Jean-Paul RIOPELLE, L'hommage à Rosa Luxemburg, 1992 (détail) [ * ]

Falsity, Ambiguity, and Metaphors

Christel FRICKE [ # ]


Against both Nelson Goodman and Donald Davidson I argue for the view that one cannot explain what metaphorically used expressions of a given language mean unless some expressions of that language, namely those which can be used metaphorically, have not only extensional meaning (may it be referential or truth-conditional) but, in addition to that, conceptual or connotative meaning. Metaphorical meaning is a phenomenon of both semantics and pragmatics. In using expressions of a language metaphorically we transcend the bounds of its conceptual scheme without thereby putting the uniqueness of this scheme into question.

Metaphorically used terms are omnipresent in everyday talk, in scientific and literary language. In most cases we don't have any difficulty in understanding, what metaphors, that is metaphorically used expressions, mean - even though we might find it difficult to spell out in detail the meaning of a metaphor. If Edward Craig tells us in the introduction to his The Mind of God and the Works of Man that « this book has grown out of a series of radio talks », we understand that, at first, there were some radio talks and these talks proved to contain potential for further development the result of which is manifest in the book introduced.1 « Growing » can be literally applied to nothing but living organisms, and so speaking of growing radio talks is literally false. But if we understand Edward Craig's use of « growing » metaphorically, then we conceive of the radio talks and the final book as of an early and a later stage of a living and developing organism, and we can then get involved in a metaphorical reflection about the conditions under which this organism has grown so nicely as well as about its nutrition etc. Sometimes, we don't even realize the metaphorical character of an expression used in a particular context.2 How do we proceed to understand metaphorically used expressions or to use expressions metaphorically in the right or a right way? And what can be said about the semantics of a language whose terms can be used metaphorically?

Since Gottlob Frege's and Bertrand Russell's work on language and logic, scholars of so called analytic philosophy interpret the terms of a language (the syntax and semantics of which is sufficently clear to be used for scientific purposes) in a purely extensional way: A one place term means the individual entities of its extension, and a n-place term means the n-tuples of its extension. Both Nelson Goodman and Donald Davidson have tried to explain the meaning of metaphorically used terms on the basis of a purely extensional interpretation of them. In my view these explanations are not convincing. Metaphorical meaning is not extensional, but conceptual or connotative meaning in the first place. In order to defend this view I shall first discuss Nelson Goodman's and Donald Davidson's explanations of the meaning of metaphorically used terms and then try to develop an alternative proposal of explanation.

Nelson Goodman's answer to the question of what metaphors mean

All languages or symbol systems that Goodman takes into account in his book Langages of Art (LA)3 fullfill the conditions of radical nominalism: These languages or symbol systems consist of nothing but concrete individual signs (signs that can be combined in particular ways) and any universe of Goodmanian discourse contains only concrete individual entities. Goodman tries to develop a nominalistic theory of metaphorical meaning. He starts his discussion of metaphorical meaning by giving the following example: « We say of a picture of trees and cliffs by the sea, painted in dull grays, that it is sad. » (LA, II.2, p.50). Thereby we do not understand the description of the picture as sad literally, but metaphorically. What does it mean to understand an expression metaphorically? What makes an expression have metaphorical rather than literal meaning? Goodman develops his answer to these questions in three steps:

  1. He first states that the description of a picture as sad, understood literally, has to be judged as false.

  2. He then compares the metaphorical use of 'sad' to the use of ambiguous terms.

  3. In order to distinguish between metaphorically used and ambiguous terms he finally develops his theory of metaphors as a transfer of schemata.

I shall briefly present and then discuss these steps.

First Step

If a term is used for predication in a sentence and if this sentence is in its literal sense false and if there is no reason to take the author or speaker of this sentence either not to know the truth or to be a lyer, we do indeed in many cases conclude that we should understand the sentence in question metaphorically. If a picture is labeled sad, we usually don't take this label to be false, even though only living beings provided with sensibility can be sad and pictures do not belong to this species. Now, even if we restrict our attention to metaphorical sentences, we have to concede that the literal falsehood of a sentence is not a necessary condition of its meaning being metaphorical. The frequently mentioned example that illustrates this phenomenon is the sentence 'No man is an island'. This sentence can be understood both literally and metaphorically, and it can in both cases be held to be true or right. Nevertheless, the metaphorical meaning of 'No man is an island' is independent of whether this sentence in its literal meaning is true or false.

Second step

How can we, in the realm of a language with a nominalistic semantics, explain the metaphorical use of a term as justified even though this use, taken literally, is false? The application of a metaphorically used term to an object is literally false if the object in question is not among those objects to which this term literally applies. If, in such a case, we do not want to conceive of the application of the term to the object as false, we can change its literal interpretation, that is, we can add the object in question (and all objects sufficiently similar to it) to those objects to which the term literally applies. If the new interpretation of a term is not supposed to replace its old interpretation but to exist together with it as an independent lexical item of the language system in question, the term will then have two different even though equally extensional interpretations; it will be ambiguous. Goodman, however, holds the view that a term, when used metaphorically, does not obtain a new extension of objects to which it literally applies and which replaces its old extension, and that it neither obtains a new, second extension and thereby becomes ambiguous: « ... metaphor is not sheer ambiguity. » (LA, II.5, p.70). But it seems to me that if the new interpretation of the term is neither supposed to replace the old one nor to be added to the language system so that the term in question becomes ambiguous, then the new interpretation of the term cannot be justified as long as the interpretation of the language to which it belongs respects the demands of Goodmanian nominalism.

Where metaphors freeze and become part of the language system to which the terms making up the metaphor belong, the result is often ambiguity. Just think of feet as a basic unit of measurement as an example. But it is neither the case that all ambiguous terms are in one of their meanings metaphorical nor that all metaphorically used terms are just ambiguous. « Applying the term 'cape' to a body of land on one occasion and to an article of clothing on another is using it with different and indeed mutually exclusive ranges but is not in either case metaphorical. » (LA, II.5, p.70-71). And the term 'sad' does not become ambiguous by being applied metaphorically to a picture.

