It is no surprise that 20th-century noncognitivism about metaphor began with the Logical Positivists.  Prosecuting their verification theory of meaning, the Positivists disdained figurative language entirely.  Although some metaphorical sentences are empirically verifiable or falsifiable on their literal readings (Bette Midler can be directly observed not to have wings, much less wings with anyone being the wind beneath them, and it is easily checked that many real men do eat quiche), some are not so (“How sweet the moonlight sleeps upon this bank!”<1>).   Much more to the point, most metaphorical sentences on their metaphorical readings are not verifiable in the ordinary empirical way (“But thought’s the slave of life, and life’s time’s fool”<2>); and so they were judged not to be cognitively meaningful.<3>

The Emotive theory
    Among the main philosophical questions posed by metaphor are these:  What is “metaphorical meaning,” broadly construed?  By what mechanism is such meaning conveyed?  How do hearers grasp the meaning, given that what they hear is only a sentence whose literal meaning is something different?
    The Positivists offered a single brutal answer to each of those questions:  There is no such thing as “metaphorical meaning” if by “meaning” one means linguistic or even propositional meaning; there is only emotive or affective significance.  Nor is there any “mechanism” by which metaphorical significance is conveyed, if by “mechanism” one means a cognitive linguistic mechanism.  Nor is there grasp of meaning; there is only the psychological effect that hearing a metaphorical utterance has on one’s feelings and attitudes.  Call this the Emotive theory of metaphor, as in the Emotive theory of ethics.
    The Emotive view is hard to accept.  First, as Monroe Beardsley points out,<4> it implies that the only relevant difference between good metaphors and nonsense strings such as “Blue why procrastination the the when of after dumbwaiter dumbwaiter” is that for whatever reason, metaphors generate affect that word salad does not.<5>  But surely there is a huge cognitive difference between metaphors and word salad:  We often not only understand them but can paraphrase them more literally; we draw inferences from them; we sometimes take ourselves to have learned new empirical facts from having heard metaphorical utterances.  It is hard to deny that there are metaphorical truths, or at least metaphorical assertions widely accepted as truths.
    Beardsley puts the objection a bit more strongly as well.  He identifies two features working in tandem:  Within a metaphorical sentence there is a conceptual “tension” (moonlight differs categorially from an animate being that could sleep, much less sleep “sweetly,” and time is not the sort of thing that could employ a fool or jester); yet the sentence is not only intelligible but perhaps even exceptionally informative or illuminating, and may express an important truth.
    A second objection to the Emotive theory is based on the prevalence of nonliteral usage.  Even if we discount uncontroversially dead metaphor, few human utterances are entirely free of metaphor; virtually every sentence produced by any human being contains importantly metaphorical or other figurative elements.  Though the point has been made many times,<6> it is still worth pausing over, because most philosophers of language try hard to repress the knowledge.
    Everyone grants that among the literal expressions in a natural language are many “dead” metaphors, i.e., phrases that evolved from what were originally novel metaphors but have turned into idioms or clichés and now mean literally what they used to mean metaphorically; they have their own dictionary entries, and in the most extreme cases, none but philologists even know of their metaphorical origins.  However:  The distinction between novel or fresh metaphor and “dead” metaphor is one of smooth degree, not of kind.  Fresh metaphors get picked up and become current, and then only very gradually sicken, harden and die.  And even “dead” metaphors that have their own dictionary entries are often not stone dead; they wear their metaphorical histories on their sleeves, and they still have some rhetorical force in virtue of their original associations: “lame duck”; “rising star”; “sick puppy”; “leap into the breach”; “mind candy”; “throw out the baby with the bathwater”; and for that matter “dead metaphor.”
    It is a further claim that ordinary speech is shot through with expressions that are metaphorical to a degree even if they are coughing up blood.  But that further claim is true; try reading through this page, or any page of any book you may pick up, and underlining each of the terms appearing on it that is in no way metaphorical.
    The upshot for the Emotive theory is that since nearly every utterance contains elements that are to some degree metaphorical, the theory would have us reject every such utterance as cognitively meaningless.  That would leave us very little meaningful communication.
    The Emotive theory primarily opposed accounts of metaphor according to which sentences do have metaphorical meanings, such as the Simile view that a metaphor roughly abbreviates the corresponding simile, and Interaction views according to which mutual juxtaposition within a sentence makes words stretch their ordinary meanings to take on new, analogical ones.<7>  But as we shall see, a number of contemporary theorists agree with the Emotivists that there is no “metaphorical meaning” if by “meaning” one means sentence meaning, yet they give much more plausible accounts of metaphorical communication.  Given the availability of such accounts, there is no reason to accept the Emotive theory, but we need not move so very far away from it either.  The purpose of this paper is to compare two such accounts, those of Donald Davidson and John Searle,<8> and to adjudicate between them.  I shall argue that a fruitful synthesis is available and is considerably more defensible than either Davidson’s or Searle’s view taken on its own.  (Some readers will find my discussion parochial, directed only toward what they consider an in-house or hothouse dispute. I shall address that concern briefly in the last section of this paper.)

Davidson’s Causal theory
    Davidson joins the Emotivist in rejecting “metaphorical meaning” and in denying the existence of linguistic mechanisms by which metaphorical significance is conveyed.  But he is no verificationist.  He grants that metaphorical sentences do have meanings whether or not the sentences are empirically verifiable.<9>  But he contends that the meanings they have are just their literal meanings (however strange those meanings may be).  “[M]etaphors mean what the words, in their most literal interpretation, mean, and nothing more” (p. 30).  When Hotspur said “Thought’s the slave of life, and life’s time’s fool,” he was saying only that thought is (in fact) the slave of life, in those words’ perfectly literal senses, and likewise that life is time’s fool, though probably he was also speaker-meaning more than those things and doubtless he was doing more than just expressing those absurd categorial falsehoods.
    For reasons which I shall try to bring out, Davidson’s view has not been widely accepted, but it has recently been rehabilitated by Marga Reimer,<10> and I agree with her that there is more to be said for it than is generally allowed.
    Davidson’s article is largely devoted to his negative case against “metaphorical meaning”; he gives a number of critical arguments, which we shall consider in the next section.  But he does sketch a positive account of the significance of metaphor.  It is an explicitly causal account, and noncognitive:

A metaphor makes us attend to some likeness, often a novel or surprising likeness, between two or more things.  (p. 31, italics added)

[W]hen ‘mouth’ applied only metaphorically to bottles, the application made the hearer notice a likeness between animal and bottle openings.  (p. 35, italics original)

[A] simile tells us, in part, what a metaphor merely nudges us into noting.  (p. 36, italics added)

This is not to deny that there is such a thing as metaphorical truth, [but] only to deny it of sentences.  Metaphor does lead us to notice what 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.  (p. 39)

    However, Davidson argued skillfully and at length against the idea of metaphorical meaning, so we must survey his arguments.

