I begin with what I am sure about. Metaphors are, indeed, for practical purposes, elliptical similes: though not all metaphors can be given the surface form of a simile, they all involve comparisons. Metaphors, of course, are not reversible, but neither are similes. Our understanding of a metaphor only begins when we realise that the literal sense is not the point (or the only point) of the utterance. The same is true of similes. In each case we have to make what we can of the utterance. How we do this is the same for both figures. What we make is the same in each case, and subject (even if not subjected) to the same controversies. Both, as Robert Fogelin has put it, express figurative comparisons: similes explicitly, metaphors implicitly.

How do we unravel metaphors? If rumour is claimed to be (like) a disease then we compare the two. However, we proceed by looking at disease and seeing what salient features might be attachable to rumour. This order is crucial. We are interested in what features are believed to be attached to disease whether or not they are. Context might determine attachability, particularly in poetry. We also need to move from talk of objects to talk of words; not because rumour and disease are funny objects, but because we work with the words. As we go about (and about) construing the metaphor, everything that can be done in the object mode can be done in the verbal mode, but things can be done in the verbal mode that cannot be done in the object mode. Words refer so we retain our objects, odd or otherwise, but words have their own features, particularly their allusive power. The Church of England as the Tory Party at prayer would be needed to underpin our responses to the claim that Marxism-Leninism was the Communist Party of Australia at prayer.

But what do we emerge with from this exercise in comparison? What does a metaphor give us? Do we emerge with a metaphorical meaning, or two, or more? What kind of meaning would a metaphorical meaning be? If we come out with no more meaning than the literal meaning we went in with, do we come out with a new way of seeing, or experiencing? That metaphors are comparisons does not give us answers to such questions. Someone says «Rumour is a disease». Salient features of disease are contagion and damage and these features seem «attachable» to rumour. So has it been claimed that rumour is a disease and, more particularly, a social disease and harmful (even fatal, potentially)? Or have we been invited to try thinking of rumour as a disease; or to have the experience of contemplating rumour as contagious and harmful; or to entertain that thought; or, to have that thought entertain us? Knowing how we manage to respond appropriately to this metaphor, insofar as we do know this, does not help us to decide between these options.

The fundamental disagreement about metaphor is between those who think a metaphorical utterance has a meaning other than, or in addition to, its literal meaning, and those who do not (most famously Donald Davidson). But there are important, though lesser disagreements. Are metaphors a case of speaker's meaning? Or do the words used metaphorically have, on that metaphorical occasion, a meaning other than their normal, literal sense? Does such a meaning attach to the metaphorical utterance as a whole or to a word or words in that sequence of words? Is the literal meaning retained or discarded?

I have not talked of special metaphoric meaning. That seems to offer no addition to the embarrassment of choice already on offer. However, elucidation of the relevant concepts is needed to make sure the options are clear.

`Inspissated' means `complexly dense' (near enough) and started life as a term of haute cuisine, so that a traditional French sauce is inspissated. Naturally enough it expanded its reach into the haute cuisine of the intellectual kitchen, so that «This book is inspissated» means that the book in question is complexly dense. What if, unsurprisingly, some tyro, hearing `inspissated' used, thinks it means `unduly complicated'. They say: «This book is inspissated». They mean, by saying this, that it is unduly complicated. Here is a case of speaker's meaning diverging from what is normally (standardly) meant. Of course, speaker's meaning need not diverge.

What does the word `inspissated' mean, as used on this occasion? It means `unduly complicated'. What if our neophyte, fresh from a lecture which has been only too easily comprehensible, says: «That was inspissated». What has been said, strictly, is that the paper is complexly dense. What is meant by the speaker is that the paper was unduly simple. What `inspissated' (the word) means on this occasion is `unduly complicated'. So in this case the speaker's meaning diverges from the standard meaning, and from the occasion meaning, and the occasion meaning diverges from standard meaning. Of course, we need have no divergence at all between the three, and, of course, the terminology is negotiable. Notice that occasion meaning attaches to a word or words: what the word or words mean on that occasion. Speaker's meaning is a matter of what is meant by uttering those words on that occasion. Standard meaning covers both what the words normally mean and what would normally be meant by uttering them.