Third step

In order to explain how metaphor and ambiguity differ, Goodman develops his view of the transfer of schemata: « An understanding of metaphor further requires the recognition that a label functions not in isolation but as belonging to a family. We categorize by sets of alternatives. » (LA, II.6, p.71) Families of related labels or terms form a schema, and all objects to which any one of the terms of such a schema applies belong to the same sphere. In order to illustrate the function of schemata in cases of ordinary, literal predication Goodman gives the following example: « ... what counts as red, ... , will vary somewhat depending upon whether objects are being classified as red or nonred, or as red or orange or yellow or green or blue or violet. » (LA, II.6, p.71-72) The phenomenon Goodman has in mind here can be understood more easily if we have in our language not only expressions and individual, concrete objects to which expressions can refer or which they can describe, but also properties of these objects and relations in which they stand to each other. Then we can say that we start with a sphere of objects which resemble each other in a particular way because they have certain properties in common; in Goodman's example this is the sphere of colourful objects. For us who see these objects they are differentiated to a maximal degree with respect to their colour, that is their colours are differentiated to the extent to which we can discriminate colours. However, the schema of colour terms which we use to describe the objects of this sphere can be more or less rich or differentiated. Even if this schema contains all the colour terms the English language actually provides, it may still not be sufficient to cope with the degree of my visual discrimination. But in most cases we rely for the classification of given colourful objects on schemata to which less than all the colour terms of a natural language belong. If a child may chose a ball among four balls in all, one of them red, one green, one blue, and one yellow, then it will be perfectly sufficient if the child says: « I would like to have the blue ball » or « I would like to have the yellow ball ». The child has no need of any further linguistic discrimination of the balls' colours.

The function of schemata for our predications and classifications of objects can be explained in the realm of a language with an exclusively nominalistic semantics, that is, without any reference to either properties or relations. But in order to be able to compare different schemata interpreted over the same realm of objects, we need an all-embracing background language which contains comparatively highly differentiated terms for classifying the objects in this realm as well as the terms of the schemata in question, the extension of which can be further analysed in more differentiated terms. For example, where the background language distiguishes between orange-red, red, and bluish red, we could introduce a new term red-n the extension of which contains all those coloured objects which the background language classifies as either orange-red, red, or bluish red.

Goodman's main point is that whenever a term is used metaphorically it is not only this term which is used for the classification of objects to which it does not literally apply; it is a whole schema of terms to which this term belongs and which is transfered to a new sphere, a sphere of different objects. That not only a single term, but a whole sphere of terms is transferred from one sphere of objects (the sphere of their literal meaning) to another sphere (the sphere of their non-literal, e.g. their metaphorical meaning) is supposed to allow for the distinction between a non-literal, e.g., a metaphorical use of a term, and a use through which the meaning of the term is simply becoming ambiguous. « How, then, do metaphor and ambiguity differ? Chiefly, I think, in that the several uses of a merely ambiguous term are coeval and independent; none either springs from or is guided by another. In metaphor, on the other hand, a term with an extension established by habit is applied elsewhere under the influence of that habit; there is both departure from and deference to precedent. » (LA, II.5, p.71).

The new interpretation of the terms belonging to a schema can be due to a transfer to an entirely different realm of objects, but it can just as well be a re-interpretation of these terms over the same or an only partly different realm. The transfer of a schema from one realm to an entirely different realm of objects is for Goodman the distinctive feature of metaphors. For example, the transfer of terms which describe different feelings (such as desperate, sad, cheerful, and happy) to a realm of non-living and non-sensible objects (including pictures and other kinds of material objects) is a strategy for metaphorically applying these terms. To describe a picture as sad is therefore metaphorical. Re-interpretation of a schema over the same or a similar realm of objects as the original, literal one is the strategy not for forming metaphors, but for irony, hyperbole or litotes.

At first sight Goodman's theory of the transfer of schemata of terms from old to new spheres of objects seems to be both convincing and of simple elegance. It promises to provide a unified explanation of different kinds of non-literal uses of terms. But a closer look at this theory reveals serious problems. There is the question whether Goodman, with his theory of a transfer of a schema underlying metaphorical predication, succeeds in establishing a distinction between the metaphorical and the ambiguous meaning of terms. I don't think he does. If it is not only one term which, when used metaphorically, is transferred to a new realm of objects and thereby obtains a new extension, if all the terms of a schema are thus transferred and obtain a new extension, the result is nothing but ambiguity; in this case, a whole schema of terms becomes ambiguous. In a language the semantics of which fullfills the strict conditions of nominalism, every kind of non-literal use of a term and thus every kind of metaphorical use of a single term or a whole schema of terms can be nothing but either false or ambiguous.

« Briefly, a metaphor is an affair between a predicate with a past and an object that yields while protesting. » (LA, II.5, p.69)

« Metaphor requires attraction as well as resistance - indeed, an attraction that overcomes resistance. » (LA, II.5, p.69/70)

« In metaphor ... a term with an extension established by habit is applied elsewhere under the influence of that habit; there is both departure from and deference to the precedent. » (LA, II.5, p.71)

I perfectly agree with all these Goodmanian formulae. Here, Goodman rightly describes the phenomenon of metaphor as determined by two opposing forces: as a justified violence to a rule, as breaking semantic laws, but a breaking that reveals a similarity between objects, a similarity that had till then gone unnoticed. Because of this similarity the breaking of semantic laws remains unpunished. Now, my objection to Goodman's theory of metaphor is this: He can only explain that the meaning of a metaphorically used term differs from the meaning it has when literally used. Goodman can explain the « protest » and the « resistance » that is being provoked by the metaphorical use of a term, but he can neither explain in which sense a term is used metaphorically with deference to it's literal use nor in which sense the metaphorical use of a term is being attracted by the objects of it's new extension.

