Sorites ISSN 1135-1349
Issue #18 -- February 2007. Pp. 76-97
Normativity without Exception: Donald Davidson on Language and Communication
Copyright © by Sonia Arribas and Sorites

Normativity without Exception: Donald Davidson on Language and Communication
Sonia Arribas

1. 1 Metaphor: The Exception to the Norm?

Language is ordinarily called «metaphorical» when it is replete with expressions that contain words with meanings different to the usual ones. It is commonly stated, for example, that in poetic language we find terms that have been removed from their usual context and inserted into a new one. According to this common understanding, it is this movement of words from one context to another that evokes in the reader or listener an aesthetic response. We have also become accustomed to hearing that, in the language of science, terms are used that originally mean one thing, but which, when applied to a new situation and directed toward a «specialized» community, end up meaning something completely different.Foot note 1 And, in more everyday terms, our sense of shock has long become anaesthetized to the bizarre twists given to words by all variety of self-appointed political orators. All of these phenomena -- those praised for their beauty or wit and those considered manipulative -- are popularly placed under the category of the «metaphorical». What this placement expresses is the belief that metaphor enacts a deviation with regard to familiar, typical or, more precisely, literal meanings; metaphor is a different -- either enlightening or distorting -- use of the meaning that we understand as valid. Given the existence of this popular conception -- which, somewhat contradictorily, tends to both extol and denigrate metaphor -- it should come as no surprise that the study of so-called metaphorical language has occupied a prominent role in philosophy, psychology, linguistics and literary criticism. In this paper, I want to use the terms «metaphor» and «metaphorical» in the popular sense of that which is «different linguistically speaking» in order to bring to light the mistaken philosophical and political presuppositions of this popular conception of metaphoricity. I want to call into question, that is, the presupposition that metaphor constitutes the exception to the norm. I want to claim that Donald Davidson's essay «What Metaphors Mean» (WMM)Foot note 2 provides the theoretical basis for this mistaken conception -- the supposed «exceptionality» of metaphor.

The idea of metaphor's «exceptionality» finds its most extreme formulation in Davidson's already classic 1978 article «What Metaphors Mean». In accordance with Davidson's famous thesis, metaphor possesses no metaphorical meaning whatsoever; it consists, rather, of a purely literal employment of words that give rise to unexpected «effects» in the reader or listener. I want to reconstruct in this opening section the main lines of argument of Davidson's article -- an article which, once again, provides the best exemplification of the thesis that metaphor produces a use of language radically opposed to the usual or literal meaning of words. In WMM Davidson formulates a compelling critique of those semantic theories of metaphor which comprehend metaphorical meaning as a meaning either added to or derived from the metaphor's literal meaning. My reconstruction of WMM will therefore allow me to briefly examine these semantic theories and to express my agreement with Davidson's critique of them. In the following section (1. 2), however, I will attempt to show how Davidson's understanding of metaphor falls prey to the same error committed by the very semantic theories he is criticizing: both undervalue metaphor by strictly separating it from literalness. In Davidson's view, literalness and metaphorical effects belong to two wholly independent spheres of communication: the sphere of linguistic meaning (semantics) and the sphere of linguistic use (pragmatics). In the remaining sections (1. 3 and 1. 4), I will investigate two of his later texts: «The Idea of a Conceptual Scheme» and «A Nice Derangement of Epitaphs». This will allow me to sketch the outline of a conception of metaphor that conceives it neither as a case of literalness (consistently with Davidson's critique of traditional semantic theories of metaphorical meaning) nor as a mere effect of the use of language (consistently with my critique of Davidson). I will understand metaphor as both meaning and use -- as something different from literal meaning but, at the same time, as something intrinsic to language.

In «What Metaphors Mean», Davidson condenses his thesis, provocative in its simplicity, in the following way: «metaphors mean what the words in their most literal interpretation mean, and nothing more.» (WMM, p. 245) Metaphor possesses no other meaning than the literal meaning of its words. In effect, however, we can distinguish metaphor from its linguistic meaning. Metaphoricity has to be conceived, therefore, not as a linguistic meaning, but as an «effect» or emotion evoked in the receiver. Davidson's argument in favor of this thesis proceeds for the most part negatively: as a critique of traditional semantic theories of metaphor and of the very notion of a discrete metaphorical meaning. It is only on the basis of these criticisms that Davidson articulates his own positive account of metaphor. I will here reconstruct this analysis of metaphor according to the same sequence of negative and positive claims. Davidson's critique of the notion of metaphorical meaning will be seen to be based on two presuppositions: firstly, on the idea that the literal meaning of the words that form part of a metaphor has to be visible in order for the metaphor to exist as such; and secondly, on the stipulation that literal meaning and metaphorical use are strictly separated. Davidson's positive account of metaphor is articulated as a «causal» or pragmatic theory -- one that conceives metaphor as a producer of effects.

Davidson's first presupposition can be stated as follows: if we are even to comprehend a metaphorical expression, then the literal meaning of the terms that make it up must remain active. Davidson defends this presupposition by criticizing two semantic theories of metaphor: the theory of «extended» literal meaning and the theory of elliptical simile.Foot note 3 According to the first theory of «extended» literal meaning, in a metaphorical utterance the literal meaning of a word is extended so as to include a new object under its scope of reference. In order to demonstrate the failure of this explanation, Davidson carries out a reductio ad absurdum by means of an example: the Biblical metaphor «the Spirit of God moved upon the face of the waters.» To explain the metaphorical use of the noun «face» here, the extended literal meaning theory states, as its name suggests, that the class of objects literally referred to by the concept «face» has been extended so as to include the object «waters». In logical terms, metaphor increases the extension of objects referred to by the word «face». In its metaphorical employment, that is, the word «face» literally means «the front part of the head» and «waters». If this account were true, Davidson contends, that is, if «face» literally referred to waters, then we would not possess an explanation of the metaphor we originally intended to explain, but an explanation of a new literal meaning for the word «face». Instead of elucidating metaphoricity, the theory of «extended» meaning provides us with an account of what is usually referred to as «dead metaphor». A dead metaphor is one in which a previously metaphorical expression has come to be accepted as literal -- as in the expression the «mouth of a river», in which the original literal meaning of the word «mouth» (the opening of the digestive tube), at least at first sight, goes unnoticed, and is instead understood as a literal designation for the point at which a river meets the sea. In a dead metaphor, the original literal meaning of the words that form the expression disappears. The theory of extended literal meaning only provides an account, therefore, of a kind of metaphor that has lost its metaphoricity. If metaphor is to be explained in a way that preserves its metaphoricity, Davidson concludes, then the original meaning of the words that make it up must remain active and be immediately perceived.

In his critique of the theory of elliptical simile, Davidson once again argues for the necessary persistence of the literal meaning of the terms that constitute a metaphor. The elliptical simile theory tries to provide an account of metaphor in terms of its corresponding simile: the literal meaning of a metaphor is seen to be equivalent to the literal meaning of the related simile, and its metaphorical meaning to the simile's figurative meaning. The literal and metaphorical meanings of the expression «Tolstoy was an infant», for instance, are equivalent to those of the simile «Tolstoy was like an infant». Davidson subjects this account to criticism for two reasons. Firstly, because in equating the metaphorical meaning of metaphor with the figurative meaning of simile it explains nothing: a simile «no more tells us what similarities we are to notice than the metaphor does.» (WMM, pp. 254-5) Even when we accept that the metaphor and the corresponding simile could have the same figurative meaning, what simile figuratively conveys «is not a feature of the word that the word has prior to and independent of the context of use.» (WMM, p. 255) We cannot determine the figurative meaning supposedly shared by the metaphor and the simile simply by affirming that it is the same in both cases. Davidson's second criticism draws attention to the misapprehensions that necessarily follow from an identification of the literal meanings of metaphor and simile. The proposition articulated by a literal interpretation of the simile «Tolstoy was like an infant» explicitly affirms, for example, a similarity. And, on its part, the proposition expressed by the literal interpretation of this metaphor is either a falsehood (since Tolstoy did not write novels when he was a child) or an obvious truth (since Tolstoy, like everybody else, had a childhood). Whatever the truth content of the literal meaning of the words that form part of metaphor might be then, Davidson concludes, it has to remain active for metaphor to exist at all. When «we take the literal meaning of the metaphor to be the literal meaning of the matching simile, we deny access to what we originally took to be the literal meaning of the metaphor.» (WMM, p. 254)