We are not done backgrounding. Take an ambiguous utterance. Standard meaning provides more than one meaning. Normally, one of those will be what the word means on that occasion, as used by the speaker. Now take the role that context can play in modifying standard meaning. In Mark Antony's speech in Julius Caesar, when he says, frequently, «Brutus is an honourable man», what `honourable' means does not change as it is repeated. What he means though, by the utterance, does change. But the context makes clear how the speech is to be understood, independently of any other evidence, with in this case, the immediate linguistic context, the repetition of the word `honourable', being crucial. The context might not make this clear, however. Literary texts in particular are likely to provide several possible contextual meanings as candidates for the author's meaning.

Let me make sure that contextual meaning is understood, as it occurs in ordinary (non-literary) contexts. Take the case of an ambiguous sentence, S, with two possible standard meanings, A and B. So if S is uttered the speaker might mean A or might mean B. The speaker might also mean both (or neither, as in the case of irony) or, indeed, whatever the speaker happens to mean. So the occasion meaning could be A, or could be B, or could be A and B, or, whatever. But a choice has to be made between the various possibilities. The occasion meaning will be just one of these possibilities; viz. A, or B, or A and B, or, whatever. Speaker's meaning allows for various possibilities in terms of what might be meant by uttering S, but, as in the case of occasion meaning, must choose between them. Since context, linguistic and non-linguistic, both adds and removes possibilities, contextual meaning offers comparable possibilities for consideration. However, contextual meaning does not involve a choice being made between the possibilities. All are equally instances of contextual meaning, of what might be meant by uttering that (ambiguous) sentence, given the context. Not all, of course, need be equally likely to have been what was in fact meant; and what was in fact meant might not be predictable from the context, being totally idiosyncratic, and so not a case of contextual meaning.

In the case of a literary text, some would opt for speaker's meaning (provided it fitted the text) as giving the correct contextual meaning (interpretation); some the contextual meaning thought to be the most aesthetically satisfying; some would think it wrong to make any choice. But all the interpretations would have equal standing as contextual meanings of that text, provided each fits that text. Rejected interpretations remain contextual meanings of that text, of what might have been meant by that text, given the context, linguistic and non-linguistic.

So we now have the possibility of a four way divergence, though it remains the case that we need not have any divergence. It is important to see that occasion meaning and contextual meaning can diverge, both from one another, and from speaker's meaning (and, of course, from standard meaning). A specific, non-linguistic context might limit, or add to, what might be meant by and within a text, given what the words mean or might mean; and, of course, speaker's meaning might be gloriously or ingloriously private, indeed, idiosyncratic. Notice that it is only occasion meaning that can attach a divergent sense to a word or words. Contextual meaning is at one with speaker's meaning in generating, not word meaning, but what might be meant by an «utterance», conceived as encompassing a text, or stretch of text. Though contextual meaning is distinct from speaker's meaning, context includes the speaker, and one (or more) of the contextual meanings might be what the speaker meant in particular cases.

We need these four concepts of meaning for the ordinary, non-metaphorical range of linguistic facts. They offer enough options for metaphorical meaning, though contextual meaning does not seem to be among the options on offer in the writings on metaphor.

Sufficiently backgrounded, we can now return to the various disagreements. There are those who believe metaphorical utterances have non-literal meaning ranged against those who believe that metaphors have (at most) literal meaning. This is the big disagreement. The former group divides between those who are for speaker's meaning and those who are for occasion meaning (word meaning). There seems to be virtually universal agreement, however, that metaphors are not paraphrasable (at any rate, interesting metaphors are not).

But why are they not paraphrasable? DavidsonFoot note 2_1 is triumphantly clear: given the absence of any meaning other than the literal there is nothing to paraphrase! With one bound Jack is free. Whereas those espousing some form of metaphorical meaning seem to struggle to have any answer, he has the knock-down answer to why metaphorical utterances are not paraphrasable.

However, Davidson would seem to have what many would see as a knock-down problem. If we lose meaning, do we not lose truth? If we lose truth, do we not lose metaphor? Surely metaphors can be true, even interesting metaphors. Where there is meaning there can be truth. Devoid of meaning, Davidson is devoid of truth.

