A Defence of the Indispensability of Metaphor

A Defence of the Indispensability of Metaphor Javier González de Prado Salas, Universidade Nova de Lisboa [email protected] *This is a penultimate draft of an article to appear in Philosophical Investigations. Please refer to the published version. Abstract I argue for the possibility of the thesis that metaphors are indispensable for grasping and expressing certain propositions. I defend this possibility against the objection that, if metaphors express propositions, once these propositions are identified they should be specifiable by nonmetaphorical means. I argue that this objection loses its strength if one adopts a Wittgensteinian, particularist view of thought, according to which grasping a propositional thought requires the ongoing exercise of a suitable skill often not characterisable by algorithmic rules. Within this particularist framework, thus, it becomes possible that metaphorical skills have an indispensable role in cognition and communication. I. Introduction In recent times, there have appeared several arguments supporting the view that metaphors can be used to express non-trivial propositional contents, either directly or indirectly.1 Many of these arguments rely on the role that metaphors play in our linguistic practices: we can use metaphorical expressions to make apparently non-defective communicative speech acts, in particular assertions. Moreover, speakers often endorse or reject what was claimed in a metaphorical utterance, as being true or false. These linguistic phenomena strongly suggest that there is something expressed or 1 Bezuidenhout (2001), Camp (2006b), Wearing (2006), Carston (2002), Stern (2006). communicated at least in some instances of metaphorical speech – although some remain unconvinced.2 In this paper, I will assume this is the case, without arguing for it. If this is granted, one can ask whether there are non-metaphorical ways of entertaining and expressing such contents. This is the question I will try to explore here. In other words, I will discuss whether there are intrinsically metaphorical propositions – propositions that may only be grasped and entertained by metaphorical means. Traditionally, this question has been connected with the possibility of paraphrasing metaphorical utterances: if there is some content that cannot be expressed but through metaphors, then literal paraphrases will not be available for those metaphors.3 My aim will be to argue for the possibility of what James Grant 4 calls the ‘Indispensability Thesis’ – that is, the view that ‘we use at least some metaphors to think, to express, to communicate, or to discover what cannot be thought, expressed, communicated, or discovered without metaphor’.5 I will defend a strong version of this thesis against the argument (presented by Grant and also found in the work of Elisabeth Camp6 and John Searle7) that, if metaphors express propositions, once these propositions are identified they should be specifiable by non-metaphorical means (in particular, one should be able to coin new terms which would be used to express such propositions in non-metaphorical ways). I will argue that this is a serious objection if one assumes a certain 2 Lepore and Stone (2010), Reimer (2001). 3 Of course, the alleged fact that metaphors cannot be paraphrased has typically been used to argue that metaphors do not express non-trivial propositional contents (Davidson (1978), Lepore and Stone (2010)). But, if we accept that metaphors do express some content, then the fact that such content cannot be paraphrased would seem to show that the content expressed by (at least some) metaphors cannot be expressed in non-metaphorical ways. 4 Grant (2010; 2103). 5 Grant (2013: 126). 6 Camp (2006a). 7 Searle (1969). conception of what it is to grasp propositional contents. However, this conception is not compulsory, and actually I will present a way of avoiding it. More specifically, I will argue that the Indispensability Thesis becomes much more plausible if one adopts a Wittgensteinian view of metaphorical interpretation, along the lines of Travis’s particularism8 (which, I will suggest, can be seen as a form of semantic relativism). The structure of the paper is as follows. I first present Grant’s argument against the indispensability of metaphor. Next, I try to unearth some of the assumptions of this argument. After that, I introduce an alternative relativistic framework which does not support these assumptions. Finally, I show how this alternative relativist view can be applied to the case of metaphors, offering a defence against the dispensability argument (and other related objections). II. The Indispensability Argument Let us grant that metaphorical utterances may express propositional contents – other than those yielded by their literal interpretation. If this is so, one may wonder why speakers would choose a metaphorical way of expressing such contents. One possible answer is that, in some cases, metaphors have some advantage over the literal alternatives. For instance, it could be that metaphors manage to express certain contents in a more economical way; or, perhaps, metaphors tend to achieve certain perlocutionary effects. Another possibility is that non-metaphorical alternatives are not available for entertaining and expressing some contents. In this paper, I intend to discuss this last possibility – the possibility that metaphors are indispensable for entertaining and expressing certain contents. In this Section, I will present an argument against such a possibility. I will call this argument the Dispensability Argument. I will focus on the formulation of the argument by Grant, 9 because he exposes the sort of argument in 8 Travis (2008). 