The second presupposition underlying Davidson's understanding of metaphor comes to the fore in his critique of two strictly semantic theories of metaphor: one of which is derived from the philosophy of Frege, the other of which attempts to translate metaphor into paraphrase. This second presupposition can be summarized as follows: the capacity of learning the meaning of words and the capacity of using these words belong to two radically distinct spheres. Or, with more specific regard to the question of metaphoricity, whereas the literal functioning of words pertains exclusively to the learning of language, metaphor emerges exclusively through the use of language. Firstly, then, the Fregean theories criticized by Davidson conceive of metaphor as possessing both a literal and a metaphorical meaning. These two meanings are connected by a rule that specifies the situation or context in which each obtains. This explanation of metaphor is derived from Frege's account of the meaning of sentences that express propositional attitudes such as belief or desire («I wish I were healthy», for example).Foot note 4 The meaning of these kinds of expression can be ascertained, Frege claims, by means of a rule that derives it from the meaning of the words of the corresponding modal sentence («I am healthy», for example). Davidson refutes this characterization of metaphor -- as the result of the application of a rule -- by devising a somewhat curious example. If we were to teach a Saturnian the word «floor» for the first time by indicating floors and simultaneously uttering the word «floor», the Saturnian would be, Davidson claims, learning the meaning of the word. If the Saturnian were to subsequently want to test her grasp of this newly acquired knowledge, she would utter the word «floor», point at a floor, and await our confirmation. This second instance would also, in Davidson's view, be one of learning a new use of the literal meaning of the word «floor». But suppose the Saturnian were then to fly back to Saturn, point at the earth, and exclaim «floor». How could we then discern whether she is learning a new literal meaning of the word «floor», that is, whether she thinks that the earth literally is a floor, or whether she is using a metaphor, as when Dante refers to the earth as «the small and round floor that makes us passionate»? We have no rule, Davidson claims, to help us in the ascertainment of the Saturnian's intention and, in order to remedy this confusion, we need to make a clear distinction between the learning of the meanings and uses of words and the effective use of these words. Metaphors, to cut a long story short, only emerge in use. Three things follow from this example: (1) when the Saturnian attentively listens to our lesson on floors, she is learning the literal meaning of the word «floor» -- or, as Davidson claims, «what is learned is that a bit of language refers to a bit of the world» (WMM, p. 251). (2) When the Saturnian points to something else, exclaims «floor» and waits to see if she has got it right, she is learning a new literal use of the word «floor». And (3) when the Saturnian finally looks at the Earth from Saturn and says «floor», she is metaphorically using a word that she already knows. In accordance with Davidson's analysis, the radical difference between cases (1) and (2), on the one hand, and case (3), on the other, is that whereas the former deal solely with the mastery and use of literal meanings through a process of learning, the latter refers to the active usage of these meanings. When we learn a language, we remain within the confines of the literal; it is only when we use language in the world that we can enter into the domain of the metaphorical (this does not mean, of course, that every use of language is metaphorical). If literal meaning is entirely governed by the conventions that must be mastered in order to understand and speak, the use of metaphor arises impulsively and only occurs outside of the boundaries of these conventions.

Davidson's critique of the Fregean theory of metaphor is grounded, then, upon a strict separation between the literal interpretation of words (understood as a purely linguistic function) and the metaphorical usage of expressions (conceived as a worldly, extra-linguistic activity). It is the same distinction that centrally informs Davidson's critique of the paraphrasability theory of metaphor -- a theory which maintains that we can grasp the full meaning of a metaphor only when we paraphrase or restate it in words that exactly convey the meaning of the original expression. Although he never says so explicitly, Davidson's critique of this theory is based upon the notion of «cognitive content». In Davidson's theory, this notion functions as the guarantee of a fundamental unity of language -- of a definite, fixed and limited set of thoughts or things referred to by locutions and determined by conventions. Because the cognitive content of each word is stipulated by convention, it follows that we can always find or create another term (or set of terms) equivalent to it: the paraphrase. In addition and opposition to this notion of cognitive content, however, Davidson also wants to maintain that the production of non-cognitive effects is essential to communication. Non-cognitive effects -- in contradistinction to cognitive content -- are never fully determined or governed by convention. They belong instead to the realm of emotions and can thus be aroused involuntarily in the reader or listener. The paraphrasability theory of metaphor mistakenly presupposes, in Davidson's view, that metaphor possesses a cognitive content over and above that of the purely literal interpretation of its terms: «If what the metaphor makes us notice were finite in scope and propositional in nature, this would not in itself make trouble; we would simply project the content the metaphor brought to mind on to the metaphor. But in fact 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. When we try to say what a metaphor `means', we soon realize there is no end to what we want to mention.» (WMM, pp. 262-3). What Davidson is proposing here is a distinction between literalness -- conceived as the univocal articulation of a finite set of ideas -- and metaphoricity -- conceived as the expression or evocation of an inexhaustible number of suggestions and emotions which are, by definition, non-paraphraseable. According to Davidson's separation of spheres, literal meaning can be completely and precisely transferred from speaker to addressee; and the addressee can in turn grasp and even restate this meaning by means of an exactly equivalent expression. Metaphoricity, on the contrary, is an extra-linguistic phenomenon, one which cannot be substituted by any other term; it takes place in the world, as an external cause, a «bump on the head» (WMM, p. 262).

It is this distinction that allows us to summarize Davidson's causal theory of metaphor. If meaning is comprehended as a strictly propositional cognitive content capable of being paraphrased, then metaphor can only possess a literal meaning. The unparaphraseable and unsubstitutable character of the effects that a metaphor produces upon its addressee are the means to gain access to other cognitive contents. Metaphor assumes in accordance with this scheme a role similar to that played by speech acts. The effect of a metaphor -- like the effect of a promise or a lie -- does not reside in the literal meaning of the words that it enunciates, but exclusively in the use that we give to these words: «What makes the difference between a lie and a metaphor is not a difference in the words used or what they mean (in any strict sense of meaning) but in how the words are used.» (WMM, p. 259) Because metaphor lacks cognitive content, Davidson claims, it tells us nothing about reality; it possesses no meaning that could be deciphered or spelled out: «what we attempt in `paraphrasing' a metaphor cannot be to give its meaning, for that lies on the surface: rather we attempt to evoke what the metaphor brings to our attention.» (WMM, p. 262) Whereas literal linguistic meaning can be articulated and repeated independently of the context of its utterance, metaphor can only be identified and interpreted in the light of a given context. Metaphor produces an inexhaustible series of emotions, it invites us to look at reality in another way, it incites us to perceive connections between things that ordinarily go unnoticed: «When we try to say what a metaphor `means,' we soon realize there is no end to what we want to mention.» (WMM, p. 263) The cognitive power of metaphor has to be clearly differentiated, however, from literal knowledge; as Davidson cautions: «[s]eeing as is not seeing that. Metaphor makes us see one thing as another by making some literal statement that inspires or prompts the insight. Since in most cases what the metaphor prompts is not entirely, or even at all, a recognition of some truth or fact, the attempt to give literal expression to the content of the metaphor is simply misguided.» (WMM, p. 263). When confronted by a metaphor, the interpreter attains a new perspective on reality, not a claim about truth or falsity.

1. 2 Language as Relation

In summarizing Davidson's critique of the various attempted reductions of metaphor -- to a literal extended meaning, to a simile, to the application of a rule, to a paraphrase -- I have brought to light the two basic presuppositions of his theory; both of which refute the idea that metaphor possesses a metaphorical meaning. According to Davidson's first presupposition -- that of the necessary activity of the literal meaning of the metaphor's terms -- if a metaphor were to possess any other meaning in addition to the literal meaning of the words that compose it, it would immediately cease to be a metaphor. According to his second presupposition -- that of a strict separation between the meaning and the use of words -- metaphoricity is characterized as a contextual and extra-linguistic phenomenon that triggers effects irreducible to a determinate meaning. These effects can be considered as «useful» but not as meaningful. In this section, I want to argue that Davidson's first presupposition (concerning the necessary persistence of the literal) is broadly correct -- if a metaphorical utterance is to be understood, then the literal meaning of the terms must, to some extent, remain active. I also want to claim, however, that Davidson's second presupposition (concerning the strict separation of the meaning and use of words) contradicts his first: the strict separation of meaning and use precludes the assumption that the literal meaning of a metaphor's terms plays a role in its metaphoricity. In order to maintain my defense of Davidson's first presupposition, I will argue that literal meaning and metaphoricity must not be understood dualistically, but instead in terms of a relation. On the basis of this understanding, I will also try to show that metaphor is comprehensible, that is, that it possesses a meaning and, as a consequence, can be normatively assessed. In order to develop these claims, I will return, firstly, to Davidson's critique of traditional semantic theories of metaphor. I will concentrate, secondly, on the above-mentioned contradiction between Davidson's two fundamental presuppositions.