To decide for or against Davidson I need first to consider the choice confronting those attracted to metaphorical meaning. This has heretofore been that between speaker's meaning and occasion meaning, between Searle most famously on the one hand, and Beardsley and Black on the other. Various objections have been advanced against metaphorical meaning as speaker's meaning, but they can be exemplified by Beardsley's objections.

He claims that speaker's meaning (as utterer's meaning `cannot account for our ability to interpret metaphorical expressions even when we know that the relevant properties were not meant by any utterer.'Foot note 2_2 In addition, he claims that occasion meaning (my terminology) fits better with the move from live metaphor to dead metaphor; and with the fact that there is a continuity between metaphorical «senses» and literal senses; and with the fact that metaphorical «senses «behave in many of the same ways as literal senses. For example ... we can devise metaphorical equivocations» (p.11). He also insists on the «rule-guided character of literary interpretation» (p.11).

What these objections all trade on is surely the essential freedom of speaker's meaning, indeed its possible total arbitrariness. This freedom is, however, incompatible with metaphorical utterance. You cannot decree that your utterance be a metaphor by fiat. Your utterance can, indeed, be totally idiosyncratic, and risk, or court, being incomprehensible, but it cannot then be a metaphor. No more than a raised eyebrow can a metaphor exist by itself. It needs a linguistic context that will sustain a metaphorical interpretation. In this it is different from ironic utterance, where my insistence that I am speaking ironically may be treated with incredulity by all and sundry, but no matter; I can insist, and it is possible that I am.

So speaker's meaning is ruled out for metaphors. It is too wide. Of course, speaker's meaning may be the same as metaphorical «meaning» in a particular case, but a metaphorical «meaning» may force itself on us in another particular case, whether or not such a «meaning» was, or could have been, the speaker's. The words themselves, sometimes perhaps with a little help from the non-linguistic context, invite a metaphorical interpretation. So speaker's meaning is also too narrow. But if metaphorical «meaning» is seen as contextual meaning, and generated by the context, linguistic and non-linguistic, then it is not speaker's meaning, it is not loose, and it is not vulnerable to Beardsley's objections.

It is true that metaphorical utterance has the potential to be highly specific, depending on the features of that particular occasion, its circumstances, and participants. Such a moment for metaphor can, indeed, pass. But the dependence on the words remains. Even so, the outsider may well be at a loss, not knowing the particular circumstances. Contextual meaning still covers such a metaphorical utterance because of its dependence on the words, on the linguistic context.

If metaphorical meaning is not speaker's meaning, is it a case of occasion meaning; a matter of what the word, or words mean as used on that (metaphorical) occasion?

Surely attaching such new senses would be ex post facto? Faced with a text we make what sense of it we can, as a text, as a whole. What might be meant by this text, by this or that passage, by particular words within the text? What might be meant by particular words, given all the other words, and whatever else is germane? We work with, and within, the context. We need answers at this level before there could be any chance of assigning metaphorical meanings to individual words, and assigning such meanings is what the alternative to contextual meaning demands.

Also, we are not always able to assign meaning to individual words. Often metaphor works with phrases as the relevant unit and it seems strained to attach a new meaning to the phrase where we are unable to do so for the individual words making up the phrase. Such cases are handled straightforwardly within contextual meaning: as what is meant, or might be meant, by uttering the phrase.

A fortiori, if we think we should work on the level of sentences, then contextual meaning seems the natural way to go, and I think we should accept that we work no lower than on that level. Take George Herbert's two lines from his poem, Virtue: «Only a sweet and vertuous soul, like seasoned timbre, never gives». Try attaching new senses to some of these words, as opposed to the whole quotation. The sounds of words, the rhythms, allusions,(both internal and external) all go to require at least the sentence as the minimal unit to which a new meaning or meanings could be attached.

If that is accepted then what would be wrong with attaching a new sense to George Herbert's lines? What would be wrong is that this case of sentence meaning is unlike standard sentence meaning, and unlike divergent sentence meaning as it otherwise occurs. Here the sentence as a whole acquires a new sense or senses but not the individual words. Whatever understanding we have of this kind of sentence meaning seems parasitic on our understanding that what is meant by uttering a sentence may be different from what is normally meant. Insofar as it is thought to be a distinct notion, it is not surprising that this has led to notions of a special kind of meaning, and then to a special kind of (metaphorical) truth. Contextual meaning carries no such danger, being a particular case of something's being what is meant, or might be meant, by uttering (in this case) a sentence, where speaker's meaning is another such particular case. There is, therefore, nothing mysterious about either of these cases.