Davidson’s defense of the Causal theory
    Argument 1: Similes.  (I list this and the next argument first because they are the only ones I find entirely unconvincing and wish to dismiss.  The rest will come in the order in which they appear in Davidson’s article.)  Davidson asks why, if metaphors have to have special cognitive contents in order to achieve what they do, do similes not have such contents?

In general, critics do not suggest that a simile says one thing and means another….  [Yet a simile] may make us think deep thoughts, just as a metaphor does….  (p. 43)

    Argument 2: Murder.  This argument is directed specifically against those theorists, typically Interaction theorists, who hold that the words in a metaphor have taken on analogically or otherwise “extended” meanings.  Citing “a famous critic”’s description of Tolstoy as “a great moralizing infant” and Genesis’ (I, 2) “the Spirit of God moved upon the face of the waters,” Davidson writes:

[I]f in these contexts the words ‘face’ and ‘infant’ apply correctly to the waters and to the adult Tolstoy, then waters really do have faces and Tolstoy literally was an infant, and all sense of metaphor evaporates.  If we are to think of words in metaphors as directly going about their business of applying to what they properly do apply to, there is no difference between metaphor and the introduction of a new term into our vocabulary: to make a metaphor is to murder it.  (p. 32)

    Argument 3: No Manual.

There are no instructions for devising metaphors; there is no manual for determining what a metaphor ‘means’ or ‘says’; there is no test for metaphor that does not call for taste.  (p. 29)

    Argument 4: Nonexplanatoriness.  To posit metaphorical meanings does nothing to explain how metaphorical usage works.

These ideas [e.g., that of metaphorical meaning] don’t explain metaphor, metaphor explains them.  Once we understand a metaphor we can call what we grasp the ‘metaphorical truth’ and (up to a point) say what the ‘metaphorical meaning’ is.  But simply to lodge this meaning in the metaphor is like explaining why a pill puts you to sleep by saying it has a dormative [sic] power.  (p. 31)

    Argument 5: Death.  (This turns Moran’s much later point (4) on its head.)  If there were metaphorical meanings, then presumably they would become literal meanings when the metaphors died.  But the literal meanings of dead metaphors are usually simple and straightforward (e.g., “He was burned up” means just “He was very angry”), while the meanings of live metaphors are normally, and vauntedly, much richer.  (p. 36)

    Argument 6: Unparaphrasability.  Although some metaphors can be paraphrased in literal terms without great loss, many are open-ended in that the relevant set of similarities is vague and indefinite, and some cannot be paraphrased at all.  As an example of the last, Reimer offers W.H. Auden’s line from “Our Bias,” “The hour-glass whispers to the lion’s paw.”<25>  My own favorite example is from e.e. cummings: “he sang his didn’t he danced his did.”<26>  The difficulty or impossibility of paraphrase is neatly explained by Davidson’s claim that there is no metaphorical meaning, for on that view there is nothing to paraphrase (pp. 30, 44-45).  Davidson adds that our uptake of a metaphor, “[w]hat we notice or see,” “is not, in general, propositional in character [at all]....  Seeing as is not seeing that” (p. 45).  Moreover, if a given sentence did have a metaphorical meaning, we would expect that that content could be fairly accurately expressed by some paraphrase, even if the paraphrase were cumbersome, boring, or both.<27>

    I said of Emotivism that there are views that equally reject metaphorical sentence meaning but give more plausible accounts of metaphorical communication.  The same is true in turn of Davidson’s purely Causal theory, and given the availability of such accounts, there is no reason to accept that theory.  But before we turn to the leading alternative, it is necessary to distinguish two strands of Davidson’s attack.
    Rhetorically, the attack is presented as a scorched-earth or zero-tolerance policy.  But actually Davidson concentrates his critical arguments on the idea that in metaphorical usage linguistic expressions change their meanings; what seems primarily to bother him is the positing of linguistic ambiguity.  And as we saw, at one point he is careful “not to deny that there is such a thing as metaphorical truth, only to deny it of sentences” (p. 39).  So far, this leaves open the possibility that there is a middle way or compromise position, that allows for metaphorical meaning in some locus other than that of sentences.
    But elsewhere in the article—and rather prominently—he forestalls the latter idea in so many words:

[Though t]he central error about metaphor is most easily attacked when it takes the form of a theory of metaphorical meaning,…behind that theory, and statable independently, is the thesis that associated with a metaphor is a cognitive content that its author wishes to convey and that the interpreter must grasp if he is to get the message.  (p. 44)

Since in most cases what the metaphor prompts or inspires is not entirely, or even at all, recognition of some truth or fact, the attempt to give literal expression to the content of the metaphor is simply misguided.
    The theorist who tries to explain a metaphor by appealing to a hidden message, like the critic who attempts to state the message, is then fundamentally confused.  No such explanation or statement can be forthcoming because no such message exists.  (p. 45)

The concept of metaphor as primarily a vehicle for conveying ideas, even if unusual ones, seems to me as wrong as the parent idea that a metaphor has a special meaning.  (p. 30)