As we saw in section 1. 1, purely semantic explanations of metaphor tend to obliterate not only the literal meaning of the words that compose it, but also the very phenomenon of metaphoricity itself. When metaphor is endowed with a supposedly metaphorical meaning that is comprehended as eclipsing the literal meaning of its words, it is reduced to a convention: metaphorical meaning is conceived either as a literal but extended meaning or as a meaning derived from a rule of application. Once this is accepted, we need to maintain that, if a metaphor is to arise, there can be no conventional link between the literal meaning of its terms and the metaphorical phenomenon itself. If the metaphorical expression is to effectively produce the series of effects that Davidson describes, then the literal meaning of the words that constitute it has to remain active.

I want to show now, however, that Davidson's second presupposition -- that of an effective dualism between literalness and metaphoricity -- (and the series of epistemological oppositions that it assumes) renders impossible any explanation of why it is that metaphor is capable of provoking effects in its addressee. If we follow Davidson then we must say, on the one hand, that meaning is solely to be located in the literalness of words, that is, in a determinate cognitive content. This content is subject to learning processes: words univocally refer to meanings and the relations between them are ruled by conventions. These meanings and conventions are, on Davidson's account, the object par excellence of semantics. If, according to this view, convention determines meaning, it follows that the latter can always be ascertained independently of the infinite contextual particularities of communication. We must say, on the other hand, that metaphoricity is itself devoid of meaning, that is, that it lacks conventional cognitive content. As such, it is the object of pragmatics; it is the tool that allows us to use language in the world in order to cause or provoke suggestions and intimations. These suggestions and intimations are essentially indeterminate and hence radically different from accepted literal meanings. They produce effects which cannot be paraphrased, effects which are as infinite as the possible situations in which words can be used. If literalness and metaphoricity are radically disconnected, however -- in such a way that metaphor, conceived as an extra-linguistic phenomenon irreducible to literal meaning, always escapes the yoke of convention -- then why does Davidson consider it necessary to perceive the literalness of a metaphor's terms? Is there not a relation between literal meaning and metaphorical effects which contradicts Davidson's second presupposition?

We can begin to answer this question by concentrating on the so-called effects produced by a metaphor. Davidson's strict separation between meaning and use restricts these effects to the sphere of use and comprehends them as emotions or suggestions evoked in the addressee. These emotions are provoked by the speaker and are totally dependent on the context of the enunciation. They are purely «sensuous» and are opposed to the cognitive content of the metaphor's own terms. It follows from this that there takes place between effects and meaning only a juxtaposition, that is, an addition without connection. It is for this reason that Davidson maintains that, with regard to the literal meaning of the terms being used, the effects function merely as an accessory. If, however, according to Davidson's first presupposition, the literal meaning of the expression has to remain active in order even for metaphor to arise, then it seems impossible to avoid the inference that the understanding of literal meaning somehow causes or is related to the supposedly pure effects. The production of effects must be construed in relation to the understanding of the literal meaning. (Later on, I will examine in more detail the nature of this relation). Furthermore, it also appears self-evident that, if the arousal of effects depends upon the context in which the locution is enunciated, the literal and active meaning cannot be free from the influence of this context. It is precisely this conception of a necessary relation between meaning and effects that renders impossible Davidson's second presupposition (of a total separation between literal meaning and the contextuality of metaphorical effects). In nuce, there is a contradiction between the first and the second presuppositions of Davidson's analysis. This contradiction can only be avoided, I want to claim, if -- in accordance with a proper understanding of the first presupposition -- Davidson's dualistic understanding of literal meaning and metaphorical effect is replaced by a relational or dialectical one. As we will come to see, this relational or dialectical understanding will also compel us to call into question Davidson's strict separation between (literal) meaning and (metaphorical) use: both meaning and use will be seen to play a role in the literal and the metaphorical.

Before spelling out the nature of this relation, however, I want to make a brief detour in order to discuss a question that Davidson himself raises: the question of dead metaphor. The main problem that Davidson faces with regard to this question is the following: how can we explain the emergence of a dead metaphor from what used to be a living one? Let's us take as an example a typical expression from sociology and political science, that of a social «role» (from the French «rôle»), which literally means «a behavior appropriate for some relations but not for others.»Foot note 5 In its historical origins, the living metaphor «role» emerged as something that was used to designate theatrical behavior in social life. The term was translated, that is, from a theatrical context -- in which it denoted a «part in a play» -- to the wider context of social life -- where it was used to qualify certain public attitudes as artificial.Foot note 6 According to the «extended» literal meaning theory, the original meaning of «role» goes unnoticed when we speak of social life because the metaphor is dead and the word «role» now literally indicates a social position. The word «role» possesses, that is, at least two literal meanings. On Davidson's account, this amounts to saying that the meaningless effects that the metaphor originally triggered when it was first used as a living metaphor have transmogrified into the numerous meaningful traits of a social conduct denoted by a dead metaphor. How is it, however, that these effects, which are by definition incommensurable to meaning, can come to signify something determinate, conventional and context-independent? Because it presupposes that metaphoricity is a phenomenon wholly extrinsic to language, Davidson's account cannot explain the transformation from living to dead metaphor.

The previous considerations pose a series of problems that will lead me to formulate, in the following section (1. 3), an alternative to the theory of metaphor that Davidson advances in WMM. For the moment, I would like to advance two characteristics of the interpretation of literalness and metaphoricity that can only be comprehended by means of a relational or dialectical understanding of meaning and use. By pointing out these two characteristics, I want to once again draw attention to the above-mentioned problems with Davidson's position. Firstly, then, we have seen that Davidson totally rejects the notion of metaphorical meaning that is developed by strictly semantic theories of metaphor. The basis of this rejection -- the presupposition of the persistence of the literal meaning of the terms that constitute the metaphor -- is undermined, however, by the long list of oppositions that derive from the second presupposition. But if -- as I have already suggested -- we abandon the strict distinction between literalness and metaphoricity, we can hold on to the thesis that literal meaning remains perceptible, without needing to oppose to it a totally alien metaphorical effect. In this way, the so-called metaphorical effects can be conceived as intrinsic to language and, hence, as comprehensible, that is, as capable of being translated into a determinate meaning. This amounts to claiming that metaphoricity should be conceived as a link between meaning and use. And, conversely and in a dialectical movement, it means that the comprehension of literalness has to be reformulated with reference to metaphoricity, that is, once again, in terms of a relation between meaning and use. Secondly, this abandonment of the strict distinction between literalness and metaphoricity implicitly refutes Davidson's claim that «there are no unsuccessful metaphors» (WMM, p. 245). If the activity of the literal meaning of words is necessary for metaphor, then -- as we will see in the next sections -- this meaning (and use), when understood in relation to metaphoricity, will provide us with normative criteria by metaphor can be assessed (without reducing it to literalness). The traditional priority accorded to literalness is the result of the dichotomy that undialectically opposes it to metaphoricity. Once literalness and metaphoricity enter into a relation of co-implication, the notion of this priority can be discarded.

The strict distinction between literalness and metaphoricity -- as belonging to two epistemologically opposed spheres -- entails an erroneous conception of both metaphor and literal meaning and, thus, of language in general. The metaphorical is produced in a world supposedly external to language. In this world, the normativity or rationality of an utterance can assume no place, since the so-called effects of language are by nature incomprehensible. Language, for its part, is reduced to a fixed number of literal meanings, determined by convention and totally context-independent. I will henceforth refer to this mistaken conception of literal meaning as «restrictive», both because it conceives literal meaning as unchangeable and as unaffected by language-use, and because it limits language to a determined and fixed amount of restrictive literal meanings. In the following sections, I want to develop a critique of the notion of restrictive literal meaning -- and of the epistemological position that grounds it -- by advancing a reconstruction of two other essays by Davidson. Without explicitly referring to the question of metaphor, Davidson articulates in these articles a very different conception of language to that contained in WMM. My aim is thus to develop this conception into an alternative theory of metaphor.