So contextual meaning should be the choice for metaphorical meaning. However, Searle's choice, speaker's meaning, is in fact proffered as possible speaker's meaning.Foot note 2_3 So perhaps contextual meaning is nothing more than that: possible speaker's meaning? Not so: possible speaker's meanings cast far too wide a net to catch metaphorical meaning, which is found only in contexts which include a linguistic context. Also, possible speaker's meanings which are candidates for metaphorical meaning are so only because they are contextual meanings. Because they are something that could be meant by the relevant utterance, given the context, then they could possibly be meant by a speaker. Such speaker's meanings are parasitic on contextual meaning. In addition, there is no guarantee that a contextual meaning will be such that it might, in fact, be uttered and meant by a speaker. There might be some taboo, or mental barrier or whatever. Again, contextual meaning is the prior notion, and a possible speaker's meaning is so only because it is a contextual meaning. So: a possible speaker's meaning can qualify as metaphorical meaning only insofar as it is a contextual meaning, and cannot qualify unless it is a contextual meaning.

So metaphorical meaning had better be regarded as contextual meaning. However, according to Davidson the only meaning a metaphorical utterance has is its literal meaning. There is more. Metaphors have no cognitive content, no cognitive content whatever, although, given that it would seem they share this feature with pictures, perhaps they might yet be allowed to be useful. Davidson's banishment of metaphorical meaning seems intuitively implausible, after all we think metaphors can be true or false, and we run them in arguments and inferences.

Davidson's extended broadside against metaphors' involving any meaning other than literal meaning is conducted, in fact, as a broadside against any notion that words have meaning other than their literal meaning. Speaker's meaning is not mentioned, never mind contextual meaning. However, the positive Davidson picture that emerges in the course of the broadside, makes it unlikely that either speaker's meaning or contextual meaning would shift him.

According to Davidson we are to respond to the metaphor, to give ourselves over to it. The effects the metaphor produces in us are what matters. Notably, `there is no limit to what a metaphor calls to our attention and much of what we are called to notice is not propositionain character' (p.44). What matters is the experience. That is why there is no meaning other than the literal; that is why metaphors are not paraphrasable. `Metaphor makes us see one thing as another... 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' (p. 45).

But metaphors seem to be capable of being true or false, and they cannot be so and lack meaning. Now Davidson is happy to accept that metaphor `does lead us to notice what might otherwise not 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, he insists that `the sentences in which metaphors occur are true or false in a normal, literal way, for if the words in them don't have special meaning, sentences don't have special truth. This is not to deny that there is such a thing as metaphorical truth, only to deny it of sentences' (p.39).

Why should it be thought that meaning which is other than standard will produce non-standard truth? If it is a different concept of meaning then the consequence might follow, just as if metaphorical meaning is mysterious then metaphorical truth might be mysterious. But speaker's meaning is not mysterious, nor is it a different concept of meaning. Standard meaning is, of course, different from speaker's meaning but that is not because there are two different concepts of meaning. Why should speaker's meaning produce only speaker's truth? What might be meant by such a claim? Speaker's meaning gives us no clues to what might be meant by speaker's «truth».

Perhaps we should regard Davidson as having done nothing other than reject the idea that metaphorical sentences acquire new meanings. So interpreted, he would be happy with metaphorical meaning otherwise understood. However, that would bring the problem of paraphrase back again, and also seems incompatible with his positive characterisation of metaphor. Perhaps his view is that metaphor may cause us, for example, to see that some thought is true. But that would be one only of many possible effects, and would be an effect of the metaphor, not intrinsic to it. The inspired thought could be called true, not the metaphor, while the inspired vision could only metaphorically be called true.