Why is it the case that the meaning of a metaphorically used term cannot be explained on the basis of a nominalistic view of linguistic meaning? The main point of Goodman's theory of metaphorical meaning is the transfer of schemata. Now, one may think that the whole conception of these schemata and of the order of terms in them might not be explicable on the background of Goodman's nominalism. But this is not, or not entirely, the case; Goodman does have something to offer in order to deal with these problems, even though what he actually offers may not be sufficient.

The first problem is to explain how we actually chose the terms that belong to such a schema from all the terms belonging to the language in question. The criterium guiding this choice cannot refer to the syntactical properties of the terms that are to be chosen, it has to be semantic: The objects of a sphere which are classified by the terms of a schema must belong to the same semantic field. But Goodman's nominalism does not allow for semantic fields. The criterium in question can only relate to objects to which certain predicates apply. As far as I can see, there is only one possibility to explain the choice of terms belonging to a schema in the context of an exclusively extensional language (and this is not a very convincing one): Terms belong to a schema if and only if there is a term in the language in question which does not belong to the schema but the extension of which contains all and only the extensions of the terms of the schema. On the basis of such a criterium we can, for example, form the schema of all the colour terms belonging to a language: All and only the objects that belong to the extension of a particular colour-term also belong to the extension of the term 'colourful'. This explanation raises a suspicion of circularity, but I do not want to deal with it here. I just want to raise the question to what extent the suggested criterium of choice can help to establish the kind of schemata we need. Which schema do we rely on if we describe a human being metaphorically as a wolf? On the schema of all terms which apply to objects to which the term 'natural kind of animal' applies? Or on the schema of all terms to which the term 'is an animal of one of the kinds mentioned by La Fontaine in his fabels'? If we try to understand the meaning of the metaphorical description of human beings as wolves, do we then have all the natural kinds of animals in mind to be found on earth, the giraffes and catarpillars and all the others, or at least all the kinds of animals mentioned by La Fontaine? I doubt that this criterium will help us to form the schemata we need for interpreting metaphors. Goodman does not raise the question of how we obtain these schemata, he relies exclusively on the persuasive power of examples.

It seems to me that the choice of terms which are supposed to be in one schema raises serious problems if it is to be explained on the basis of a purely extensional language. Another problem arises when we try to explain the order of terms in a schema. Again, this order has to be a semantic order, not a syntactical order (as the traditional or conventional order of letters in the alphabet or the alphabetical order of terms in a dictionary - see LA, II.6, p.77/8). But how are we going to obtain a semantic order of the terms of a schema, given that we are not allowed to rely on properties or relations of objects in the realm of that schema, properties and relations which are measurable?

Goodman has shown - first in his The Structure of Appearance (SA) and then in chapter IX. of his Problems and Projects (PP) - that there is a solution to the problem of establishing a linear order of objects which respects the demands of nominalism and exclusively extensional semantics.4 But his solution to this problem operates with a trick: Goodman characterizes the individual entities in the realm over which the language we use for expressing our empirical knowledge of the world as qualia. He defines qualia as individual entities which are observable and which stand to other qualia in relations of apparent matching, that is « apparent indifference », or apparent non-matching, that is « apparent difference » (see PP, p.424). Goodman characterizes the matching-relation as being both reflexive and symmetrical, but not transitive.5  

This characterization may appear to be contra-intuitive. Nevertheless, it can be made plausible with the help of an example provided by C. I. Lewis. Imagine three samples of material A, B und C (as in a tailor's book of samples) which are identical with respect to their size, the material they are made of, their weight and their weaving structure and which are very similar with respect to their colour. Indeed, theses samples are so similar that we cannot distiguish neither A from B nor B from C. But as soon as we put B aside and compare A to C with respect to their colour we become aware of a minimal difference between their colours. That is, A and B apparently match just as B and C do, but A and C are nevertheless apparently different. Goodmans characterizsation of the relation of non-matching as both reflexive and symmetrical, but non-transitive turns out to be plausible.6 If A apparently matches B and B apparently matches C but A apparently does not match C, then not only A and C, but also A and B and B and C are different - even though not apparently.

In so far as they are observable and comparable individuals Goodman's qualia have something conceptual about them. Nevertheless, Goodman does not conceive of qualia as having observable properties - a feature that does not make it easier to understand in which sense qualia are supposed to be observable. On the basis of his conception of the relation of apparent matching between qualia and two additional relations between qualia, namely the relation of betwixtness and the relation of besideness (see PP, p. 428, SA, p. 216-227) Goodman succeeds in establishing a linear order of qualia. The establishment of this order fullfills the conditions of extensionality of the language in question; it does not rely on measurable properties of objects. Observing qualia and ordering them in a linear manner by testing their apparent matching, their betwixtness, and their besideness does not rely on any ascription of predicates to qualia.

Unfortunately, Goodman's theory of qualia and their linear ordering (which fullfills the conditions of nominalism and extensional semantics) is of little help for the explanation of the order of terms in a schema such as Goodman needs it in his theory of metaphor. A world of qualia can be fully and correctly described with the help of terms of a purely extensional language. In a world of qualia and its description there is no room for metaphorical uses of terms - and no need of it. But this is not the main problem of his theory.

My main objection against Goodman's nominalistic theory of metaphor is the following: We may - on the background of a nominalistic ontology and with the means of an exclusively extensional language - be able form schemata of terms and we may as well have semantic criteria to establish a linear order of these terms (at least when these terms apply to observable qualia). But as soon as the terms of such a schema are seperated from the realm of objects to which they literally apply in order to be transferred to some other realm, these terms just lose their meaning. As soon as these terms are used to classify objects in a new realm they obtain a new extensional meaning. But this new meaning has nothing whatsoever to do with their old meaning. There is no way in which Goodman could explain how, in such an exclusively extensional language, the new use of a term shows any deference to its old use or is attracted by the objects of its new extension. After having lost its old meaning and before having obtained a new meaning, i.e. in the process of being transferred from one realm of objects to another, a term of a Goodmanian language is just a meaningless sign - if it is not just a bare particular and then even loses the status of a sign. But if terms lose their literal meaning on the way to being transferred to a new realm of objects, they automatically lose their metaphorical potential.