The Pragmatic theory
    Searle joins Davidson in rejecting metaphorical sentence meaning and the linguistic ambiguity view.  But as against Davidson, Searle’s own account takes seriously the idea that metaphorical utterance is genuinely linguistic communication rather than mere causation, and it posits a cognitive mechanism that computes something well worth calling metaphorical meaning.
    I shall call Searle’s view the Pragmatic theory, for he sees metaphor as simply a species of indirect communication in the style of Gricean implicature and/or his own more broadly Gricean theory of indirect force.
    Searle had previously offered a “conservative” account of how indirect speech acts are performed and understood.<28>  The speaker utters a sentence grammatically marked for one range of illocutionary force but primarily means something by it that has a different force or at least a characteristically different locutionary content.  The hearer proceeds in two stages, first using Gricean reasoning to determine that the speaker is trying to convey something other than what her/his sentence literally means, and then using further Gricean reasoning augmented by principles of speech-act theory and by mutually obvious contextual assumptions to work out the intended force and content of the utterance.
    Turning to metaphor:

The problem of explaining how metaphors work is a special case of the general problem of explaining how speaker-meaning and sentence or word meaning come apart....  Our task in constructing a theory of metaphor is to try to state the principles which relate literal sentence meaning to metaphorical [speaker’s] utterance meaning....  [But i]n our account of metaphorical utterance, we shall need to distinguish it not only from literal utterance, but also from those other forms in which literal utterance is departed from, or exceeded, in some way.  (“Metaphor,” pp. 92-96)

Searle vs. Davidson
    Searle and Weak Davidson are in complete agreement.  Both deny that linguistic expressions have special metaphorical meanings, and both hold that metaphor can be understood using apparatus already on hand in mainstream philosophy of language.  But I do not see why Davidson should, or how he could, dispute Searle’s view that there is metaphorical speaker-meaning.  He does argue, as Reimer emphasizes, that what some metaphors convey is not propositional at all, and if he is right in that, then Searle’s account cannot be the whole story.  But for now, the big disagreement is over the existence of metaphorical meaning tout court, and genuinely cognitive/linguistic mechanisms by which it is conveyed.  Let us see, then, how Searle might rebut Davidson’s Arguments 3-6 against “metaphorical meaning.”

    Ad Argument 3 (No Manual):  As if directly inspired by Davidson’s flat assertion that there are no instructions or rules for generating or for interpreting metaphors, Searle produced quite a number of such rules, and so far as they go they are plausible.  Davidson added the qualification, “no test for metaphor that does not call for taste”; very likely Searle would concede that point, since he makes no claim to completeness and does not predict that even a final set of principles will give perfectly determinate results.  Nonetheless, there are instructions and rules.<30>

    Ad Argument 4 (Nonexplanatoriness):  Davidson’s unfavorable comparison of metaphorical meanings to literal meanings was to literal sentence meanings, and the objection was obviously directed against special metaphorical sentence meanings.  Whether or not the Pragmatic theory is otherwise adequate and whether or not its explanations are correct, Searle has shown that the notion of metaphorical speaker-meaning does figure substantively in explanations of how metaphorical usage works; it works by the hearer’s Gricean computation of metaphorical speaker-meaning (based on initial perception of literal sentence meaning).  So the objection has no force against Searle.

    Ad Argument 5 (Death):  Searle says little about dead metaphors, but his account at least suggests a way of dealing with Davidson’s problem.  While a metaphor is still alive, it may be comparatively open-ended; the third step of Searle’s interpretation procedure will not have eliminated all but one or two of the possible speaker-meanings.  But perhaps part of the dying process involves a constriction of just this sort.  For whatever social reason, one of the speaker-meanings hardens and squeezes out the others, and that meaning becomes conventional rather than something that needs to be calculated in Searle’s way.  Of course, this idea concedes something to Davidson, viz., that prior to rigor mortis, there was more to the metaphor than just the corpse that is the new literal meaning.  To accommodate that, Searle would have to say something about how there can be more to an open-ended speaker-meaning than to a definite one—for one would expect there to be less, e.g., if one thinks of an open-ended speaker-meaning as the disjunction of possible definite meanings.  I shall return to this problem of open-endedness in the next section.

    Ad Argument 6 (Unparaphrasability):  Davidson’s remaining appeal was to unparaphrasability and downright nonpropositionality.  Searle’s account is unfriendly here.  He grants that often we use metaphor precisely because there is no handy and accessible literal expression that means the same thing, but he argues that if something is a linguistic meaning at all, in principle it could be formulated (however cumbersomely) in some language or other.
    Since I sympathize with the latter principle, I think Searle wins this round as well.  But as Reimer insists, there is a deeper issue about nonpropositionalness.  Searle’s account is propositional to the core, since all speaker-meaning is meaning that so-and-so.  If Davidson is right that what we notice or see in metaphor “is not, in general, propositional in character,” then by Searle’s own principle aforementioned, it is not a linguistic meaning of any kind, not even a speaker-meaning.
    Though a qualification, Davidson’s interpolated phrase “in general” (“not, in general, propositional in character” (p. 263)) makes his claim fairly ambitious, indeed false.  Perhaps many poetic and other literary metaphors are so rich as to be nonpropositional in their purport, but everyday metaphors used casually by ordinary people are often perfectly paraphrasable in context.  Quite often the speaker certainly does mean something, possibly something quite specific and unambiguous.  The imprecation “You pig!” can mean different things in different circumstances, but in each type of circumstance it is perfectly paraphrasable: the hearer is grossly fat; the hearer is a filthy slob, the hearer is a glutton, the hearer is shiftless and indolent, the hearer is stubborn and unreasonable (“pigheaded”).<31>  So I believe Davidson has overstated his case by overlooking plain facts of speaker-meaning.
    On the other hand, just as Davidson says, writers who strew fresh literary metaphors, far from always having determinate speaker-meanings, may have no speaker-meanings or other propositional intent at all.  That does not make the metaphors any less good or useful, because metaphor does sometimes have the quasi-perceptual character noted by Davidson.  In some cases metaphor affects one’s literally perceptual set.  (In other, intermediate cases, the metaphor just puts one in a different intellectual frame of mind for thinking about the topic at hand.)  And that is a telling point against Searle.
    For that matter, the Emotivists were right to allude to affect, even if they were wrong to insist that metaphor is entirely noncognitive.  For many metaphors, a good deal of their force and effect is affective rather than cognitive.  Of course there is no reason for Davidson not to grant that.  Searle can grant it too, though it does not sit as well with his militantly cognitive view as it does with Davidson’s already purely causal theory.