9 Grant (2013: 146-148). Grant also offers several arguments against different specific proposals of an indispensable role for metaphors in cognition and language. I will not deal with these other which I am interested in a clear and explicit way. However, this argument is also suggested by Searle and by Camp10 – and its spirit lurks in much of the work on metaphor, and especially in the debate about paraphrase. 11 The Dispensability Argument goes as follows. Assume that certain metaphorical assertions express some propositional content. That is, they represent some way things may be. Let us say that a contentful metaphorical assertion characterises something as having some property (maybe things are more complicated, but for the sake of simplicity I will assume this unadorned view).12 Understanding a metaphorical assertion, therefore, will allow us to know what property is being attributed. But, once we know what the relevant property is, we can always create a new term which will be then used to express that same property in a non-metaphorical way. So, we will be able to attribute that property to things – that is, to express the same content that was expressed by the metaphorical utterance – without having to resort to the original metaphor. In this way, the metaphor is not indispensable to express that content. Note that this argument is quite strong, because it can be applied, in principle, to any sort of content one takes metaphors to express – no matter how rough, vague, open-ended or ad hoc such contents may be (in principle, it seems possible to coin ad hoc terms so as to denote ad hoc properties). This is how the argument is synthesized by Grant: Consequently, any metaphor that expresses something and is understood is dispensable. In the case of any such metaphor, we can identify the relevant properties, and therefore we can arguments here. 10 Camp (2006a: 16), Searle (1969: 20). 11 See Davidson (1978) and Lepore and Stone (2010). 12 Grant (2013: 90-1) acknowledges that metaphors are not always used to characterise things as having properties (e.g. metaphors can be used in questions or embedded contexts), but he focuses on characterising uses. For the sake of simplicity, I follow him in doing so. coin a non-metaphorical expression for the property. And if we can coin a non-metaphorical expression for the property, then we can use it to express whatever we expressed with the metaphor. 13 The essential point for Grant is that we are able to ‘identify’ the property expressed by the metaphor. It seems reasonable to say that identifying a property (at least in the sense relevant for Grant’s argument) means knowing how something is characterised when it is characterised as having that property. 14 That is, we have identified the property, in the relevant way, if we know what it is for something to instantiate it. The Dispensability Argument leaves room for a weak version of the Indispensability Thesis. According to this weak thesis, metaphors may be necessary for identifying some properties (or for enabling others to identify them). This view is defended, among others, by Camp.15 Grant, on the contrary, has serious doubts about it: although he admits it is a possibility, he is not optimistic about the prospects of finding good arguments in its support. 16 A stronger indispensability claim is that metaphors may be necessary for talking and thinking about certain properties even after such properties have been identified. Both Camp and Grant reject this strong Indispensability Thesis – which is ruled out by the Dispensability Argument. Whether metaphors are needed for an initial identification or not, Camp and Grant agree that, once a property is identified, one can just coin a new term which would be used to express the property without needing to appeal to any metaphor. The stipulated use of this new term would be that it is used to express (non-metaphorically) the property we had previously identified. Analogously, once 13 Grant (2013: 147-8). 14 This knowledge may be implicit or know-how. At any rate, identifying a property must not be confused with identifying some object as instantiating the property. 15 Camp (2006a). 16 Grant (2013: 148). we identify the property, we can have thoughts involving it (that is, we can think of things as having such property), without needing the mediation of any sort of metaphorical thinking. I will be concerned here with this strong Indispensability Thesis. Camp and Grant do not only, as we have seen, dismiss this thesis: they seem to take for granted that it is obviously wrong (in general, the thesis does not find many supporters in the literature).17 Grant says: to my knowledge, no advocate of indispensability does explicitly and directly claim that there are properties such that, even when we have identified them, we cannot then name them or coin a predicate or adjective with which to characterise18 something as having them.19 Camp is equally categorical when she affirms: It is of course true that after the speaker has gotten her hearer to identify the relevant property by metaphorical means, she can then introduce a new term which denotes it.20 17 Searle had already presented similar ideas when formulating his ‘Principle of Expressibility’ – according to which whatever can be meant can be said. Searle allows for the possibility that some language lacks words for expressing (literally) certain contents. But he argues that, whenever this happens, one ‘can in principle at least enrich the language by introducing new terms or other devices into it’ (1969: 20).Thus, Searle rejects the strong Indispensability Thesis. 