Now, one might think that Goodman has an argument that allows him to reject this criticism, namely his distinction between representation and representation-as, that is the distiction between signs that denote objects of a certain kind and signs of a certain kind. Goodman introduces this distinction in the context of his analysis of pictorial signs; 'representation' means for him 'pictorial sign': « A picture that represents a man denotes him; a picture that represents a fictional man is a man-picture; and a picture that represents a man as a man is a man-picture denoting him. » (LA, I.6, p.27/8). Representation is denotation of something by a pictorial sign, the reference of a pictorial sign to actually existing objects. However, representation-as is not necessarily denotation or reference; a picture pictures an object as being of a certain kind by being itself a picture of a certain kind, and if there is no object to be pictorially referred to, the picture remains nevertheless a picture of a certain kind. What kind of picture a given picture is depends on its syntactical properties, on some of the physical properties of the sign-token. Goodman's defense against my main objection to his theory of metaphor as a transfer of schemata could rely on his distinction between representation and representation-as in the following way: Whenever a term or a schema of terms loses its reference to a particular sphere of objects, the sphere of its literal extension, to be transfered to a new sphere of objects, the sphere of its metaphorical extension, the terms in the schema do indeed lose their literal meaning, namely their reference to objects, but they do not thereby lose their syntactical properties; syntactically, they remain signs of the same kind. And these signs remain signs of the same kind even when they obtain their new, metaphorical meaning. Thus, Goodman could try to explain in which sense the metaphorical use of a sign is a use with deference to its literal use by referring to the sign token which, while being transfered from one realm of objects to another, does stay a sign of the same kind.

It may not be obvious in which sense this line of argument might lead to a defeat of my objection to Goodman's theory of metaphor. Doesn't Goodman himself underline that metaphor is not a syntactical, but a semantic phenomenon? Where the term-signs that are either literally or metaphorically used, are not pictorial but conventional with respect to their syntactical properties, where they are term-signs of a natural language, it will certainly be difficult - if not impossible - to explain the deference of their metaphorical use to their literal use by referring to their syntactical properties. But in the special case of pictorial signs the situation is different: The pictorial representation of a woman receiving a letter which is brought to her by her maid may mean metaphorically the Virgin receiving the message of her pregnancy with Jesus from an angel.7 Where pictorial signs are used metaphorically their syntactical properties, as they remain the same for both their literal and their metaphorical use, can indeed be seen as establishing a relation of deference between their literal and their metaphorical use. This is so because their syntactical properties are determined by the way the objects to which they refer in their literal use look or would look if they existed.

But this defense of Goodman's theory of metaphors, which is limited to the special case of pictorial signs anyway, is of a kind that Goodman himself does not have at his disposal. And this problem has nothing to do with his theory of metaphor but it is a consequence of his conception of pictorial signs. To put it briefly, Goodman denies that there is any difference in kind between signs of a natural language on one hand and pictorial signs on the other with respect to their being conventional. I am not convinced by this theory. But it may indeed be the case that, on the basis of his radical nominalism, there is no way to conceive of pictorial signs as syntactically non-conventional. And the defense against my objection to Goodman's theory of metaphor - limited as it is to the case of pictorial signs and their metaphorical use - relies on a conception of syntactical features of pictures as non-conventional and motivated by visual properties of the objects to which these signs literally apply. My objection against Goodman's theory of metaphors thus remains the same: The terms of a language whose semantics meet the conditions of Goodman's strict nominalism cannot be used metaphorically.

Donald Davidson's answer to the question of what metaphors mean

Davidson is certainly not a nominalistic and his view of linguistic meaning differs considerably from that of Goodman. But there is one point about which they both agree (together with many other analytic philosophers): They insist on a purely extensional interpretation of the terms of a language. Davidson holds the originally Fregean view that only in the context of a sentence do linguistic expressions have meaning. He conceives of the meaning of a sentence as its truth condition.8 The sentence 'Socrates flies.' means that Socrates flies, and the sentence 'Julia is the sun.' means that Julia is the sun. In order to write a dictionary where fixed meanings of the vocabulary of a language are given, the meanings of the terms of that language have to be abstracted from the meanings of the true sentences in which they occur. As Davidson holds the view that sentences in which terms are used metaphorically are usually false, he can say that the meaning of these sentences and of the terms they contain does not have to be taken into account when the dictionary meaning of these terms is to be fixed.9  

Davidson does not deny the phenomenon of metaphor and he is even willing to allow for metaphorical expressions in a scienticifally adequate language.10 In his famous paper « What metaphors mean » he holds the view that not only do we rely on the literal meanings of words for understanding metaphorical expressions but « that metaphors mean what the words, in their most literal interpretation, mean, and nothing more ».11 He thereby denies that in forming or understanding metaphors we make use of « semantic resources beyond the resources on which the ordinary depends ».12 The ordinary is, for Davidson, the literal, and thus he consequently rejects the idea of metaphorical meaning as distinct from literal meaning.

This denial of metaphorical meaning (even though not of metaphors) has to be understood against the background of his conception of literal meaning as exclusively truth-conditional and extensional and of his ideas concerning the learnability of a language. Human beings with their limited lifetime and limited intellectual capacity can learn a language only if the vocabulary of this language is limited. Terms of a language can be used metaphorically in many different ways. If one allows that terms of a language can have, in addition to their literal meaning, metaphorical meanings, and that they can have as many different metaphorical meanings as there are different possible metaphorical uses of these terms, then one will risk a multiplication of the elements of the vocabulary of the language and thereby put the learnability of the language into question. In addition, the view of the metaphorical meanings of a term as meanings additional to its literal meaning comes down to nothing but the view that terms which are used metaphorically become - by being so used - ambiguous. And this view has already been rejected. Thus, it seems as if Davidson rejects metaphorical meanings as additional meanings for very good reasons.