The rapprochement
    Now, here is the synthesis I mentioned in beginning.  Each of our two theorists is right about something important:  Searle is right in that there certainly is metaphorical speaker-meaning, and his view of how that meaning is discerned is plausible so far as it goes.  Davidson is right in that a metaphor’s accomplishments often transcend the propositional.  I contend that the Causal and Pragmatic theories can be combined into a single and more comprehensive view that will respect both insights.
    Notice that cases of metaphor lie on a certain scale.  At one end of the scale, metaphorical utterances have determinate, clear and obvious speaker-meanings even though they remain metaphorical.  But some metaphors are a little more open-ended; we may be sure of one or two properties that the speaker is ascribing, but not sure of the others even though the metaphor is plainly richer in content than just the one or two.  Further along the scale, metaphors are more open-ended still, until they become ineffable in that although the speaker evidently means something, there is no paraphrasing it.  Finally, some metaphors just go indeterminate, and we feel there is no propositional speaker-meaning at all.
    Clearly the original end of the foregoing scale is the Searle end and the latter end is the Davidson end.  The earlier on the scale a metaphor occurs, the better Searle’s account will apply to it.  The later it occurs, the more force Davidson’s Arguments 3 and 6 will have against Searle and the more we will be inclined to fall back on the Causal theory.  So perhaps we should begin by acknowledging that metaphors vary along the dimension I have described, and then suppose that Searle’s view correctly characterizes metaphors lying at the early end of the scale and for some distance along it; but at some point or points, Searle’s rational reconstruction runs out, brute causality takes over, and Davidsonian muteness is the appropriate response.  (There would be the further consequences, noted in our original critique of the Causal theory, that such Davidson-end metaphors cannot be misinterpreted and that they are not themselves true or false.)
    Brute causality may also play some role fairly early along the scale.  For in addition to a particular speaker-meaning conveyed and interpreted by Searle’s means, there may be a penumbra of Davidsonian quasi-perceptual noticings and shifts of mental set that elude Searle’s explanatory apparatus and can be explained only in some noncognitive way.  Consistently with that, they may contribute to the apparatus by helping the hearer to tamp down possible speaker-meanings.  (I have not yet addressed the problem of open-endedness mentioned in the last section, but will defer that for a few paragraphs.)
    We might brilliantly call this irenic combined view the “Pragmatic-Causal theory.”

    The combined theory avoids each of three objections that have been made against Searle.  First, Moran (1997, p. 263) has complained that Searle has failed “to elucidate the specifically figurative dimension of metaphor”; even taking into account that on Searle’s view a speaker may intend an indefinite range of meanings, “[n]o degree of indefiniteness alone will add up to power or insightfulness.”
    The latter statement is not obvious.  As has been widely observed, part of the function of fresh metaphor is to make the hearer work at interpreting it, which requires exercising the imagination.  (This is no longer true once a metaphor dies; that is a second reason why as Davidson says, the dead metaphor is impoverished.)  Surely some of the metaphor’s power and insightfulness is explained by the participatory and imaginative nature of the hearer’s interpretive process.  But I agree with the thrust of the criticism.  What has the Pragmatic-Causal theory to add?
    Open-endedness does contribute here, though in a more active way than Searle envisages.  I now hypothesize on behalf of the Pragmatic-Causal theory that open-endedness occupies its own explanatory niche.  We saw that there can be more, not less, to an “open-ended speaker-meaning” than to a definite one, and that suggests that the phrase is a misnomer.  What is really going on is that speaker-meaning has been outrun again; Searle’s apparatus cannot really explain open-endedness even if it is compatible with it.  But it does not follow that the open-endedness is a matter of Davidsonian aspect-perception and/or affect, for there is an intermediate, still somewhat cognitive category.  As has been pointed out by Simon Blackburn and no doubt others,<32> an open-ended metaphor stands as an invitation to explore the proffered comparison.

Thus when Romeo says Juliet is the Sun we can profit from the metphor indefinitely: we can move among the respects in which someone’s lover is like the Sun: warm, sustaining, comforting, perhaps awesome, something on which we are utterly dependent….  This process is quite open-ended….  The metaphor is in effect an invitation to explore comparisons.  (p. 174)

    On to the second objection to Searle that is avoided by the Pragmatic-Causal theory:  Moran (op. cit.) and D.E. Cooper<36> note that if metaphorical meaning is simply speaker-meaning, then it is determined by and confined to the speaker’s intentions.  Yet in cases of fresh metaphor, as Cooper says (p. 73), “even a quite definite speaker-intention does not finally determine the meaning of a metaphor.”  Moran adds that “the interpretation of the light [the metaphor] sheds on its subject may outrun anything the speaker is thought explicitly to have had in mind” (p. 264).
    A first reply to this is to balk at “explicitly,” and point out that not all of a speaker’s intentions are ones that the speaker did explicitly have in mind.  Often we come to realize that we spoke or acted with a certain intention even though we had been largely unaware of at the time.  But even if we are wary of this appeal to shadowy tacit intentions, again the Pragmatic-Causal theory can help in the same way as before:  Although a Davidsonian penumbra cannot determine the meaning of a metaphor in the strictly propositional sense, it can contribute to the overall effect of the metaphor, and in a way that is entirely independent of the speaker’s intentions.  Indeed, it seems obvious that this often does happen.