1. 3 «Passing Theory»: Meaning and Belief

The relation between language and world is conceived in strictly dichotomous terms when the world is conceived both as external to language and as capable of exerting an epistemological impact on knowledge. In his 1974 article «On the Very Idea of a Conceptual Scheme» (CS)Foot note 7 Davidson uses a number of deflationary arguments in order to show the unintelligibility of this conception. The fundamental characteristic of the dualist conception of language and world, Davidson claims, is that language is conceived as a system or conceptual scheme that maps or projects itself onto a worldly reality. This model recurs, Davidson maintains, in different systems and conceptions of philosophy, all of which, as we shall see, coincide in comprehending language as a set of meanings organizing a supposedly external world. It is stated, for example, that language configures empirical or sensuous data, that it provides us with a point of view or that it endows us with a number of categories by means of which we can comprehend or analyze reality. According to Davidson, these various formulations of language as a conceptual scheme entail a thesis of relativism: given that the world is structured by conceptual schemes, different schemes can arrange the world in divergent or incommensurable ways. In linguistic terms, different languages speak about the world in ways divergent from and untranslatable into other languages. In order to call into question -- and ultimately abandon -- the idea of a conceptual scheme, Davidson discusses the two main forms in which the untranslatability thesis is defended. Firstly, the form of complete untranslatability, which is said to occur when no set of categories of one language can be converted into another; and, secondly, that of partial untranslatability, in accordance with which some categories can be translated, and some cannot. Davidson's critique of the thesis of complete untranslatability dismantles a concomitant dualism between language and world, that is, it dismantles Davidson's own second presupposition in WMM. Davidson's considerations on partial untranslatability prefigure a non-restrictive conception of literal meaning. This conception of a non-restrictive literalness is more thoroughly elaborated in Davidson's «A Nice Derangement of Epitaphs» (NDE),Foot note 8 to which I will dedicate the last pages of this section.

One common manifestation of the thesis of complete untranslatability of different languages (or of the incommensurability between different conceptual schemes) is the claim that translatability is not transitive. According to this idea, given three languages, A, B and C, when A can be translated into B and vice versa, and B can be reciprocally translated into C, it does not follow that A and C can be translated into each other. Davidson argues against this claim by questioning the presupposition that -- from the perspective of an A speaker, and given that A is not translatable into C (or vice versa) -- an A speaker is entitled to affirm that C has been translated into B. For in order for an A speaker to verify the validity of the translation of C into B she would have to recognize that the expression is part of C. Without this act of recognition A would never be able to verify whether a B speaker who claims to have successfully translated an expression from C to B is pretending or telling the truth. An A speaker could simply assume that the B speaker is not lying, but she would never be able to know whether the expression was actually translated. For A to be able to recognize what it is that a B speaker has translated, A has to recognize the expression as part of language C. And this already entails that A and C are translatable.

A second argument in favor of complete untranslatability is postulated by those philosophers of language and science who, inspired by Kuhn,Foot note 9 maintain that when different observers possess different conceptual schemes their perspectives about the world are also radically different. Davidson suggests that this relativistic position is generally articulated as an alternative to Strawson's thesisFoot note 10 that the distinction between concept and world is linguistic -- that is, that it is only within language that we can distinguish between a concept and its referent. Davidson shares Strawson's thesis, and in order to defend it makes a detour. The followers of Kuhn could attempt a criticism of Strawson, Davidson conjectures, by arguing the following: the analytic-synthetic distinction in language states that analytic sentences are true or false by virtue of their meaning alone, and that synthetic sentences are true or false because of their empirical content; if we accept Quine's critique of this distinction,Foot note 11 it follows that we should also give up the idea that meaning is independent of our beliefs about what is true or false; if meaning radically depends on belief, then whenever the speakers of a language change a sufficient number of beliefs, they will not simply exchange falsehoods for truths (or vice versa); they will rather adhere to a new set of meanings, that is, to a new conceptual scheme. Davidson proceeds to question this reasoning by pointing, by means of an example, at its basic presupposition. Suppose some philosophers were to advocate that instead of using terms that apply to psychological states we use only those terms, already present in our language, that refer to the equivalent physiological events in the brain. According to the followers of Kuhn, a change of scientific belief would have here compelled the creation of a new language or conceptual scheme, one radically incommensurable with, or untranslatable into, the old language or conceptual scheme. Davidson replies to this contention by claiming that, when scientific advances like this one compel a change in the use of language, this in no way means that, after the change, we are employing a new conceptual scheme incommensurable with the old. For, if that were the case, we would not be able to determine whether the words used in the new language actually refer to physiological events -- they could just as easily be synonyms of the old psychological terms. The criteria of the old language (the one that has apparently been superseded) must be presupposed if we are to be able to recognize that the supposedly new terms refer to physiological events. The Kuhnian argument in favor of untranslatability and, coterminously, in favor of the idea of a conceptual scheme, has to be discarded.

The third derivation of the argument for complete untranslatability discussed by Davidson also stems from the abandonment of the analytic-synthetic distinction in language: even if we give up this distinction, and drop its related conception of meaning, we can still hold on to the notion of empirical content. We can still maintain that all sentences -- and not only synthetic sentences -- have an empirical content. Empirical content can be interpreted on this view as the world, as reality, as the density of experience, as the totality of sensory stimuli or, more generally, as something outside language which, in order to be known at all, requires the imposition of an organized structure. Concomitantly, a conceptual scheme can be described as that which imposes the required order or form onto the empirical content. This formulation -- denominated by Davidson as the «third dogma of empiricism» -- merely reproduces, he claims, a dualism of world and language that is as untenable as the one informing the thesis of intransitivity and conceptual relativism. The argument proceeds in the usual manner: Davidson, firstly, presents the relationship between conceptual scheme and empirical content before, secondly, going on to prove its unintelligibility. This relationship is usually expressed, according to Davidson, in one of the following ways: either (1) conceptual schemes organize reality or experience or (2) they fit reality or experience. The dictum that conceptual schemes organize or put in order reality or experience can only possess a meaning, Davidson claims, if we take it to state that conceptual schemes organize objects within reality or sense-data within experience. For reality or experience, taken as a whole or as a single object, cannot be organized, just as it is not a closet itself that is organized, but the things inside it. Even if this criticism is accepted, however, a response remains open to those who defend the idea of a conceptual scheme. On the assumption that there exist different conceptual schemes organizing a manifold of objects within reality or experience, it could still be affirmed that, within one conceptual scheme, there might be some concepts which apply to determinate objects or sense-data and which do not have correlate concepts in another conceptual scheme. Davidson replies to this objection by claiming that we only perceive the lack of correlation between conceptual schemes if we acknowledge that there are a series of items (objects within reality or sense-data within the realm of experience) referred to by concepts in both conceptual schemes. Without this background of shared understanding, there is no possibility of a comparison of concepts. The first version of the third dogma of empiricism is, then, indefensible. In the second version, (2) above the emphasis is no longer placed on concepts that individualize and hence organize reality or experience, but on sentences that fit or correspond to it. To affirm that sentences fit reality or experience, Davidson claims, is to conceive of the latter as the place where we can find evidence for the affirmation. But if to offer this kind of evidence is to find the affirmation to be true, then to say that a sentence fits reality or experience is simply to say that it is a true sentence. But, what is it, Davidson asks, that makes us recognize a sentence as true? A fact in reality or experience? The evidence that we already possessed? This is the surprising answer:

Nothing (...) no thing, makes sentences and theories true: not experience, not surface irritations, not the world, can make a sentence true. That experience takes a certain course, that our skin is warmed or punctured, that the universe is finite, these facts, if we like to talk that way, make sentences and theories true. But this point is put better without mention of facts. The sentence `My skin is warm' is true if and only if my skin is warm. Here there is no reference to a fact, a world, an experience, or a piece of evidence. (CS, p. 194)