Nevertheless, the view that the dictionary of a natural language does not mention any metaphorical meanings of terms, cannot be defended. There are many so called frozen metaphors that have become part of the vocabulary of a given language. Elimination of these elements of the vocabulary risks to reduce considerably the wide range of application of the language or to destroy the possibility of its application all in all.13  

How, then, does Davidson reply to the question what metaphors mean? What makes a sentence which is literally false a metaphor is a particular way of its being used: « What distinguishes metaphor is not meaning, but use ... . »14 But how does Davidson characterize the metaphorical use of a sentence, how does he distinguish metaphorical from literal use? By which means can he establish this distinction, given his purely extensional concept of meaning? « No doubt metaphors often make us notice aspects of things we did not notice before; no doubt they bring out surprising analogies and similarities to our attention ... . »15 But how do metaphors make us notice these aspects and analogies?

For Davidson metaphors do not have a fixed cognitive content: « ... there is no limit to what a metaphor calls to our attention, and much of what we are caused to notice is not propositional in character. »16 In characterizing the cognitive content of a metaphorically used term as complex, unlimited and somewhat vague, Davidson does not reply to the question what determines this content. This content is not the meaning of the term, but it must somehow depend on its meaning and on the context of its use. « Metaphor makes us see one thing as another by making some literal statement that inspires or prompts the insight. »17 But how can a literally false sentence prompt such an insight? Does Davidson admit that, when we know the meaning, i.e. the truth condition of a sentence and when we know the meaning of the expressions it contains, we also know the conceptual content of this sentence and of the expressions it contains? Does he admit that it is this conceptual content on which we rely in order to decide whether the sentence, used literally in a particular context, is true or false? And does he refuse to admit conceptual meaning (of sentences and expressions contained in sentences) only for the reason of trying to avoid the ontological implications of such an admission? But what is such ontological avarice good for if it makes the answer to the question what metaphors mean impossible? In a later paper Davidson says explicitly that, in his paper on « What metaphors Mean » he only intended to reject metaphorical meaning as additional (to literal) meaning without thereby giving an answer to the question he had raised.18 And in his paper « Locating Literary Language » he explicitly admits: « In my essay 'What Metaphors Mean' ... I was stubborn about the word 'meaning' when all I cared about was the primacy of 'first meaning'. »19 He explains his distinction between first and other (non-first) meaning as follows:

« Sometime too hot the eye of heaven shines » means that the sun sometimes shines too brightly. But the first meaning of « the eye of heaven » purports to refer to the one and only eye of heaven. We can tell this because Shakespeare (we assume) intended to use words that would be recognized by a reader to refer to the one and only eye of heaven (if such a thing existed) in order to prompt the reader to understand that he meant the sun. We may wish to use the word « meaning » for both the first meaning and what the metaphor carries us to, but only the first meaning has a systematic place in the language of the author.20  

Here, Davidson explicitly allows for non-first, non-literal, and that is metaphorical meaning. But it seems to me that, on the basis of his monistic conception of sentence meaning as extensional meaning, he can neither explain what metaphorical meaning is nor in what way it depends on literal meaning. The reasons for this failure are very similar to those of Goodman's failure to explain what metaphors mean.

What, then, do metaphors mean?

In order to explain what metaphors mean and how metaphorical meaning depends on literal meaning, it seems to me that we have to admit a dualism of linguistic meaning. The dualism I have in mind is not a dualism of literal and metaphorical meaning in the first place but a dualism in the realm of the literal: the dualism of conceptual or connotational (literal) meaning on one hand and of extensional (literal) meaning on the other. The explanation I am suggesting makes use of both a Goodmanian and a Davidsonian element of thinking about what metaphors mean: of Goodman's idea that metaphorical meaning has its ground in a transfer of schemata and of Davidson's idea that in using or understanding expressions metaphorically we rely, first of all, on the literal meaning of these expressions and that metaphorical meaning, if meaning at all, is second-order meaning. But both these ideas can only be defended coherently on the basis of the semantic dualism I suggested above.

If we leave the phenomenon of frozen metaphors out of account and concentrate instead on creative metaphors (on metaphors in which the creative potential of language use is manifest), we can say that no term as such is metaphorical. A term is metaphorical only if it is used in a particular way. It is this use which creates the semantic tension characteristic of most metaphors. Thus, the phenomenon of metaphor is, and here I think I agree with both Goodman and Davidson, intrinsically pragmatic. But this by no means excludes that there is a semantic dimension to the phenomenon of metaphor too. Nor does it exclude that there is a pragmatic dimension to literal meaning. Compare the following two sentences:

(L) The sun is shining today.

(M) Juliet is the sun.

Tokens of the same term « the sun » are used in both (L) and (M). (L) makes sense if we understand it literally (and uttered at daylight); it is true or false depending on whether, at the time and place of its utterance, there is sunshine in the sky or not. « The sun » here is the brightest star in the sky, the star we see at daylight if the sky is not covered. With (M) things are different: If we take it literally and have to decide whether it is true or false, we will certainly judge it to be false. Julia is a human being and one and the same thing cannot be a human being and a star at the same time. But (M), taken literally, is not false in exactly the same way as (L), taken literally, is false (if uttered at daylight at a time and place where the sky is covered). (L), taken literally, makes sense, independently of whether, at the time and place of its utterance, it is actually true or false. (M), however, taken literally, doesn't even make sense; we cannot imagine conditions under which (M) could possibly be true. Thus, as long as we understand (M) literally and realize that it doesn't make any sense at all, the question whether it is true or false doesn't really arise. It is this absence of sense (rather than the absence of truth) which makes us try to understand (M) metaphorically.21  

How do we represent « the sun » in (L) and how in (M)? In order to answer to these questions we have to spell out elements of the conceptual meaning of the term « the sun ». If we want to understand the meaning of the term as it occurs in (L) and (M), a simple indication of its extensional meaning, the identification of a particular object, wouldn't be quite satisfactory. In the case of (M) the extensional meaning of « the sun » alone would be of no help in order to make metaphorical sense of the sentence. And in the case of (L) this meaning alone would leave us without any idea of the criterium which allows us not only to make sense of the sentence, but also to decide whether, at the time and place of its utterance, it is actually true or false. If we understand (L) literally and make sense of it, we represent the sun as the brightest star in the sky, the origin of daylight, visible at daytime when there are no or few clouds in the sky. In addition to that, we might represent the sun as creating the opposition of light and shadow, as the source not only of light, but also of warmth, or as a source of a radiation dangerous for the uncovered human skin. However, in order to make sense of (M), we have to represent « the sun » in a somewhat different way: We have to represent it as a bright object and as a source of warmth, but, in addition to that, as the center of a system and as an object which has power over human beings and their fate.