Is he right, so interpreted? Those who attach metaphorical meaning to words are claiming that the words on that occasion of their use have a meaning other than their standard meaning. Just as the word `inspissated', used by someone who has the wrong notion of what `inspissated' means, will have a sense other than its standard sense. They will also (generally) mean by the utterance, in which no doubt `inspissated' will be the star turn, something other than what is standardly meant. If they have expressed, in so doing, the belief (say) that my categorisation of meaning is unduly complicated then their claim is, of course, false, while what they have (strictly) said (that it is complexly dense) might by some be thought to be true. Surely, here, true is true and false is false, and the twain are no more likely to meet here, than they are anywhere else.

Another example of nonstandard meaning's allowing for truth, is irony. Someone says, of an embarrassingly elementary paper, that it was inspissated, and opines, by saying that, something like that it was embarrassingly elementary. Someone could disagree and think the opinion false, or agree and think it spot-on true. This is surely plain old, boring old truth, spot-on or otherwise.

Truth has been attached in these two examples in the one case to sentences (as uttered) and in the other to what has been meant by uttering the sentence. In the examples given, the point against Davidson could have been driven home this way: what has been (strictly) said (that the paper was inspissated) is not what the sentence means, in Unduly Complicated's case, nor what is meant by the sentence, in the case of Complexly Dense's ironic utterance. But what is, in fact, meant could as well, though no better, have been said directly. Why then, given this equivalence, should direct opining be true or false but its deviant sibling be denied this excitement?

It is true that metaphors differ from the examples above in that they are (generally) not paraphrasable. But being paraphrasable is neither necessary nor sufficient for truth. `This little rooster is red' is not paraphrasable. I pluck this example fairly randomly from many on offer. `Rhett Butler was a gentleman'. `Dalziell is a drongo'. `He's a ratbag but he's me mate'. But it might be thought that metaphors are in principle, if they are so, not paraphrasable. Why, though, should this difference, if indeed it exists, have as a consequence that such metaphors have only literal meaning? Trying to explain what a ratbag is, in practice is very like trying to explain what a metaphor means. We grope, and continue to grope, never doubting that we know.

Those clinging to this difference, should consider some literal comparisons. `She looks like Bette Davis', where it is known how Bette Davis looks, will be loaded with information, which cannot, in principle, be paraphrased. I might well say, finally having found this Bette Davis look-alike, all hope suddenly gone: `She doesn't at all!' More was meant by saying what was said, than was said.

This talk of meaning should not be resisted. Someone is claimed to be like a colleague, where the context makes clear that what is being claimed is that they are kind. There should be no unease in treating this as a case of speaker's meaning, where what is meant by saying `He is like Alec' is that the person is kind. Now this is paraphrasable; but what if the claimed likeness is in respect of my unusual walk. Then I think they mean more by what they have said than they have said, but that more is not paraphrasable.

There is, then, no problem in principle in metaphors meaning other than their literal meaning; nor is there a problem in principle in what they mean or might mean not being paraphrasable; nor in their being true or false. However, Davidson seems to have the advantage that he has a clear explanation for metaphors not being paraphrasable: nothing other than the literal is meant, so there is nothing to paraphrase.

Searle believes that metaphors are intrinsically not paraphrasable, `because without using the metaphorical expression, we will not reproduce the semantic content which occurred in the reader's comprehension of the utterance' (p.123). But this would seem to hold equally of ironic utterances and they do not seem to be intrinsically not paraphrasable.

Black refers to Toynbee's `No annihilation without representation', and says that this `could no doubt be spelled out to render his allusion to the familiar slogan boringly explicit...[but]...something of the force and point of the original remark would then have been lost.'Foot note 2_4 However, it would not be boring to someone who did not know of the allusion, and would surely be egregious rather than boring to those who did. Once again, though, ironic utterance would turn out not to be paraphrasable on this test.

What these examples demonstrate is that we should not confuse a metaphor's being paraphrasable with its being replaceable without loss. A poem is not replaceable by a literal paraphrase. If the poem has meaning, what is meant by the poem would not exhaust the features that make it a poem, make it valuable. No more is the claim that someone is AC/DC, replaceable; yet it is no trouble to say what it means. So a metaphor can have meaning even though there is more to a metaphor than its meaning.