To believe something to be true does not imply that there is an object in experience or reality that is related to this belief and which makes it true. Nor do we possess an epistemological device that would allow us to relate or fit together an object and a belief with the aim of making the belief plausible. On the contrary, everything we need in order to be certain about a belief resides in the meaning of the sentence itself. To know that our belief is true is to know the language in which the sentence is uttered. If, after all of this, we still wanted to maintain the idea of a conceptual scheme -- as the idea of a set of sentences that fit reality or experience in ways incommensurable to other sets of sentences -- we would have to reformulate it so as to affirm that sets of sentences in different languages possess different truths. This last thesis, however, Davidson claims, cannot be valid. For to maintain that truth is defined within language means to apply the predicate «is true» to all the possible sentences for which truth is in question, including those from other languages. This is Tarski's famous Convention T,Foot note 12 which claims that a linguistic statement can only be true when it can be translated into another language. The sentence «snow is white», for example, is true in English if, and only if, snow is white (p); and true in another language when we can find in it an equivalent expression that can be translated into «snow is white» and substituted as p. Sentences can be understood and asserted to be true, that is, only in relationship to other sentences -- either from the same language or from other languages. Once again, the idea of a conceptual scheme as a non-translatable framework of truth has been refuted: truth cannot be conceived independently of translation.

Davidson's analysis has shown that the idea of a conceptual scheme -coterminous with the idea of conceptual relativism or complete untranslatability, and with a dualistic conception of language and world- cannot be justified by any of its most common philosophical defenses. This idea cannot be maintained, that is, either as an intransitivity between languages, as a relativism derivative from a change of belief or as a «third dogma of empiricism». What remains to be addressed, however, is the thesis of partial untranslatability. This thesis maintains that the basic conceptual presuppositions of differing views -- even if they are intelligible to one another -- are exclusive to their respective conceptual schemes. In reply to this thesis, Davidson inquires what it is that makes people capable of being understood and of understanding others. He tries to develop a theory of the conditions of intelligibility, that is, that gives an account of the interconnection between the attribution of belief and the interpretation of sentences. Communication and understanding is to be conceived as a two-fold process in which we can interpret others only because we attribute beliefs to them, that is, only because we attribute commitment to their sentences. By implication, this is a process in which we can trust the assertions of others only because we are able to extrapolate the meaning of what they are saying. This theory finds its most complete development in NDE. In this article, Davidson explicitly refutes the thesis of partial untranslatability and advances -- as against his own restrictive account of literal meaning in WMM -- a revised conception of literal meaning and, as a consequence, of metaphor. I will deal with Davidson's refutation of partial untranslatability at the end of this section. I want now to concentrate on his reworked conception of literal meaning.

The title «A Nice Derangement of Epitaphs» makes a pun on Mrs. Malaprop's infamous «a nice arrangement of epithets»; the topic of Davidson's article is precisely that kind of linguistic use that deviates from literalness. A malapropism, for instance, is a linguistic utterance in which an appropriate term is replaced by a term with a similar sound but a completely inadequate meaning. Despite their anomalousness, these utterances succeed as a form of communication. If successful communication occurs whenever the hearer understands the speaker in the way the speaker had in mind, then all it takes for communication to succeed, in the example at hand, is for the hearer to understand the malapropism as a malapropism. These preliminary remarks alone already indicate that what Davidson is envisaging here is an alternative, non-restrictive, definition of literalness. Whereas in WMM Davidson denied the possibility of meaning outside convention, in NDE he contends that the understanding of meaning can arise even in a supposedly irregular -- in this case, malapropistic -- circumstance. And whereas before he declared meaning to be context-independent, he now argues that «we want a deeper notion of what words, when spoken in context, mean.» (NDE, p. 434). Davidson's goal is, then, to reformulate the notion of literal meaning, without thereby making it indistinguishable from the speaker's intended meaning.Foot note 13 This distinction between literal meaning and the speaker's intended meaning cannot be given up for, if this were done, all meanings would be left to the intention of the speakers, that is, no room would remain open for linguistic understanding.

Instead of beginning with a direct refutation of the restrictive definition of literalness, Davidson prefers to interrogate the basic requirements that a new definition of literalness should satisfy once successful communication is accepted in supposedly irregular cases. A definition of literalness, Davidson claims, has to account for the fact that literal meaning «comes first in the order of interpretation.» (NDE, p. 435). I shall, for the time being -- in order to avoid the customary restrictive definition of literalness -- give the name «first meaning» to the literal meaning that needs to take the first place in the series of possible interpretations of an expression. First meaning is the meaning which -- given the possibility of different and alternative ways of interpreting an utterance -- seems to come closest to the intention of the speaker, and which, equally importantly, will be interpreted by the hearer as the speaker intends. First meaning is simultaneously determined, then, from both the speaker's and the listener's perspective. From the speaker's perspective, it is determined both by the intention to say something and by the intention that the hearer understand what is said (Davidson denominates the latter as «self-referring» or double intention). From the hearer's point of view, first meaning is determined when what the speaker says is understood, and when this understanding corresponds to what the speaker intended. In short, first meaning arises when a speaker's belief in and interpretation of an utterance corresponds to a hearer's.

Having adopted the concept of first meaning (instead of the common conception of literalness) as the basis of his theory of meaning, Davidson proceeds to analyze and modify three explanations of «irregular» cases that derive from the common conception of literalness. According to the first explanation, «first meaning is systematic», that is, it is immersed in a system of semantic relationships to other meanings. The second explanation states that «first meanings are shared», that is, in addition to belonging to a system -- as the first explanation stipulates -- they are also shared by speaker and listener. The third and final explanation maintains that «first meanings are governed by conventions and regularities.» (NDE, p. 436) We can begin to see the problems that these three proposals encounter when we consider the fact that malapropisms, despite their supposed irregularity, possess meaning:

[M]alapropisms and the like (...) introduce expressions not covered by prior learning (...) Malapropisms fall into a different category, one that may include such things as our ability to perceive a well-formed sentence when the actual utterance was incomplete or grammatically garbled, our ability to interpret words we have never heard before, to correct slips of the tongue, or to cope with new idiolects. These phenomena threaten standard descriptions of linguistic competence (including descriptions for which I am responsible). (NDE, p. 437. My emphasis)

Malapropism, along with any idiosyncratic use of words, seems call into question those descriptions of communication which -- like the three explanations summarized above -- reduce meaning to a restrictive account. Even if Davidson only mentions it in passing (in his critique of the third explanation), it is important to note here that metaphor can be understood to fall within the category of unconventional uses here under discussion. Metaphor invokes -- although not always -- novel forms of expression, it includes words that can only be properly understood within the context of their utterance, and it sometimes interferes with or suspends grammatical rules. Davidson's avowal of responsibility in the above-quoted passage has, then, to be comprehended as a reference to his own former theses in WMM.

According to the first explanation of first meaning -- that which construes it as forming part of a system -- the hearer is capable of correctly interpreting a novel use of language only because she is in possession of a finite system of first meanings governed by rules. Davidson partially accepts this theory of intelligibility: the hearer does make use of a finite and recursive set of rules when she encounters novel uses or enters into different contexts. He adds, however, the following qualification: this finite and recurring set can never be known by the hearer in its totality. On the contrary, this system -- which Davidson insists on calling a theory -- is always geared to the occasion and created ad hoc. According to the second explanation of first meaning, the latter forms part of a system that is shared by both speaker and hearer. Davidson does not want to give up this thesis completely: speaker and hearer do share the same «theory» when they communicate -- a theory made up not of fixed, context-independent conventions, but of context-adapted rules of use that are perpetually recreated in the act of understanding itself. This sharing of a theory does not amount, Davidson maintains, to the sharing of a language. It merely entails «the interpreter's and speaker's understanding of the speaker's words.» (NDE, p. 438)