The context of Shakespeare's play « Romeo and Juliet » where Romeo utters (M) in the second scene of the second act helps us to understand how we should represent the sun in order to make sense of (M):

« ...
But soft! What light through yonder window breaks?
It is the East, and Juliet is the sun!
Arise, fair sun, and kill the envious moon,
Who is already sick and pale with grief
That thou her maid art far more fair than she.
... » (Act II, scene 2, verses 2-6)

Romeo's metaphor picks up the light-and-darkness imagery associated with Romeo and Juliet and this imagery will be further developed throughout the whole play.22 Taken apart from its context « Juliet is the sun. » may be misunderstood as a compliment of limited originality used by a lover who is young, romantic, and of conventional imagination. But J.W. Draper has associated the plot of the whole play as well as the personal characters of its principle figures to XVIth century astrology. The text contains detailed information about the time of Juliet's birth. Following the « horoscope of her nativity ... her planet should be Mars or the sun » - and with the latter Romeo associates her here.23 Thus, the seemingly conventional metaphor contains a hint to a fate dominated by heavenly powers. Indeed, throughout the play, Juliet is characterized in terms of light and brightness.

Whereas Juliet is associated with the sun, with light, Romeo is associated with darkness. Here, Romeo passes from darkness to light and life brought about by the presence of Juliet: There is a dimension of birth and rebirth in the scene, due to the conceptual connotations of « the sun ». These connotations are underlined at verse 50 in this same scene 2 where the idea of Romeo's rebirth is expressed through the topic of his new baptism:

« ...
...O, be some other name!
What's in a name? That which we call a rose
By any other word would smell as sweet.
So Romeo would, were he not Romeo called,
Retain that dear perfection which he owes
Without that title. Romeo, doff thy name;
And for thy name, which is no part of thee,
Take all myself.

«                           I take thee at thy word.
Call me but love, and I'll be new baptized.
Henceforth I never will be Romeo. »
(verses 41-51)

These are only hints to the conceptual connotations which « the sun » obtains in the context of Romeo's talk in this scene, in this play written at that time. There may be more elements of the conceptual meaning of « the sun » as I have mentioned, and it is to the readers and audience of Shakespeare's play to reflect about these elements. Whether Shakespeare himself has already had all these elements in mind, all the connotations we bring about or even constitute in our reflection about the conceptual meaning of « the sun » - this is a question which, to my mind, is of secondary importance. The interpretion of conceptual meaning of a term is a process which can be described as a working out or bringing about its possible connotations.

For most of us, astrology has lost its power. This doesn't hinder us from referring to it in order to understand the connotations of Romeo's seemingly trivial compliment comparing Juliet to the sun. But when we talk about the weather and wonder whether the sun is out or not, there is certainly no association of the sun with the powers that former cultures used to attribute to this star.

The interpretation of a term in the course of which we aim at understanding its conceptual content can be analysed in terms of a construction of a semantic field. A semantic field consists of a lexical field and of a domain of conceptual content which can be used to represent and thereby organize a realm of objects.24 The conceptual content of such a field contains not only representations of properties of objects and of relations between objects but also representations of properties and relations that people have - in former and present times - attributed to these objects as well as representations of (both literal and metaphorical) ways in which terms belonging to that field have been used. A semantic field can be conceived of as a network of interrelated terms which mutually determine their conceptual content.25 Thus, the domain of conceptual content characteristic of the semantic field to which the term « the sun » belongs does not only contain the properties and relations attributed to the sun by modern physics but also actual and former common sense ideas about the sun as well as all the mystical and religious powers that astrology and ancient cultures have attributed to it.

In cases both of literal and of metaphorical use or understanding of a term, we interpret its conceptual content by constructing a semantic field around it. The elements of the domain of conceptual content of this field are all possible elements of literal conceptual meaning of the lexical elements of this field. Thus, both our literal and our metaphorical understanding of the conceptual content or meaning of a term are, in the first place, cases of an understanding of literal conceptual meaning. But whereas a term is literally used or understood to represent objects of its common extension, the metaphorical use or understanding of a term is based on a transfer of the semantic field to which it belongs (on a transfer of the lexical field and its domain of conceptual content) to a realm of objects which is commonly organized by an entirely different semantic field. As a result of this transfer, a realm of objects is newly organized by the terms of a semantic field to the literal extensions of which the objects of this realm do not belong. In the case of the above quoted Shakespearian example the semantic field to which the term « the sun » belongs and which is commonly used to represent heavenly bodies is transferred to the realm of the characters in the play; it is mainly Romeo himself and Juliet whose relations are newly organized in terms of the semantic field of the sun and other heavenly bodies.26  

Insofar as the procedures of both literal and metaphorical interpretation include the construction of a semantic field around the term to be interpreted, both procedures include a pragmatic element. The pragmatically interpreted conceptual content of a term used literally serves as a criterium for deciding whether the sentence in which it occurs is, at the time and place of its utterance, true or false. In order to decide whether the sentence « The sun is shining today » is, at the time and place of its utterance, true or false, we represent the sun as the brightest star in the sky, seen at daylight if there are no or few clouds, and we thereby know how to proceed for making this decision: We look at the sky at the time and place where the sentence is uttered in order to see whether there are clouds or not, whether we can see the sun or not. The conceptual content of a metaphorically used term, pragmatically interpreted with the help of a semantic field, however, is used in the first place not as a criterium for deciding about the truth or falsity of a sentence, but as a means for reorganizing a realm of objects which do not belong to the extension of this term. When Romeo characterizes Julia metaphorically as the sun, Julia does not thereby become an object in the extension of « the sun ». Instead, her position and function among the characters of the play is compared to the position of the sun among the heavenly objects and to the function of the sun for both these objects and the people on earth and their fates.