18 Presumably, Grant means that the new term can be used to characterise non-metaphorically things as having the property (otherwise, it would not be much of an argument against the indispensability of metaphors). 19 Grant (2013: 148). 20 Camp (2006a: 16). So, according to Camp and Grant, metaphors might, at most (and unlikely, in Grant’s opinion), enable speakers to make a first identification of some property – so to say, the right of entry to thinking about that property would be provided by metaphor. But, at any rate, once the property is identified, the metaphor can be discarded – without expressive loss.21 Both Camp and Grant think that this is quite uncontroversial. However, in this paper I will try to argue against this conclusion, and against a hasty rejection of the strong Indispensability Thesis. Of course, the strong Indispensability Thesis is trivially false if metaphor is characterised simply as a figure of speech, in which some linguistic expression receives an interpretation different from its literal one. Certainly, this sort of figure of speech may be made dispensable by coining new terms without alternative interpretations. But there is no need to remain at such a superficial linguistic level when characterising metaphor. At least since the work of Lakoff and Johnson, it is common to treat metaphor as a cognitive, as well as linguistic, phenomenon, involving a distinctive way of thinking about some topic. 22 From this perspective, it becomes a possibility that, even when new terms are coined specifically to refer to some property, thinking about such a property keeps relying on the exercise of metaphorical cognitive skills. This is the possibility I will explore here. In the following Section I will examine the assumptions made by Camp and Grant when rejecting the strong Indispensability Thesis, and I will consider a view of what it is to identify a property in which these assumptions are actually compelling. Then, I will present an alternative picture that, however, makes such assumptions far more dubious. 21 Of course, as Camp argues (2006a: 14-6), the metaphor may still be needed to allow other speakers to identify the relevant property. But, after this is achieved, the metaphor would be dispensable for communication (once the participants in the conversation have identified the property, they can resort to non-metaphorical ways of talking about it). 22 Lakoff and Johnson (1980). Also, Gibbs (2011), Gentner and Gentner (1983). III. The Dispensability Assumption The argument against the strong Indispensability Thesis which I have presented above is based on the assumption that, even if non-literal (e.g. metaphorical) mechanisms are required to identify a certain property in the first place, once we have succeeded in identifying the property, we can go on to talk and think about it without needing to resort again to the non-literal mechanisms that would have allowed us to make the initial identification. In other words, even granting that an agent may need certain specific mechanisms or abilities so as to make a first identification of some property, when the property gets to be identified, the agent would be able to talk and think about it just in virtue of her competence as a literal speaker – that is, just by means of the general skills characteristic of mastery of literal discourse, without having to go back to those specific skills or mechanisms that were employed in the initial identification. I will call this assumption the Dispensability Assumption. This assumption may seem, at first sight, fairly plausible. So, for instance, I may identify and label one of the little pots in my kitchen as containing salt – instead of sugar – by tasting its contents. Once I have identified and labelled the pot, I can proceed to use it – to distinguish it from the sugar pot, cook with it, and so on – without needing to taste its contents anymore. The intuition behind the assumption is that identifying a property is something that can be done, so to say, once and for all. So, we would identify the property at some moment, and from then on we would be ready to think and talk about that property without having to re-identify it on each new occasion of use. The original identification would allow us to think and talk about the property in all possible situations that we may encounter. Another way of spelling out this intuition is to say that, when a property is identified for the first time by an agent, a suitable concept denoting it can be inscribed in her 'mental lexicon'. After that, the agent may think about the property just by tokening that concept, without having to reinscribe it each new time. This is compatible with the view that concepts are often constructed online, drawing from contextual information pertaining to a particular occasion of use. Although the relation between these sorts of ad hoc concepts and specific linguistic expressions may not be stable across different contexts, once a particular ad hoc concept is grasped by an agent, that ad hoc concept could in principle be encoded and retrieved so as to figure in her thoughts on new occasions – without having to re-construct the concept again each time. 23 Neither Camp nor Grant offer detailed reasons in favour of the Dispensability Assumption. They – as most people in the literature tend to do24 – just seem to take it as obviously right. The most I can do is to try, as charitably as I am able, to figure out what view of human language and thought makes this assumption so appealing. Then, I can try to propose alternative views which leave the assumption without much of its allure. This is what I will, in turn, do. Let me, then, start by examining what conceptions of property identification make the Dispensability Assumption look plausible and well grounded. One possibility that immediately comes to mind is that the learner manages to identify the property as a result of an explanation– as a result of grasping some rule for the identification of the property. For instance, a learner may be told that a bachelor is an unmarried man – so having the property of being a bachelor amounts to having the property of being a man and the property of being unmarried. The identification of one property would be related to the identification of some other properties. 