If Davidson only modifies the conceptions of first meaning as systematic and shared, he certainly disagrees fiercely with the idea that first meaning is learned by conventions. This explanation has typically been put forward so as to preserve the distinction between literal or first meaning -- conceived as furnished by conventions -- and the speaker's intended meaning -- conceived as resulting from the speaker's whims. As I indicated at the beginning of this section, Davidson also construes the distinction between first meaning and speaker's intended meaning as essential to communication. Nevertheless, if we comprehend this distinction as coterminous with the dichotomy conventional / unconventional we are deprived of an explanation of those unconventional uses of language, such as malapropism and metaphor, where meaning does seem to exist. In order to propose an alternative to the conventionalistic understanding of meaning, Davidson takes recourse to the following example (developed by Keith Donnellan):Foot note 14 JonesFoot note 15 is an individual who affirms that «Smith's murderer is insane,» thereby meaning that a man, who happens to have murdered Smith, is insane. According to Donnellan, this sentence is true even if the man Jones is thinking of did not in fact murder Smith, as long, that is, as Smith's murderer -- whoever this is -- is actually insane. A traditional, conventionalistic explanation of Jones's sentence would reject Donnellan's interpretation. It would take him to be implying that the meaning of the sentence is only a matter of the speaker's intention, without any connection with the truth of the «facts». This explanation would counter Donnellan's interpretation by claiming that literal meaning is entirely conventional and that the speaker's initiative is completely arbitrary: given that Jones identified the wrong person, the sentence is false. Donnellan responds to this accusation, however, by pointing to the danger of reducing successful communication to the initiative of the speaker. This initiative is always accompanied by the speaker's intention and expectation that her words be interpreted as they were meant (by what Davidson denominates as «self-referring» intention). And furthermore, Davidson claims, to prevent the danger of the speaker's intention taking over «[a]ll that is needed (...) is a firm sense of the difference between what words mean or refer to and what speakers mean or refer to.» (NDE, p. 439) Following Davidson's contention, we only have to distinguish in Jones' sentence between the falsity of his intended meaning and the truth of what the sentence actually means. If the murderer of Smith -- whoever he is -- actually is insane, then, even if Jones wrongly identified this murderer, he still said something true using a false sentence. The first meaning is true; it is only Jones's intended meaning that is false. It follows from this that if meaning were to be determined solely by convention and if Jones were to point at the wrong person, his sentence would be false. The sentence is, however, when the context is taken into account, true. As there is no danger of this truth being left entirely to the whim of the speaker, the idea of convention as a safeguard of meaning can easily be discarded.

In order to develop an alternative account of first meaning and intelligibility, Davidson combines this critique of the third explanation of first meaning with the modifications made to the first two explanations. Meaning should be understood, he contends, as a process: a speaker and interpreter can only enter into communication because each possesses their own respective prior theory -- a set of beliefs and rules of interpretation that have been learned. Armed with their respective prior theories, speaker and listener attempt to understand one another. The speaker then utters something with the «self-referring» intention and expectation of being interpreted in a certain way. Let us suppose that this utterance is a malapropism; and that the interpreter's prior theory does not contain its meaning. What the interpreter now has to do is adjust her prior theory so as to incorporate all the contextual information available to her -- information that allows her to engage in conjecture concerning the speaker's own prior theory and behavior. Conversely, the speaker needs to include in her prior theory reflections about the constitution of the hearer's, as well as about the hearer's reaction to her speech. When communication succeeds, both speaker and interpreter have substantially modified their prior theories in accordance with this adaptive process. They now use, instead of prior theories, what Davidson calls passing theories: «For the hearer, the prior theory expresses how he is prepared in advance to interpret an utterance of the speaker, while the passing theory is how he does interpret the utterance. For the speaker, the prior theory is what he believes the interpreter's prior theory to be, while his passing theory is the theory he intends the interpreter to use.» (NDE, p. 442) Whatever is understood and meant to be understood at the moment of the utterance is the first meaning: this is the only agreement of or convergence between speaker and interpreter. Even if the passing theory has been reduced to such a brief moment and such a narrow field of application, it still has to be denominated, Davidson insists, a theory: «when a word or phrase temporarily or locally takes over the role of some other word or phrase (as treated in a prior theory, perhaps), the entire burden of that role, with all its implications for logical relations to other words, phrases and sentences, must be carried along by the passing theory.» (NDE, p. 443) Because the passing theory that provides access to first meaning is totally geared to the occasion, the boundary between «knowing a language» and «knowing our way around in the world generally» (NDE, p. 446) -- a boundary defended by Davidson himself in WMM -- is continually eroded in a constant, and constantly unconventional, process of modification. On this new account, communication does not succeed because of any fixed set of linguistic rules, but because we are always (re)learning and modifying our beliefs and modes of interpretation.

Having reconstructed this theory of intelligibility and communication, we can conclude this section by making a brief return to Davidson's reply (in CS) to the thesis of partial untranslatability. If, in any act of successful linguistic communication, speaker and hearer share a passing theory, then any differences of belief or interpretation can only be conceived as differences in relation to this more fundamental agreement. Those who defend the thesis of partial untranslatability conclude from it that differences in belief amount to differences between conceptual schemes. If, however, as Davidson argues in NDE, there is «no general principle, or appeal to evidence, [that] can force us to decide that the difference lies in our beliefs rather than in our concepts,» (NDE, p. 197) then the thesis of partial untranslatability is as philosophically unfounded as that of complete untranslatability.

1. 4 Inversion and Normativity

Davidson's conception of a prior and a passing theory provides an explanation of intelligibility in the extreme case of malapropism. In this section, I want to formulate -- on the basis of this conception, and on the basis of the conclusions drawn in CS -- a view of metaphor explicitly opposed to the one developed by Davidson in WMM. The first question that needs to be asked is whether metaphor can be said to possess meaning. In order to advance an affirmative response to this question I will seek to refute Davidson's two fundamental presuppositions in WMM: firstly, that of a dichotomy between language and world, which entails the exclusion of metaphor from language and understanding; and, secondly, that of a restrictive conception of literal meaning. Secondly, I would like to examine again the question of dead metaphor. A transition that was inconceivable in terms of the basic presuppositions of WMM -- that from living to dead metaphor -- will be seen to become easily comprehensible when looked at from the perspective of NDE. Thirdly and finally, I would like to ask whether a theory that maintains the necessary persistence of literal meaning (the first presupposition of WMM) can open up the possibility of a normative consideration of metaphor. The relation between literal meaning and first or intended meaning can serve as the basis, I want to argue, for such a consideration: the normativity of metaphor arises as a consequence of its being conceived as both understandable and constitutive of language.

The central feature of WMM is Davidson's exclusion of metaphor from the realm of language and understanding. As we have seen, this exclusion results in an epistemological dichotomy between literal, linguistic meaning and metaphorical, non-linguistic and worldly effects. Whereas language is formed by a set of literal meanings that can be learned, the world is the place where language is used by speakers. It follows from this dualistic account that, if language is constituted by conventions or rules that link words and meanings with things and facts in the world, then its structure must be that of a system; and that metaphor, conceived as operating outside the bounds of linguistic convention, must be said to lack any systematic meaning. The external world appears as a reality that, in order to be known at all, requires the imposition of the structure or order of a language. It also follows from this idea that -- whether it is maintained that there is only one conceptual scheme or as many as there are languages -- the world (or that which we know about it) is relative to the particular conceptual scheme in use. That is, whatever is real with respect to one conceptual scheme might not be real with respect to another. As we have seen, Davidson's article CS demonstrates the inconsistency of this epistemological division between language and world. This division is always articulated by means of the idea that one or many conceptual schemes organize or fit reality. Davidson successfully shows that this separation must imply a thesis of conceptual relativism or linguistic untranslatability, and he demonstrates that, in its two most common formulations, this thesis is indefensible. It can be sustained neither by the claim of linguistic intransitivity, nor by that of an untranslatability of incommensurable point of views, nor by that of an untranslatability tantamount to what Davidson terms the «third dogma of empiricism». Since Davidson's WMM relies on the same dichotomy proven inconsistent in all these cases, we can now affirm -- from a perspective worked out by the latter Davidson himself -- that its fundamental argument has to be rejected.