The function of a metaphorically used term and the corresponding semantic field can be compared to that of a model, a structure thought of in a certain conceptual content which can be used to represent objects of different realms and thus serve to organize different realms of objects. The metaphorical potential of a term and its semantic field lies in its capacity to serve as such a model for the organization of different realms of objects. Every term and its corrsponding semantic field may have some metaphorical potential, but this does not imply that every semantic field can serve as a model for organizing every possible realm of objects. The model must fit in some sense to the realm of objects that is supposed to realize it. Therefore, the metaphorical application of a term to a particular object or to objects of a certain kind can be right or wrong, just as the literal application of a term to a particular objects or to objects of a certain kind can be true or false. Kittay gives the following example for the way in which sentences containing metaphorically used terms can be right or wrong: « Relative to a scheme in which cities are thought of as women, 'The Venice of the Renaissance was a noblewomen' would be true while 'The Venice of the Renaissance was a washerwomen' would be false. »27 However, given that 'The Venice of the Renaissance was a noblewomen' is, literally understood, false or senseless and that we rely on the conceptual literal content for both our literal and metaphorical interpretations of this sentence and the expressions it contains, I suggest speaking of the sentence as metaphorically right rather than true because I can thus avoid the contradiction of judging this sentence as both false (or senseless) and true.

Only if a term is metaphorically used in a correct way to represent an object or objects of a cerain kind does it make sense, only then metaphorical meaning emerges. Metaphorical meaning is conceptual or connotative meaning; it is the conceptual content of the thought representing an object or objects of a certain kind by means of a term which does not literally apply to these objects but which is a correct metaphor for them. Romeo rightly represents Julia as the sun; this representation corresponds perfectly well to her position in his social universe and helps to understand the logic of the drama. But had he represented her as a shooting star in the context of the very same drama, his metaphor would not have been right, it wouldn't have made any sense. The term « shooting star », applied metaphorically to the Julia of Shakespeare's play, wouldn't have any metaphorical meaning.

Insofar as the interpretation of a metaphorically used term starts from constructing its literal meaning with the help of a semantic field, the emerging metaphorical meaning of this term is a kind of second order meaning (as Davidson has suggested). It seems to me that the account of metaphorical meaning I am suggesting here does not provoke any of Davidson's objections. The metaphorical meaning a term obtains when used in a metaphorically right way is a kind of momentary meaning, the result of a process of interpretation and metaphorical reflection about the literal meaning of the term and the object or objects to which it is metaphorically applied. This process is neither a process of change of literal meaning of the term in question, nor a process of change from one (literal) to another (metaphorical) meaning of it, nor a process of establishing a new, additional meaning, namely a process of an emerging ambiguity of this term. Applying a term metaphorically in a right and convincing way does not lead to a lasting change of the vocabulary of the language to which it belongs and thus does not put the learnability of the language into question. Such an application may leave no traces in the language system at all. However, if a term is used metaphorically and if this use is right and convincing and if it happens in a prominent place (like in a Shakespearean play) then it may well leave a trace in the language: It may enter the semantic field of this term insofar as this field may contain a representation of the conceptual content this term has had in former metaphorical applications. If « the sun » is applied metaphorically today, the interpretation of its conceptual content may rely on a semantic field which contains somewhere at its borders former prominent metaphorical applications of this term, among them Romeo's « Juliet ist the sun ».

Davidson may have had in mind a process of interpreting and understanding a metaphorically used expression very similar to the one I am defending here. And this is what Sue Larson has suggested to me when I had the opportunity to dicuss this problem with her. But then Davidson should admit that there is more than extensional literal meaning to words, namely conceptual or connotative literal meaning. His conception of metaphorical meaning as second order meaning makes sense only if this second order meaning (and the underlying first order meaning) are conceived of as conceptual or connotative meaning. Davidson may well refuse to call the conceptual content of terms meaning because there is no way to spell out what this meaning precisely is. One cannot have a systematic theory of the conceptual content of a term (independent of whether it is used literally or metaphorically). Therefore, there are no rules to tell us how to interpret metaphors. But admitting that interpreting a metaphorically used term brings about a second order meaning of this term and describing this interpretation as a process in which our imagination is deeply involved,28 hardly goes together with the claim that metaphors do not make use of semantic resources beyond the literal, where the literal is conceived of in terms of purely extensional meaning.29  

Kittay proposes to analyse the distinction of metaphorical meaning from literal meaning of a term and of the corresponding semantic field in terms of a distinction between two incommensurable conceptual schemes. But this aspect of her theory of metaphors is at odds not only with Davidson's very idea of a conceptual scheme (which she rejects) but also with his conception of metaphorical meaning as second order meaning (which she accepts). Metaphorical meaning can rightly be analysed as second order meaning because we interpret a metaphorically used term on the basis of the conceptual scheme of the language to which it belongs and which determines its literal meaning. If, in the course of this interpretation, metaphorical meaning emerges, we do indeed transcend the bounds of this conceptual scheme, we try to represent certain objects in a way not provided for by this scheme. Nevertheless, we do not thereby establish a new conceptual scheme incommensurable with the first. Incommensurable conceptual schemes are supposed to be incomparable and untranslatable into each other. If the interpretation of metaphor involved a move from one concepual scheme to another, incommensurable with the first, we would never be able to understand a metaphorically used term in the right way and to be at the same time aware of the fact that, understood literally, it would be false or senseless. In trying to spell out the conceptual content thought of in a metaphorically used term we rely on the lexical means and the corresponding conceptual (literal) content of the very same scheme to which this term literally belongs. This does not imply that metaphorically used terms can be translated into literally used terms without any loss of conceptual meaning. The meaning of a metaphorically used term is indeed incommensurable from the point of view of the conceptual scheme to which this term literally belongs. But this is not because this term, when used metaphorically, belongs to a different conceptual scheme but because spelling out the conceptual content of a metaphorically used term is an endless task. There are no limits to interpreting the conceptual content of a metaphorically used term, it is a task for both our imagination and creativity, and therein lies the charm of metaphors.