25 Certainly, this seems a plausible way of identifying properties: if I know what it is to be red, and what it is to be a house, I will know what it is to be a red house: I will be able to identify the property associated with the complex ‘red house’ (in this case, the rule of identification is embodied by the compositional rules of the language). The ability to follow this sort of algorithmic rules of identification would be part of the competence of literal speakers. A Wittgensteinian worry would seem to lurk in the area, in the form of a regress of rules: we 23 See Carston (2010). 24 For instance, Searle (1969) when introducing the ‘Principle of Expressibility’. 25 This view can be related to decompositionalism, which characterises most concepts as complexes built from some basic primitives. See Vicente and Martínez-Manrique (2010). may demand further rules regarding the identification of those other properties that figured in the explanans (the learner may ask for additional rules for identifying what it is to be unmarried – which, in turn, may lead to yet further demands for explanations). However, this is not a worry that should concern us here, because we are assuming that the learner is already able to identify many other properties – those expressed by literal uses of predicates in her language. Remember that the learner we are interested in here is a competent literal speaker; what we want to find out is whether that is enough for identifying any property that might be expressed through metaphorical speech. The Dispensability Argument will apply nicely to properties which can be identified through such rule-like computations. After we had found the rule for the identification of the relevant property, we would be able to characterise things as instantiating it, without having to resort to metaphorical mechanisms (in general, without needing to resort to the mechanisms that lead us to the discovery of the rule in the first place). If it can be shown that the set of properties already expressible by literal speech provides us with enough resources so as to be able to identify any thinkable property through rule-like explanations, the Dispensability Assumption will be well grounded. This view of property identification, therefore, offers a strong motivation for the Dispensability Argument. But, is this the only conceivable way of identifying properties? Clearly not. For one thing, we have seen that this sort of rule-like identifications presupposes the previous identification of other properties. This suggests that at least some basic properties (which would be the base for further explanations) have to be identified by some other means.26 One obvious alternative possibility is that properties are identified through examples. Individuals would manage to identify a property after getting familiarized with a sufficient number of relevant instantiations of it. So, a learner would be shown several examples of red things and in that way she would find out which 26 Although it might be that, for any property, we could always find an identificatory explanation in terms of other properties. What is not possible is to give (non-circular) explanations for all properties at the same time property is picked out by the term ‘red’. In a similar way, Camp27 notices that many properties are singled out through ostensive identifications. The speaker would identify the relevant property by means of a demonstration, that is, by pointing to some relevant instantiation. For example, I can identify a particular tone of voice by pointing to somebody speaking in such tone (or by mimicking it myself).28 The temptation here is to say that the examples themselves determine what property is being exemplified. This would support the Dispensability Assumption: once we find a relevant set of instantiations associated with the property expressed by the metaphorical predicate, we can just introduce a new term which we will use to refer to the property specified by those examples. With the help of this new term, we will be able to express that property without the involvement of metaphorical mechanisms (even if we arrived in the first place at the identifying set of examples through metaphorical means, as Camp argues). But we know since Wittgenstein’s rule following considerations that this temptation should be resisted. Any given set of examples can be expanded in an indefinitely large number of alternative ways, always in accordance with some regularity. Thus, no series of instantiations of the property does fix, on its own, a unique and definite property. 29 It is necessary that those instances 27 Camp (2006a). 