In WMM, meaning conforms to what I have termed a restrictive definition; it is conceived, that is, as something context-independent that is learned by convention. In NDE, by contrast, the definition of meaning is expanded so as to include within itself so-called non-conventional uses of language, such as malapropism and metaphor. Davidson here describes intelligibility as a communicative process that takes place between two participants: speaker and hearer. (His description must also be understood as valid, of course, for conversations in which numerous speakers and hearers participate). These roles can be understood as constantly interchangeable: speaker and hearer come to and initiate the conversation by means of a prior theory. This theory consists of a certain amount of information (beliefs about oneself, about the other, about the world, about the particular topic under discussion) and of communicative skills that enable mutual understanding. The interaction proceeds when the speaker and hearer express their beliefs and when each of them modifies their respective prior theory so as to incorporate new information. When speaker and hearer actually succeed in understanding one another, when a conversation «goes somewhere», it is because they coincide in a passing theory -- one which is ceaselessly undergoing revision and being adapted to the new information that becomes available. If the speaker uses a word or utterance that is not part of the hearer's prior theory, for example, he must be construed as having the intention and expectation that this expression be understood in the way that he intends. In the course of the conversation, therefore, or even at the same moment in which she utters the unfamiliar expression, the speaker provides the hearer with the necessary clues about how to interpret it. These clues, and the hearer's reaction to them, now form part of the speaker's own passing theory. In turn, the hearer -- who is in principle unfamiliar with the utterance -- also assumes that the speaker means something and that what she understands corresponds to this meaning. Because in the course of the conversation the speaker gives the hearer clues about how to interpret the utterance, the latter has the capacity to create her own passing theory. The meaning of the unfamiliar sentence is attained, therefore, when the meaning intended by the speaker converges with the meaning interpreted by the hearer. If, in WMM, meaning is understood to result from the learning of conventions, and metaphor is understood as having no meaning other than the literal meaning of its terms, then, in NDE, meaning becomes possible -- with regard to both literal and metaphorical utterances -- whenever speaker and hearer converge in a passing theory. Prior theories show us how to interpret utterances in advance; it is the only the process of communication, however -- the constant revision of beliefs and modification of meanings -- that provides us with the passing theory that enables us to understand novel utterances. Passing theories are always being discarded and recreated as communication proceeds. Nothing allows us to maintain that prior theories are necessarily shared: each participant enters into communication with different knowledge and skills. The only thing that is shared is, then, the passing theory, that is, meaning is only attained when communication succeeds, when the hearer understands what the speaker says. A passing theory is only valid on the occasion in which it is used; it cannot be fixed or detached from the context in which it originated.

In NDE, Davidson employs this theory of meaning in order to explain the linguistic use of malapropism, and in order to conclude that malapropisms possess linguistic meaning. What I want to show now, following Davidson's explanation, is that linguistic meaning can also be accorded to metaphor. With this in mind, I would like to refer back to the example from Donnellan discussed above, in order to advance a comparison of it with the case of metaphor. In this example, Jones utters that «Smith's murderer is insane»; although he is thinking of the wrong person, he nonetheless says something true. This example is both similar to and different from the case of metaphor. Jones does not realize that he is using a sentence with two meanings: his intended meaning being false, since he is thinking of the wrong person; and the first meaning of his sentence being true, since Smith's murderer is insane. In the case of metaphor we are also dealing with a sentence with two meanings, but in this case the sentence is uttered deliberately.Foot note 16 Metaphor consists of an intended first meaning -- which has to be understood as such by the hearer -- and a second meaning -- which also has to be understood as such, that is, as non-intended and literal. Davidson's explanation of the case of Jones can therefore be extended to metaphor so as to propose the following thesis: in metaphor, the dichotomy between first meaning as conventional and literal and second meaning as unconventional and non-literal is inverted. The hearer understands the first and intended meaning as such, even if it is unconventional and non-literal. To understand and accept a first meaning does not mean, however, to erase the second, non-intended meaning. For the second meaning -- commonly called literal -- is also understood by the hearer: it remains active. In contradistinction to Jones's statement, however, where first meaning and intended meaning did not correspond, in the case of metaphor, the hearer realizes the correspondence between intended meaning and first meaning. This explanation of metaphor maintains the distinction, then, between first and intended meaning, but calls for an abandonment of the dichotomy between the conventional and the unconventional.

It is possible to rephrase this explanation in terms of the distinction between passing theories and prior theories, and in terms of the dialogue between speaker and hearer. I want now to examine this process, firstly from the perspective of the hearer, and secondly from that of the speaker. The hearer's prior theory provides in advance the approximate meaning of the majority of expressions emitted in conversation. When the speaker utters a metaphor -- Max Weber's «politics is a strong and slow boring of hard boards»,Foot note 17 for example -- with the intention that it will be interpreted to mean that politics requires tenacity and steadfastness, it is quite possible that this meaning will not be provided by the hearer's prior theory. The hearer therefore adjusts this prior theory so as to accommodate all the information available to her; what the speaker previously said about politics, what both of them said about related issues, what the speaker seemed to say about, etc. Armed with all the facts and contextual information available to her, the hearer then formulates a hypothesis as to what the metaphorical expression could mean: either that politics is boring (mistakenly taking the metaphor for a pun), that it consists in drilling wood (mistakenly taking it literally), that it is a gradual, step-by-step process or that it requires a lot of effort, etc, etc. The hearer thus manages to interpret the speaker, given all the contextual information, in the way she thinks intended by the expression. This is the hearer's passing theory. As the conversation goes on, however, the hearer's view as to what the speaker intends might be transformed -- she might come to interpret the utterance as meaning, for example, that politics requires a charismatic character. This change is only possible because the hearer's passing theory is always geared to the occasion. If we now observe the same process from the speaker's perspective, we can see that, since her intention is to be understood, she tries to give -- taking into account her own view of what constitutes the hearer's prior theory -- all the information relevant for the understanding of the metaphor. When the speaker's and the hearer's passing theories converge, communication succeeds and the metaphor, rather than merely producing effects or suggestions, is understood.

On the basis of this analysis, we can also provide an explanation of how a living metaphor transforms into a dead one. Let us consider once again the metaphor «role». At a certain point it came to be used outside a theatrical context as a living metaphor. The hearer (or reader) of this metaphor had to realize that, in addition to the literal meaning of its words provided by her prior theory (that is, «part in a play»), the expression had to have -- in order even to make sense -- another meaning. This other meaning was the one intended by the speaker (or writer) who, most probably in the context of a discussion concerning social relationships, meant to suggest something like «feigned behavior», «proper function» or «expected pattern of action» etc. The hearer therefore acknowledged that this other meaning came first in the order of interpretation, and that the second (literal) meaning, although still active, was not intended by the speaker. It is precisely when this acknowledgment starts to occur immediately that the metaphor becomes effectively dead (although it can always be, of course, revitalized). Historically, for instance, the metaphor «role» has been repeatedly employed by different people both in a social and a theatrical context. Whenever somebody hears this metaphor today, her prior theory will tell her that, in a social context (in a sociology classroom, for example), «role» means «appropriate behavior»; and, in a theatrical context (when reading the list of actors in a program, for example), that it means «part in a play».

A number of other implications can be derived from this metaphorical reconstruction of Davidson's conception of prior and passing theories. We can extrapolate from Davidson's analysis, for instance, that in every process of communication the meaning of the words has to be distinguished from the speaker's intended meaning, and that both are equally constitutive of understanding. This article, for example, is for the most part written so that the meaning of the words roughly coincides with their intended meaning. Intended meaning and literal meaning can, however, be separated. Consider the following examples. When I use the word «first» in the expression «first meaning», its intended meaning in no way differs from its literal meaning: that which comes before others in a series or order. Even when the convergence between intended and literal meaning is obvious, this is only acknowledged in the context of its occurrence: it is intended here that «first» has a selecting function, not a temporal connotation (as in «the first female politician to gain power»). The literal meaning of the word «first», therefore, although it certainly belongs to the reader's prior theory, already presupposes an accommodation to the specific context of its use, that is, it already requires a passing theory in order to be understood. A passing theory is an unavoidable requisite whenever expressions appear the words of which do not correspond to the intended meaning given them. If a metaphor is employed, then the reader, in order to understand it, has to be able to distinguish between the literal interpretation of its words and their intended first meaning. If specific philosophical or political terms and expressions are employed, it is most probable that these words or expressions possess other meanings when used in other contexts and by other people. A field linguist, an English teacher, a multiculturalist and a Davidsonian, for example, would offer very different interpretations of Davidson's sentence «there is no such thing as a language». If the context is a quote from one of Davidson's articles (NDE, p. 445), then the interpreter will (hopefully) receive the necessary clues as to a successful interpretation with the help of my reconstruction of Davidson's article. These will allow her to modify her prior theory, to discern what Davidson intends the sentence to mean, and to decide whether she agrees with it or not. In each of these cases, the intended meaning must be distinguished from the meaning of the words that constitute the expression: Davidson does not want to claim that language does not exist, but that those traditional theories of language which claim that meanings are always furnished by conventions are wrong. Davidson's dynamic conception of meaning purports to claim that -- in the case of metaphor and other linguistic uses, as much as in the case of expressions that simply mean what their words literally mean -- communication can succeed. Communication succeeds when both speaker and interpreter are capable of distinguishing between the literal meaning of words and their intended meaning, of acknowledging when these two meanings coincide and of making clear (speaker) and ascertaining (hearer) the intended meaning when they do not coincide. When first meaning and intended meaning do coincide, there is no need for a complex mechanism of interpretation; all that is required is an implicit commitment both that the speaker means what the words that she uses says and that the reader acknowledges this. When first meaning and intended meaning do not coincide, speaker and interpreter have to modify their passing theories so as to render them convergent; these modified theories will then give them the new intended meaning. The literal and the metaphorical therefore belong to a continuum of meaning and use -- one in which every employment of language, even if it is transformed by its intended meaning, remains comprehensible so long as it answers to what the words literally mean.