    And the third objection:  As Searle admits (pp. 116-17) and as is emphasized critically by Stern (op. cit.), the comparisons that underlie metaphor are often themselves metaphorical and when pursued, they sometimes bottom out in brute “fact[s] about our sensibility.”  E.g., emotionality is compared to temperature, and gentle, kind or pleasant personality traits are compared to degrees of sweetness.  “[W]e just do perceive a connection… [the properties are] associated in our minds….”  But is this not a huge Davidsonian explanatory gap deep in Searle country?  The principles of conversation say nothing about brute psychological associations.
    First, notice that on Searle’s model the Gricean principles do not themselves have to advert to the associations even when Searle’s method as a whole recruits them.  The first step in his strategy is what gets the hearer to the lemma that the speaker’s utterance was metaphorical.  Only then does the method instruct the hearer to start looking for comparisons; the purely Gricean part is over.  And, second, there is nothing wrong with tacit appeal to brute associations when it is mutually known by hearer and speaker that people do habitually make those associations.
    In any case, third, the Pragmatic-Causal theory is only partly pragmatic, and does regularly appeal to the Davidsonian penumbra of built-in associations and aspect-perception.  So even if Searle’s own project were vitiated by the need to make such an appeal, the Pragmatic-Causal theory takes that need in stride.<37>

New objections to the Pragmatic-Causal theory
    Practitioners of other sorts of theories of metaphor may feel that there is no significant difference between Davidson’s and Searle’s accounts in any case.  In particular, the Pragmatic theory itself faces several further criticisms that would also be objections to the Pragmatic-Causal view.  I shall close this paper by briefly reviewing several of these, though with no hope of allaying all the concerns behind them.  My main purpose is just to speculate as to how far the Pragmatic-Causal view may be defended.

    Objection 1: Anti-Grice.  After a quarter century’s reign, Grice’s theory of conversational implicature and the Gricean “conservative” approach to indirect force have seriously been called into question.  “Relevance theory,” though it began as a development of and/or friendly amendment to Grice’s own apparatus of conversational “maxims,” has become a vigorous competitor.<38>  More recently and more pertinently, Wayne Davis has argued in detail that Grice’s apparatus is too vague and feeble actually to generate the implicatures and (especially) the indirect speech acts that have ostensibly been explained by it.<39>  (It is easy for a hearer to begin, “[The speaker] couldn’t mean that, because it is too obviously false and we all know that”; we know that something is up.  But then there is the positive part of figuring out just what it is that is up, and here is where Grice’s maxims fail.<40>)  So too, Davis says, with Searle on metaphor: Searle makes his second and third steps look far easier than they would be in real life.  Davis contends further that in reality, conventional elements enter into the positive stage, especially in the interpretation of indirect speech acts.<41>
    I am afraid that Davis is right about all that.  But notice that (as he intends) his critique of Grice applies to all cases of implicature and of indirect force.  Now, no one denies the existence of either phenomenon itself, so the failure of Gricean explanations of implicature and indirect force signifies only that we need a better theory of those things; Davis’ problem is everyone’s problem.  There is no present a priori reason to doubt that an eventual adequate account of implicature and indirect force will extend to cover metaphor as well.  Moreover, Davis’ appeal to a weak sort of conventionality may help to mark senescent metaphors’ nearness to death.

    Objection 2: Embedding.  Recall Moran’s Geach-style objection to Davidson: metaphorical clauses embed, e.g. in conditional antecedents, and what is thus impacted is not just their literal contents but the metaphorical content that would have been intended by their utterers.<42>  The problem for Davidson was that he could admit no content for such a conditional antecedent to express.  The problem for Searle is that Gricean reasoning always starts with the literal content of the speaker’s whole utterance; a hearer cannot read an implicature out from under sentential embedding.  Moreover, the condition intuitively expressed by the metaphorical antecedent is not given by the literal meaning of the embedded clause, but is the apparent metaphorical content.
    Two replies may be made to this.  First, it is already known that there are what Stephen Levinson calls “intrusive constructions,” operators which yield compound sentences whose truth-conditions depends on the implicatures rather than the truth-conditions of the operands:<43>  “Driving home and drinking several beers is better than drinking several beers and driving home”; “If each side in the soccer game got three goals, then the game was a draw”; “She either got married and had a child, or had a child and got married; I don’t know which”; “Because the police have recovered some of the gold, they will no doubt recover the lot.”  Carston (op. cit.) argues that Relevance theory can handle such data even though traditional Gricetheorie cannot.
    Second, White (op. cit., pp. 187ff.) points out that the Geach-style argument makes a substantive assumption: It assumes that the antecedent clause in the relevant conditional has the same content, whatever that was, as would have been intended by an utterer of the original.  Though entirely natural, that assumption is open to question.  Since White himself argues emphatically that the locus of metaphor is always the entire sentence and not any word, phrase or clause contained in it, he finds it easy to deny that metaphorical antecedents are “detachable”: “if we were to consider either the antecedent or consequent of such an hypothetical detached from the hypothetical,…it would not, in general…[have] the same reading as is required to make sense of the whole hypothetical utterance” (p. 189).

    Objection 3: Analogy Mechanisms.  Interactionists such as Ross (op. cit.) and Kittay (op. cit.) call our attention to a class of lexical phenomena, sometimes called “analogical,” that indisputably involve meaning and meaning shift but are addressed neither by Davidson’s view nor by Searle’s.  They are pervasive; they occur in nearly every sentence that comes out of our mouths.  And there is a powerful if sketchy theory of the analogy mechanisms that generate the new meanings when two terms are juxtaposed that have not previously co-occurred.  Thus, there are metaphorical sentence meanings whether we like it or not, and the Pragmatic-Causal theory ignores them.
    The premises are true.  It has been indisputable since Aristotle that words take on paronymous meanings and that this happens by way of various analogical relations.  Moreover, there do seem to be interactive analogy mechanisms that function on their own, independently of speakers’ intentions or hearers’ interpretive strategies.  But not all generated analogical meanings are metaphorical meanings, and it is not clear that any are.  Indeed, in both Ross’ and Kittay’s accounts of metaphor, metaphorical meanings are the result of a sort of second-order operation on analogical meanings.  “A [metaphorical] transference of meaning is not a simple displacement of an atomistic meaning but a move from one system to another….  [M]etaphorical meaning is a second-order meaning…” (Kittay, pp. 138, 141).  Kittay goes on to propose an account of metaphor as a second-order phenomenon, based on semantic field theory.  So from the pervasiveness of analogy-generated paronymy it does not follow that any metaphorical sentence meanings are produced in this way, and it remains an open question whether any second-order theory of metaphor based on an analogical theory of first-order meanings is superior to the Pragmatic-Causal theory.