28 Camp (2006a: 11) further argues that, sometimes, the learner will not be in a position that allows her to get acquainted with the property through direct demonstrations. In these cases, metaphors may offer indirect access to an experience of the property, by relating it to relevantly similar experiences that are more familiar to the learner. This is why metaphors may be indispensable for identifying some properties (remember, this is what I called weak Indispensability Thesis). 29 So, for instance, imagine a speaker being shown many examples of red things, described to her as ‘red’: a red picture, a red car, a red umbrella and so on. Nevertheless, when she is shown a new object, say a red apple, she may insist that it does not instantiate the property picked out by the term ‘red.’ Maybe she interpreted that ‘red’ referred to the colour red in cars, pictures and are presented to a suitable learner who is able to develop a suitable discriminatory sensibility, in virtue of which she will be able to project the series to new cases in an appropriate way. This is to say, the learner will be able to know whether a new example is actually an instantiation of the property specified by the past series of examples. So, in any new situation in which the learner is to think or talk about the property, she will have to make use of the capacities – the suitable responsive sensibility – that allowed her to identify the property on past occasions. Therefore, when considered under a Wittgensteinian light, ostensive identification does not seem to offer motivation for the intuitions behind the Dispensability Assumption. Nevertheless, it could still be that this Wittgensteinian outlook only applied for certain basic or primitive properties, and that the Dispensability Assumption held for the identification of further properties, identified on the basis of these primitive ones. So, it could be that the properties expressed by metaphors can always be reduced, with the help of algorithmic computations, to properties already expressible by literal means. However, if the Wittgensteinian perspective introduced above is generalized, there emerges an alternative view on property identification – and, in general, on human cognitive and linguistic abilities –, which does not support the intuitions underlying the Dispensability Assumption. In particular, this view conceives most of the abilities involved in property recognition as nonalgorithmically decomposable. I turn now to discussing this alternative approach. IV. A particularist Alternative In the end of the previous section, it was suggested that individuals with a proper sensibility will be likely to identify the relevant property when presented with a sufficient number of relevant instantiations of it. Here, what is required from individuals is not necessarily that they grasp any specific algorithmic rule, but rather that they respond in a suitable way to the examples shown to umbrellas, but green for apples. Of course, this interpretation would be extremely odd in ordinary circumstances, but it is not ruled out by the examples; see Kripke (1982). them. It is expected that individuals with similar biological constitutions, and similar training, will eventually identify the same property from a series of paradigmatic instances. One may extend this point by saying that, in general, coming to appreciate some regularity, or learning to identify a property, amounts to mastering a suitable skill – to exercising the right responsive sensibility. In some cases, having the ‘right sensibility’ may consist in grasping some rule, or proceeding in accordance with some step-by-step algorithm. But, in other cases, what this ‘right sensibility’ is might not be enlightened at all by appeals to rules. Actually, it is one of the main tenets of both moral30 and semantic particularism31 that some complex human capacities are not properly characterisable in terms of general rules or principles – an idea already present in Wittgenstein’s thought. Charles Travis has applied thoroughly this Wittgenstenian perspective to the study of human cognition and communication. According to a possible interpretation of Travis’ particularism, what sort of thing counts as having a certain property is an occasion-sensitive matter, which cannot be settled without considering the circumstances surrounding the attribution of such a property. So, on each new occasion, in order to be able to assess whether some situation accords with the way the proposition describes the world (in order to decide whether a certain thing counts as instantiating the property attributed by the proposition), one would have to pay attention to features of the surroundings of the particular occasion of use. For instance, the question whether the proposition ‘The leaf on the table is green’ describes correctly a given leaf would receive different answers on different occasions for asking it: a brown leaf painted in green may count as green for the purposes of decorating my living-room, but not for a botanical study. A speaker is skilful enough – has developed a suitable sensibility – if she generally agrees with what competent individuals take as a reasonable understanding of a given utterance on a particular occasion. Crucially, Travis claims that, at least for most aspects of human thought and language, the 30 See McDowell (1998), Dancy (2004). 