirely different.

I think it is best not to think of this as a necessary failing of every possible metaphor-understanding mechanism, but to think of this as a failing of certain speakers' uses of certain metaphors. I say take Searle's mechanism, elaborate and awkward as it may be, and use it as a device that assigns metaphorical meaning to utterances. I say utterances, rather than sentences, because of course context will be important. I also think this is really an unnecessary distinction because I think ultimately semantic-pragmatics and pragmatic-pragmatics are inseparable. So, in the end, our conclusion seems rather simple. What metaphors mean is what competent speakers, in general, think they mean.