    Objection 4: Metaphorical Thought.  I.A. Richards pointed out,<44> and Lakoff and Turner (op. cit.) emphasize at length, that metaphor is not essentially a linguistic phenomenon, for we can think in metaphor, indeed very richly, without speaking or hearing speech.  Lakoff and Turner go farther, and contend that metaphor is essentially a feature of thought, and only accidentally and derivatively linguistic.  Yet the Pragmatic-Causal theory treats the entire issue of metaphor as a problem about the interpretation of speech, and does not apply in any obvious way to silent thought.
    It is indeed obvious that we often think in metaphor.  It is far less obvious that this is essential to metaphor rather than itself derivative.  Do languageless creatures ever think in metaphor?<45>  I know of no evidence that they do, though it is an empirical question.  But this does not answer the objection, for even if metaphorical thinking is only (“only”) internalized speech, it does happen, and the Pragmatic-Causal view as developed so far gives no account of it.
    What may help is to point out that analogues of implicature and indirect force occur in silent thought as well.  Sarcasm and irony certainly do.  And we certainly find ourselves thinking such things as “I must get downstairs,” meaning that I need to find a toilet, or (while driving) “Would this person mind getting in one lane or the other?”  It may be said that examples of this kind are examples, not of thoughts themselves, but of verbal imagery; we are imagining the words in which we might only indirectly express the actual thoughts we are having.  But that is hardly obvious, and the matter of nonverbal implicature and indirect force needs a good deal more investigation.

    Objection 5: Degrees.  Can the Pragmatic-Causal theory accommodate the nasty prevalence of nonliteral usage, in particular the nonexistence of a difference save one of smooth degree between real metaphors and “dead” metaphors?  Davidson’s theory by itself does so, I believe:  Nearly every natural-language sentence has some “penumbra,” however insignificant, of associations and aspect-perception in addition to its literal content.  But Searle’s view seems to introduce a sharper distinction, since either speaker-meaning diverges from sentence meaning or it does not.  Moreover, criticizing what they call “the Pragmatics Position” (p. 125), Lakoff and Turner argue that a view like Searle’s is committed to each of two unacceptable claims: the “Deviance Position,” according to which “all concepts and conventional language are nonmetaphoric, and we make metaphors only by deviating from normal conventional usage” (p. 124), and the “Fallback Position,” the idea that since normal language use is nonmetaphorical, “we look first for the literal meaning of a sentence…, and seek a metaphorical meaning (that is, a paraphrase, only as a fallback, if we are not content with the primary literal meaning” (p. 125).  This is, indeed, Searle to the life.
    In response, let us remember a fundamental fact that may be obscured by Lakovian rhetoric (though it is not forgotten by Lakoff and Turner themselves): On anyone’s theory, metaphor is derivative, and presupposes a prior meaning.  For that reason, on pain of regress, there must have been entirely nonmetaphorical utterances even if there are no longer any.  Moreover, there are former metaphors that are truly dead, in that no one but the odd philologist knows that their current senses began life as metaphorical.
    Now consider a present-day utterance.  Most likely it will not be entirely literal, even if we properly ignore the truly dead former metaphors.  But, n.b., so far as the utterance is metaphorical and “live” to any degree, it has a literal meaning that is available to hearers.  (If the hearers could hear no literal meaning behind the metaphor, it would not be metaphor for them, but truly dead.)  Lakoff and Turner write as if “literal meaning” in normal language use is a myth—and one sees their point, if the opposing idea was supposed to be that all normal language use is purely literal—but again, when normal language use is to any degree metaphorical, a prior literal meaning is there and available.
    From Searle’s viewpoint, what this means is that the computation process is more tedious than at first we imagined.  For the hearer must compute the little, nearly dead speaker-meanings from the (barely heard) literal sentence meaning, and then recursively calculate the fresher metaphorical meanings from those.  This complicates not only the recovery process, but also Searle’s idea of speaker-meaning itself, for there will now be remoter speaker-meanings impacted within the first one that diverges from the literal.
    However, the complication will not be as awful as might be feared.  For the little, “dead” or moribund metaphors are in standard usage, indeed may be clichés (as are the examples I originally gave of “dead” metaphors—“lame duck,” “rising star” et al.).  Well-entrenched custom will help the hearer in Searle’s third stage of processing to tamp down the competition and discern the speaker’s actual meaning from among the possible ones displayed as the output of stage two.  And no doubt the Davidsonian penumbra will help where there is remaining underdetermination.  (In issuing this blithe vote of confidence, remember, I am speaking only of the hearer’s getting from the utterance’s barely-heard Ur-literal meaning to its everyday quasi-literal meaning that includes the Lakovian “dead” and moribund metaphorical elements; getting from the latter meaning to further, fully metaphorical speaker-meaning will be as hard as it would otherwise be.)

Conclusion
    The Pragmatic-Causal theory is more defensible than is either Davidson’s theory alone or Searle’s theory alone.  And, I have argued further, it is tenable in the face of the most obvious objections aimed directly at it.
    For the record, I suspect that “metaphor” does not constitute a single natural kind.  (For example, some theorists, such as Fogelin (op. it.) and White (op. cit.) see a more substantive distinction than do Lakoff et al. and I between “dead” metaphor and real metaphor; some would distinguish personification from metaphor; some would write off cummins’ “he sang his didn’t he danced his did” as nonsense rather than metaphor; etc.)  And so it is possible that there will be no single, unified theory of metaphor.  It may be harder to get the taxonomy right than to give a decent theory of any one of the taxa.
 
 

1  The Merchant of Venice, V, i, 54.

2  Henry IV, Part I, V, iv, 81.

3  For now I shall continue to speak sloppily of sentences’ being metaphorical or not, but this usage will be refined below.

4  “Metaphor,” in P. Edwards (ed) The Encyclopedia of Philosophy, Vol. 5 (New York: Macmillan, 1967).