31 Travis (2008). sort of skill required to appreciate what is relevant and reasonable in a given situation cannot be codified by mechanical rules, since it is sensitive to contextual factors in an open-ended way. According to Travis, how the world is characterised by some proposition does not depend merely on some specific, narrowly constrained, and clearly defined contextual parameters (such as the time of utterance, or the individual uttering it): the influence of the surroundings is much more pervasive, and, no matter how many such narrow parameters we fix, there can always arise new contextual differences that would yield different understandings of the proposition expressed. There is an indefinitely large range of possible questions about how the world is being described by the proposition expressed, which would have to be answered by looking at features of the occasion of expression. MacFarlane has proposed reconstructing Travis’ particularism as a form of relativism.32 On relativist views, the truth of proposition does not only depend on the world of evaluation, but also on further parameters.33 For instance, on a relativist account of taste predicates, the truth of the proposition ‘Artichokes are tasty’ does not only depend on the world of evaluation, but also on the standard of taste with respect to which the proposition is assessed. In this way, that proposition can be true as assessed by James, who likes artichokes, but false from the assessing perspective of Mary, who hates them (even if James and Mary inhabit the same world). In a relativistic picture, there is no definite answer to the question whether a given proposition is actually true, in abstraction from a specific context fixing the value of the relevant parameters. In the same way, what counts as instantiating a certain property would depend on the contextually determined value of the parameters.34 The same object could count as instantiating a given property (e.g. tastiness) 32 McFarlane (2009). A similar reading of Travis’ particularism is offered by Predelli (2005). 33 See MacFarlane (2014), Recanati (2008), Kölbel (2015). 34 Different versions of relativism have different views about how to fix the parameters of the circumstances of evaluation. According to moderate forms of relativism, the relevant values for the parameters are those determined from the context of expression of the proposition (e.g. the when the issue is assessed from some context but not from another. In order to model Travis’ particularism as a form of relativism, a contextual parameter called 'counts-as' parameter is introduced in the circumstances of evaluation. This additional parameter would keep track of what counts, in each circumstances of evaluation, as being in accordance with the way the proposition describes the world (e.g., what counts as being a green leaf). So, the proposition ‘The leaf on the table is green’ will be true with respect to a context where the leaf on the table counts as green. Of course, the value of this parameter is not in general determined by any narrowly constrained feature of the context: being able to assess the count-as parameter requires mastering a suitable skill, rather than grasping some specifiable mechanical rule. The particularist views that I have presented – and which I have just characterised as a form of relativism – clearly clash with the Dispensability Assumption. According to these particularist views, being able to express and think about some property amounts to exercising a mature sensibility in a way that allows one to know what counts as having the property, on each particular occasion of use. And, crucially, the fact that I know what counted as relevant instances of the property on past occasions – the fact that I could identify the property on those occasions –, does not guarantee that I will be able to keep doing so in new contexts, if I stop exercising the proper sensibility which allowed me to succeed in those past identifications. Property-identification, therefore, would be an ongoing business: there is no point where the parochial capacities that lead the speaker to the correct recognition of a property can be sent to rest, without the risk of being unable to recognize the property at all anymore. No matter how many parameter of taste relevant in the context of utterance). By contrast, radical versions of relativism hold that the value of the parameters is fixed from the context of assessment of the proposition. Although nothing in my argument hangs on this, for ease of exposition I will assume that the parameters are fixed from the context of expression. MacFarlane (2009) calls this sort of view ‘nonindexical contextualism’; I follow Recanati (2008) instead in referring to it as moderate relativism. times in the past you have been able to tell what it is for something to have some property, there may always appear a new context of use in which your mature sensibility has to be recruited again so as to find out what is a relevant instance of the property in those circumstances. According to a relativist reconstruction of particularism, this is so because, in order to assess whether a given entity instantiates the property expressed by a certain predicate on a given occasion, it is necessary to be able to fix adequately the parameters relevant for such an assessment; and, on a particularist approach, being able to fix the value of the relevant parameters involves the exercise of a mature sensibility. Thus, a suitable sensibility needs to be at work not only when the property is first discerned, but all along the agent's thinking (and talking) about the property. If anything, identifying a property means becoming able to start exercising a suitable sensibility in order to think about the property on different occasions – rather than getting to a position where one can feel free to brush aside the exercise of such sensibility, as not relevant any more for entertaining thoughts about the property in question. My purpose in this Section has not been to offer a persuasive argument in favour of particularist views of human speech and cognition, but