So why are metaphors generally not paraphrasable? Consider a feature of non-metaphorical language. `Red' enables us to refer to an element of our colour experience. `Bette Davis eyes' allows us to refer to such eyes. However, `It's red' isn't paraphrasable, and intrinsically so, because it depends on the link with experience; as, for another example, does our understanding of what a burning pain is like. The experience can be picked out, but cannot be put into words. The way someone looks can be picked out, but cannot be put into words. Metaphors enable us to pick out more complex features of reality. It has been claimed that the tango is a metaphor for the way men and women relate. If it is indeed so then that view of an aspect of reality, and how it is experienced, has been presented by the metaphor, but cannot otherwise be presented.

Metaphors can make very specific links with features of reality. Leave the metaphorical utterance the same but change its context, then it has changed. The metaphor only exists in context, so the metaphorical utterance can be the same and the metaphor different, if the context is different. What metaphors convey is a singularity.

Is there still a problem for the devotees of meaning? Simon Blackburn thinks so, stressing that metaphors are open-ended, indefinite. Referring to Romeo's calling Juliet the sun, he claims that our response to the metaphor `is quite open-ended. Shakespeare need have had no definite range of comparisons which he intended, and it is quite wrong to substitute some definite list and suppose the exploration is complete. The metaphor is in effect an invitation to explore comparisons.

But it is not associated with any belief or intention, let alone any set of rules, determining when the exploration is finished.'

The first point to make on indefiniteness is that literal meanings can be indefinite. Any literary text can receive more than one interpretation. It is not thought that there is a limit to the number of possible, even sensible, even eminently satisfying interpretations. Nor is it thought that this means that complex texts lack meaning. By parity of reasoning, the fact that there need not be just one «meaning» to a metaphor, is no reason for denying that metaphors have meaning. They can have «meanings», so they can have meaning. Presumably the controversies that surround the interpretation of texts will gather round the interpretation of metaphors: is there a correct interpretation?; is it the author's?; and so on and so on.

If it is insisted that the indefiniteness that attaches to metaphor attaches to the individual indefiniteness of the list of comparisons rather than there being an indefinite number of lists; to the individual «meaning» rather than a range of «meanings», the response is that this feature is also found in literal language. Open-texture and family resemblance are relevant theoretical notions that immediately spring to mind.

Another response to the indefiniteness problem is this, using the «comparison» approach for illustration. That A is like B need not tell us much. In context, that A is like B might tell us a good deal, to the extent of shocking us, by way of telling us, via a literal comparison, that A is evil. But the information might be less specific. In context, that A is like B tells us that A has some of the salient features of B, warmhearted to a fault etc., but does not pick out just which of those features A shares with B. Context is generally important here, as much to rule out as to rule in possible features. Figurative comparisons exhibit the same range within the informative band. The relevance of these facts to the indefiniteness issue does not depend on accepting the «comparison» approach. All that is needed for their relevance is the indefiniteness possible in that approach.

Susan Haack has stressed the cognitive usefulness of metaphor, and she has emphasised that `it is precisely because metaphorical statements are unspecific or open-textured that they are apt for representing novel conjectures in their initial and undeveloped stages, and for prompting investigation of what might be specific respects of resemblance.'Foot note 2_5 She has outlined elsewhere in some detail the exploratory usefulness of metaphor by way of her recounting how she worked her way towards her «foundherentism»; and the indispensible role played in that by the notion that `the way a person's beliefs about the world support one another is rather like the intersecting entries in a crossword.'Foot note 2_6 The responses above accommodate these important truths.

The attention just given to indefiniteness should not be seen as countenancing the degree of open-endedness to a metaphor's invocations that Robert Fogelin has characterised as allowing a `drift into the Davidsonian void' (p.112) of possible comparisons (or experiences or whatever). Any open-endedness in metaphor is always subject to constraints, sometimes powerful constraints. Fogelin has done a splendid job of elaborating the role of context in shaping the interpretation of metaphor, most markedly in the way a poem sets limits to our understanding of its metaphors (see pp.108-112).

Which brings me back to contextual meaning. I have argued that if meaning other than literal meaning inhabits metaphor, it has to be contextual meaning, not speaker's meaning, not word or sentence meaning (occasion meaning). I am now able to claim more boldly that metaphors do have meaning other than literal meaning and that this is contextual meaning. If metaphors are to be allowed the possibility of being true, then metaphors had better have meaning.