5  There is the fact that most metaphors are at least grammatical sentences, but the Positivists were rarely impressed by superficial grammaticality alone.  Also, not all metaphors are grammatical sentences.

6  Particularly by George Lakoff and Mark Johnson, in Metaphors We Live By (Chicago: University of Chicago Press, 1980).  See also George Lakoff and Mark Turner, More Than Cool Reason (Chicago: University of Chicago Press, 1989).

7  I know of no one who currently accepts the simple Simile view, but a sophisticated and illuminating Simile theory is defended by Robert Fogelin in Figuratively Speaking (New Haven, CT: Yale University Press, 1988).  Recent Interaction theories include those of J. Ross (Portraying Analogy, New York: Cambridge University Press, 1981) and Eva Feder Kittay (Metaphor, Oxford: Clarendon Press, 1987); 20th-century Interactionism goes back to Max Black’s “Metaphor,” in Models and Metaphors (Ithaca: Cornell University Press, 1962).  A further, different kind of theory that features metaphorical sentence meaning is the quasi-indexical view defended by Josef Stern in Metaphor in Context (Cambridge, MA: MIT Press, 2000).

8  Davidson, “What Metaphors Mean,” in S. Sacks (ed.) On Metaphor (Chicago: University of Chicago Press,1978); Searle, “Metaphor,” in A. Ortony (ed.) Metaphor and Thought (Cambridge: Cambridge University Press, 1979).

9  He distances himself from the general Positivist attitude as well:  “Metaphor is a legitimate device not only in literature but in science, philosophy, and the law: it is effective in praise and abuse, prayer and promotion, description and prescription” (p. 31).

10  “Davidson on Metaphor,” Midwest Studies in Philosophy XXV (2001), 142-55.

11  “Metaphor,” in C. Wright and R. Hale (eds.) A Companion to the Philosophy of Language (Oxford: Basil Blackwell, 1997).

12  Twelfth Night, I, i, 1.

13  This point was once made to me by Franklin Goldsmith.

14  Devotions upon Emergent Occasions, No. 17.

15  (He had been less explicitly anticipated in this by Simon and Garfunkel, in their 1966 song “I Am A Rock.”)  Will later generalizes: “In my opinion, all men are islands. And what’s more, now’s the time to be one. This is an island age.”  (At one point he also switches metaphors, adopting a television-updated version of Shakespeare’s standard “stage” trope: “I was the star of The Will Show. And The Will Show wasn’t an ensemble drama. Guests came and went, but I was the regular. It came down to me and me alone.”

16  Reimer (p. 152) defends Davidson against this objection by insisting that the proposition over which Donne and Arnold disagree “needn’t be a proposition expressed by the metaphor itself…[or] even be a proposition meant by the author of the metaphor….  Arnold may well have succeeded in conveying (to his audience) that he himself believed that we are alienated from one another.  But…it would be a mistake to take this as implying that the metaphor itself—or even its author—‘means’ that this is so.”  Reimer is clearly right to point out the failure of the latter implication, but she does nothing to show that Will and Arnold do not mean to deny what Donne asserted (or what he at least meant), and so far as I can see, they did mean that we are alienated from each other.

17  The Structure of Metaphor (Oxford: Basil Blackwell, 1996), p.199.

18  King Lear, I, i, 21.

19  Richard III, IV, iv, 365.

20  Nelson Goodman, “Twisted Tales; or Story, Study, and Symphony,” Synthese 46 (1981), 331-350.

21  King Lear, I, iii, 19.

22  Reimer (op. cit., p. 148) considers this reply, but is not convinced.  She rejoins in good Davidsonian fashion by “deny[ing] that there must be something that a speaker who uses a simile means. That is, there needn’t be a proposition, even the one literally expressed, that the speaker intends to communicate” (italics original).  Well, true, there need not be, but that holds of any sentence whatever, figurative or not.  Any sentence can be tokened without the utterer’s meaning anything by it at all, as in delirium or when testing a microphone or practicing elocution.  The question is, what a normal utterer of a given sentence in an everyday context would most probably mean by it, and it seems clear to me that the normal utterer of a simile would mean at least the relevant resemblance claim.
    Reimer anticipates a second possible reply to the Simile argument: that similes do have special cognitive contents just as metaphors do, in that the point of uttering a simile is never simply to make the bare resemblance claim.  Reimer rejoins (p. 149) that this is a non sequitur; that there is a further point to uttering the simile hardly entails that that point is for the speaker to express some special cognitive content.  Further argument would be required, especially in light of the now familiar point that the most interesting similes are themselves figurative (see, e.g., Fogelin, op. cit.).

23  For an elaborate account of some such mechanisms, see Ross, op. cit.

24  In this Davidson follows Ted Cohen (“Figurative Speech and Figurative Acts,” Journal of Philosophy 72 (1975): 671-84).  “[M]etaphorical meaning is somehow constructed out of literal meaning, but not according to any function.  In this respect metaphor differs from other figures.  Irony, for instance…” (p. 672).

25  Line 1 of “Our Bias,” in (e.g.) The Collected Poetry of W.H. Auden (New York: Random House, 1945).  Reimer adds the qualification that the inscrutability of this may be due to “the lion’s paw”’s being an allusion, to a line of Shakespeare’s.  If such allusion was intended, I think the reference must be to the opening of Sonnet 19 (Auden’s poem is itself a sonnet): “Devouring time, blunt thou the lion’s paws.”  (Context supplies an “even if” or “even though,” so that the line means roughly, “Time, though you may ravage even the fiercest beast,….”)  Plugging “the fiercest beast” back into Auden’s line, it is still not clear what that line would mean.  The theme of the poem is, I conjecture(!), human beings’ freedom from the present moment, compared to ways in which lower animals are stuck in their present.  If so, Auden’s poem may have been meant as a partial corrective to Shakespeare’s.  On that reading, the word “whispers” would receive the emphasis.  (But there is still the question of why Auden would have changed Shakespeare’s plural “paws” to the singular, unless to make it near-rhyme with “for” at line 3.)
    Actually the matter is considerably more complicated.  For the line quoted by Reimer is not the final or authorized version, even though it did appear in print more than once (and, according to Reimer’s own reference, was even anthologized by Norton).  The final version, which also appeared in print more than once and was then codified and authorized in Collected Poems, ed. E. Mendelson (London: Faber and Faber, 1976), is, “The hour-glass whispers to the lion’s roar.”  Being no Auden scholar, I have no idea when or why the change was made, though clearly it makes a better rhyme with “for.”  (In his Foreword to Collected Shorter Poems (New York: Random House, 1966), Auden says, “[I]t makes me wince when I see how ready I was to treat –or and –aw as homophones,” though he does not mention “Our Bias.”)  But this final version of the line could not be an allusion to Sonnet 19, save a cryptic one confined to Auden’s mind.  And it restores full inscrutability, though it is consistent with the theme aforementioned.
    (There are two other changes:  Lines 7-8, originally “Has never put the lion off his leap / Nor shaken the assurance of the rose” became “Has never put one lion off his leap / Nor shaken the assurance of a rose.”)