It is possible to derive from this account a final insight into the minimal normative constraints inherent in all communication. For if a speaker, in order to communicate with others, has to commit herself to making her utterances understandable, she must employ words with a determinate literal meaning, and introduce enough contextual information so as to convey what she intends these words to say. To make herself understood, therefore, a speaker has to both abide by the usual meaning of words, and commit herself to showing how these meanings relate to the intended meaning introduced into them. A hearer, conversely, if she wants to recognize and understand what is being said, must abides by the literal meanings of the words and follow and ascertain those hints of the speaker that relate these words to their intended meaning. Understanding compels both an acknowledgment of the literal (usual) meaning of words and an endorsement of their intended meaning. (We can say the same thing when a new word, unknown to the hearer, appears in a conversation. This word's meaning will only be understood if the speaker provides enough clues with regard both to what the word literally means and is intended to mean, and if the hearer acknowledges that both of these meanings coincide. That both literal and intended meaning are constitutive of communication is revealed by the fact that, when a new word is introduced into a discourse and is intended as a metaphor, the hearer can in no way ascertain this metaphoricity unless she also understands the literal meaning of the word). Communication is subject to an inherent normativity: it necessarily depends both on the speaker's and hearer's commitment to the literal meaning of the words employed, and on their endorsement of the intended meaning as that which is actually meant. The inherent normativity of language is nothing more than the mutual presupposition of, and unavoidable commitment to, both literal and intended meaning; a presupposition and commitment to a sense of both their inextirpable difference and necessary connection. Without committing themselves to the literal, speaker and hearer simply cannot understand the usual reference of words; without committing themselves to the intended meaning in its necessary relation to the literal, speaker and hearer cannot understand one another on each new occasion of utterance.

If we accept, in contradistinction to the Davidson of WMM, that metaphor forms a part of language, then this account of the inherent normativity of language must of course be seen to compel an insight into the intrinsic normativity of metaphor. Metaphor is use of language in which the literal meaning of the words is related to, but does not coincide with, their intended meaning. Because the speaker wants to be understood, she commits herself to using words with a known literal meaning; but because she also intends to say something else with these words, she gives enough contextual information about how to relate this intended meaning to the literal meaning. And, from the hearer's perspective, metaphor is understood both because she recognizes the literal meaning of the words and because, with the help of the contextual information provided by the speaker, she acknowledges their intended meaning. It can consequently be maintained -- in contradistinction to conceptions of metaphor as non-constitutive of language, or as a deviant and merely ornamental use of language that can be paraphrased -- that metaphor is subject to a properly normativity. Like any other form of linguistic communication, metaphor makes use of literal meanings; and, when it is recognized as metaphor, it deliberately and explicitly formulates the speaker's intended meaning in a discursive manner. In order even to be understood, metaphor must respond both to literal meaning and its relation to intended meaning. As against Davidson's conception of it in WMM (as the mere producer of extra-linguistic effects), metaphor can be said to possess a meaning, and to be inherently normative, because of the relationship that it establishes between the literal and intended meaning of the words that constitute it. Once we have refuted Davidson's earlier thesis according to which metaphor has no meaning other than a literal one, we can therefore term the intended meaning of metaphor «metaphorical». Metaphor has an intended metaphorical meaning that can only be comprehended in relation to the literal meaning of the words that constitute it. The linguistic normativity of metaphor resides precisely in this relation.

Sonia Arribas
Instituto de Filosofía (CSIC)
<arribas1234 [at] yahoo [es] es>

[Foot Note 1]

Mary Hesse, Revolutions and Reconstructions in the Philosophy of Science, Bloomington: Indiana University Press, 1990.

[Foot Note 2]

Donald Davidson, «What Metaphors Mean», in Inquiries into Truth and Representation, Oxford: Clarendon Press, 1984, pp. 245-264.

[Foot Note 3]

There are numerous versions of both theories. For a contemporary version of the theory of «extended» literal meaning, see Nelson Goodman, Languages of Art, Indianapolis: Bobbs Merrill, 1968. The theory of elliptical simile dates back to Aristotle, Rhetoric, III, 1406 b, and Cicero, De Oratore, III, 38.

[Foot Note 4]

Gottlob Frege, «On Sense and Reference», in Philosophical Writings, ed. by M. Black and P. T. Geach, Oxford: Blackwell, 1962, pp. 56-78.

[Foot Note 5]

Richard Sennett, The Fall of the Public Man, London, Boston: Faber and Faber, 1986, p. 33. In the theatrical context, this metaphor originated from another metaphor: the «rôle» or «paper» was literally the piece of paper that contained the part which each actor had to memorize.

[Foot Note 6]

The Baroque origin of this conception of social and political life is explained in the chapter «Theatrum Mundi: teatro, máscara y escena política,» in José M. González García, Metáforas del Poder, Madrid: Alianza, 1998, pp. 97-142, and in Alexander Demandt, Metaphern für Geschichte, Sprachbilder und Gleichnisse im historisch-politischen Denken, Munich: Beck, 1978, chapter VI.

[Foot Note 7]

Donald Davidson, «On the Very Idea of a Conceptual Scheme», in Inquiries into Truth & Interpretation, Oxford: Clarendon Press, 1984, pp. 183-198.

[Foot Note 8]

Donald Davidson, «A Nice Derangement of Epitaphs», in Truth and Interpretation. Perspectives on the Philosophy of Donald Davidson, ed. by Ernest Lepore, Oxford, New York: Basil Blackwell, 1986, 1989, pp. 433-446.

[Foot Note 9]

T. S. Kuhn, The Structure of Scientific Revolutions, Chicago: University of Chicago Press, 1962.

[Foot Note 10]

P. F. Strawson, The Bounds of Sense, London: Methuen, 1966.

[Foot Note 11]

W. V. Quine, «Two Dogmas of Empiricism,» in From a Logical Point of View, Cambridge, Mass.: Harvard University Press, 1961.

[Foot Note 12]

A. Tarski, «The Concept of Truth in Formalized Languages,» in Logic, Semantics, Matamathematics, Oxford: Clarendon Press, 1956.

[Foot Note 13]

«Intended meaning» is the meaning that the speaker wants to give to her words, and which will be so understood by the listener.

[Foot Note 14]

Quoted by Davidson from Keith Donnellan, «Putting Humpty Dumpty Together Again,» The Philosophical Review, 77, (1968).

[Foot Note 15]

Jones is the character invented by Wilfrid Sellars to exemplify «philosophical behaviorism» in Empiricism & The Philosophy of Mind, Cambridge, London: Harvard University Press, intro. by Richard Rorty, study guide by Robert Brandom, 1956, 1997.

[Foot Note 16]

Only, that is, when it is realized to be a metaphor so interpreted.

[Foot Note 17]

Max Weber, «Politics as Vocation», in From Max Weber: Essays in Sociology, ed. by H. H. Gerth and C. Wright Mills, London: Routledge, 1948, 1991, pp. 77-128, p. 128.

Back to the SORITES portal