1. See Craig, Edward: The Mind of God and the Works of Man. Oxford 1987, p. 1.

2. Johnson and Lakoff give many examples for expressions in our everyday language which we use and understand metaphorically without realizing their metaphorical character. See Lakoff, George and Johnson, Mark: Metaphors we live by. Chicago and London 1980.

3. Goodman, Nelson: Languages of Art. An Approach to a Theory of Symbols. Brighton 1981.

4. See Goodman, Nelson: The Structure of Appearance. Third edition with an Introduction by Geoffrey Hellman, Dordrecht 1977 (SA); Goodman, Nelson: Problems and Projects. Indianapolis 1972 (PP).

5. See SA, p. 209.

6. See C. I. Lewis: Mind and the World Order. New Vork 1956, p. 363/4. See also SA, p. 212.

7. The painting I have in mind here is Johannes Vermeers Mistress and Maid from the Frick-Collection in New York. It was Amelie Rorty who mentioned to me the idea of interpreting this painting as a representation of the Annuntiation. Thus interpreted this painting of Vermeer's has both a literal and a metaphorical meaning.

8. See e.g. Davidson, Donald: « Reality without Reference. » In: Platts, Mark (ed.): Reference, Truth and Reality. Essays on the Philosophy of Language. London et al. 1980, p. 131-140.

9. See Davidson, Donald: « What Metaphors Mean. » In: Davidson, Donald: Inquiries into Truth and Interpretation. Oxford 1990, p.245-264 (« Metaphors »), here p. 257: « If a sentence used metaphorically is true or false in the ordinary sense, then it is clear that it is usually false. »

10. « Metaphor is a legitimate device not only in literature but in science, philosophy and the law; it is effectice in praise and abuse, prayer and promotion, description and prescription. » (See Davidson, « Metaphors », p. 246.)

11. Davidson, « Metaphors », p. 245.

12. Davidson, « Metaphors », p. 245.

13. See Lakoff, George and Johnson, Mark: Metaphors we live by.

14. Davidson, « Metaphors », p. 259.

15. Davidson, « Metaphors », p. 261.

16. Davidson, « Metaphors », p. 263.

17. Davidson, « Metaphors », p. 263.

18. See Davidson, Donald: « Reply to Oliver Scholz. » In: Stoecker, Ralf (ed.): Reflecting Davidson. Donald Davidson Responding to an International Forum of Philosophers. Berlin/New York 1993, p.172-3.

19. Davidson, Donald: « Locating Literary Language. » In: Dasenbrock, Reed Way: Literary Theory After Davidson. New York 1993, p.307, footnote.

20. Davidson, « Locating Literary Language », p. 300. See also Davidson, Donald: « A Nice Derangement of Epitaphs. » In: LePore, Ernest (Hsg.): Truth and Interpretation. Perspectives on the Philosophy of Donald Davidson. Oxford 1986, S. 433-446.

21. In this respect I disagree with Louise Röska-Hardy's speech-act theory of metaphor. Röska-Hardy holds the view that we are willing to try to make metaphorical sense of a term in a sentence which, at first sight and taken literally doesn't make any sense, if and only if we take the utterance of the sentence to be a speech act by a rational being produced intentionally. But this is not a necessary but only a sufficient condition for the intention of interpreting a term metaphorically. In order to try to make metaphorical sense of a term in a sentence we do not depend on a speaker's intention. We can try to understand a term used metaphorically in a sentence even if we know that the sentence in question was uttered by a speaker of limited linguitic competence (by a speaker, for example, to whom the language in question is foreign) or by a machine producing arbitrarily sequences of words. See Röska-Hardy, Louise: « Metapher, Bedeutung und Verstehen. » In: Danneberg, Lutz, Graeser, Andreas und Petrus, Klaus: Metapher und Innovation. Die Rolle der Metapher im Wandel von Sprache und Wissenschaft. Bern, Stuttgart, Wien 1995, p. 138-150.

22. See Shakespeare, William: Romeo and Juliet. Ed. by T. J. B. Spencer. London/New York 1967; here verses 19-22.

23. See Draper, J.W.: « Shakespeare's 'Star-Crossed Lovers'. » In: Andrews, John F. (ed.): Romeo and Juliet. Critical Essays. New York a. London 1993, p. 285-306, esp. p.293.

24. See Lyons, John: Introduction to Theoretical Linguistics. Cambridge 1968.

25. For the way in which terms of a semantic field mutually determine their conceptual content see Saussure, Ferdinand de: Cours de Linguistique Générale. Publié par Charles Bally et Albert Sechehaye, avec la collaboration de Albert Riedlinger. Ed. critique par Tullio de Mauro. Paris 1982.

26. Here I follow Eva Feder Kittay who has suggested explaining what metaphors mean with reference to a transfer of a semantic field from one realm of objects to another. See Kittay, Eva Feder: Metaphor. Its Cognitive Force and Linguistic Structure. Oxford 1987.

27. See Kittay, Metaphor, p.313.

28. Davidson, Metaphor, p. 245.

29. I profited from discussions of Davidson's conception of meaning not only with Sue Larson, but also with Achille Varzi and other members of the Department of Philosophy of Columbia University where I was invited to read a paper about the subject. Further help came from discussions with Louise Röska-Hardy who shares Sue Larson's suspicion that, concerning the understanding of metaphors, Davidson and I have more in common than I think.

Previous Article /
Article précédent

Top / Début
Table of Contents / Table des matières
Next Article /
Article suivant