26  Line 4 of “anyone lived in a pretty how town,” in Complete Poems 1913-1962 (New York: Harcourt, Brace, Jovanovich, 1972).

27  Some readers, notably Kittay (op. cit., pp. 97ff.), have attributed an additional argument to Davidson, based on the thesis that literal sentence meaning is independent of context.  But Davidson does not hold any thesis so general as that.  What he does claim, in Argument 4, is only that sentences have their literal meanings independently of the uses to which they may be put.  He gives the example of lying; no one would suggest that when a sentence is uttered as a lie, it takes on a new “deceit meaning.”  Nor would anyone suppose that when the sentence is shouted from a mountainside to test the echo, it takes on a special “acoustic meaning.”

28  “Indirect Speech Acts,” in P. Cole and J.L. Morgan (eds.), Syntax and Semantics, Vol. 3: Speech Acts (New York: Academic Press, 1975).  I characterized Searle’s approach as “conservative,” and discussed it at length, in Ch. 7 of Logical Form in Natural Language (Cambridge, MA: Bradford Books/MIT Press, 1984).

29  The Gricean strategy is not the only first-step option.  Some metaphorical utterances are not in any way defective; there are other contextual cues, such as the kind of discourse that is taking place.  Searle observes that “when reading Romantic poets, we are on the lookout for metaphors” (p.114).  And as Kittay (1987, p. 76) notes, metaphors can be explicitly flagged as such (“metaphorically speaking”).

30  Stern (op. cit.) reminds us that Davidson has always been skeptical about the possibility of codifying “conversational implicature” and Gricean reasoning generally.

31  My goodness, what a comprehensive indictment of pigs.  In each case, I would argue, the metaphor is one of those that exploits an inaccurate popular stereotype.  But there are some subtleties too: Searle reminds us (p. 116) of the differences between “Sam is a pig,” “Sam is a hog,” and “Sam is a swine.”

32  Blackburn, Spreading the Word (Oxford: Oxford University Press, 1984), pp. 171ff.

33  Titus Andronicus, I, i, 314.

34  Richard III, V, iii, 194.

35  Romeo and Juliet, III, iii, 54.  (Unfortunately a mixed metaphor, since in the immediately preceding line Friar Laurence has called his philosophy “armour.”  In any case, Romeo responds, “Hang up philosophy! / Unless philosophy can make a Juliet.…”  Well.)

36  Metaphor (Oxford: Basil Blackwell, 1986).

37  Incidentally, the theory of metaphor known to me that is closest to mine is that of Roger White (op. cit.), though his is a good deal more subtle.  White too (a) rejects metaphorical sentence meaning and (b) defends propositional speaker-meaning but (c) insists that the interesting and creative achievement of a good metaphor is nonpropositional.  He also argues, correctly in my view, that the locus of metaphor is whole sentences, not words or even phrases within them.  And he has further interesting and detailed things to say about how authors exploit multiple ambiguities in developing a metaphor or set of them over an extended stretch of discourse.  Highly recommended.

38  Initially, D. Sperber and D. Wilson, Relevance: Communication and Cognition (Cambridge, MA: Harvard University Press, 1986); for a massively helpful presentation of the current Relevance critique of Grice (and much else of value), see Robyn Carston’s Thoughts and Utterances: The Pragmatics of Explicit Communication (Oxford: Basil Blackwell, 2002).  For an alternative critique of Grice, C. Gauker, “Situated Inference versus Conversational Implicature,” Noûs 35 (2001): 163-89.

39  Implicature (Cambridge: Cambridge University Press, 1998).

40  Davis points out that philosophers of language have missed this important lacuna in Grice’s theory because, whenever we look at an example, we already know what would normally be implicated by an utterance of the sentence in question, and so we take it that there is a reasonable route to that implicatum, and are not moved to ask ourselves how, exactly, the positive calculation is worked out.

41  Searle had himself admitted that (“Indirect Speech Acts,” loc. cit, pp. 75-78).  For an early and strong argument for the conventional element in indirect force, see Jerry L. Morgan, “Two Types of Convention in Indirect Speech Acts,” in P. Cole (ed), Syntax and Semantics, Vol. 9: Pragmatics (New York: Academic Press, 1978), for extended discussion, see Ch. 7 of my Logical Form in Natural Language, loc. cit.

42  Jonathan Cohen presses a similar but not quite so well focused objection against Searle, in “The Semantics of Metaphor,” in Ortony, op. cit., pp. 65-66.

43  Levinson, Presumptive Meanings: The Theory of Generalized Conversational Implicature (Cambridge, MA: MIT Press).  Such examples had previously been noted in L.J. Cohen, “Some Remarks on Grice’s Views about the Logical Particles of Natural Language,” in Y. Bar-Hillel (ed.), Pragmatics of Natural Language (Dordrecht: D. Reidel, 1971), and D. Wilson, Presupposition and Non-Truth-Conditional Semantics (New York: Academic Press, 1975).

44  The Philosophy of Rhetoric (London: Oxford University Press, 1936), p. 94.

45  Assuming, as I do contra Davidson and the early Sellars, that languageless creatures think at all.