only to show that it is possible to develop such alternative views – and that they undermine the assumptions underlying the Dispensability Argument. In the next Section, I will show how these alternative views, when applied to metaphorical discourse, offer new strengths to the possibility that metaphors have an essential function in expressing and entertaining certain propositional contents. V. Particularist Metaphors According to the particularist approach I have presented above, having and expressing contentful thoughts requires mastering some appropriate skill. A further possibility is that, for some propositions, the skill required is some distinctively metaphorical skill. Since Lakoff and Johnson’s seminal work,35 it has become common to think that metaphor is characterised by the exercise of certain cognitive abilities, rather than merely being a linguistic device. In particular, I will follow Lakoff and Johnson in conceiving of metaphorical skill as the ability to make metaphorical projections or mappings across different conceptual domains. So, the possibility I want to put forward is that, for thinking about certain properties, the suitable sensibility whose exercise is required consists in the ability to make metaphorical mappings – which would vindicate the strong Indispensability Thesis. This mapping view applies straightforwardly to the sort of simple metaphor I am focusing on, that is metaphors with the form ‘x is F’, where x, which belongs to the target domain, is metaphorically attributed a property F from the source domain. However, the model can be easily generalized to cover other types of metaphors. As Lakoff and Johnson emphasize, metaphorical projections are primarily conceptual phenomena, and do not need to be directly reflected in the superficial form of the sentences used metaphorically. In particular, the target domain of the metaphor is not always explicitly mentioned in the sentence. Sometimes we have to rely on contextual clues in order to figure out what domain is targeted by the metaphor. For instance, in Sylvia Plath’s metaphor ‘Boarded the train there's no getting off’, it is the context provided by the rest of the poem that allows the reader to realize that the target domain is pregnancy and the changes it brings to the life of women. According to the mapping view, the metaphor invites us to think about this target domain by projecting our understanding of journeys and train travel (e.g. if you cannot get off the train, you cannot change your destination). Similarly, the mapping view can be extended to metaphors with non-declarative grammatical form. For example, in a metaphorical question, our grasp of what is being asked would rely on a projection of our understanding of some source domain. Likewise, in a metaphorical wish the object of the wish is characterized via a metaphorical projection from the source domain. Take Keats’ 35 Lakoff and Johnson (1980). Also, Lakoff and Turner (1989). metaphor ‘Oh, for a beaker full of the warm South.’ On a possible interpretation of this metaphor, what is being wished for is the emotional comfort brought by wine, and we are to think about such emotional comfort in terms of physical warmth.36 I will not discuss in detail here what metaphorical projections may exactly be. The general idea is that in metaphors we understand some concept in a target domain in terms of our understanding of some other concepts in a source domain. For instance, in the metaphor ‘The French teacher is a sergeant’ one describes a teacher by means of a projection of certain properties characteristic the military world (e.g. discipline, authority) – while other features of sergeants (e.g. being paid by the army) remain out of the scope of the projection. It should be stressed that I am not assuming that metaphorical projections are realized as durable links among neural networks,37 or that we should identify them with some type of causal process connecting cognitive states. The view I want to consider is more neutral and just holds that in metaphorical discourse our way of structuring our thought about some target domain is based on our way of thinking about a certain source domain. So, I am not presupposing a literal physical understanding of the notion of projection. In general, there does not need to be any specifiable rule or step-by-step algorithm that determines how a metaphorical projection between two domains is to be performed. It is well known that any two things are similar and dissimilar in many different respects, and that we can establish many alternative correlations and projections between any two structures. Understanding a metaphor is not just making one of the many possible projections, but appreciating which of these is the relevant one in that case. Thus, mastering metaphorical discourse involves being sensitive to 36 According to Lakoff and Johnson (1980), one of our basic ways of conceptualizing emotions is to think of affection as warmth. This metaphorical mapping would be grounded in the association between our embodied experiences of emotional affection and physical warmth. 37 As suggested by Lakoff (2008). what is salient in a given situation – in particular, what mappings, what correlations, are salient. And appreciating salience often implies exercising highly occasion-sensitive capacities that are not characterisable in a rule-like way. 38 The context-sensitivity of metaphorical mappings can be captured by a relativistic model of metaphor. According to such a relativist model, the extensional value of metaphorical propositions (their correctness or truth value) would be given by a function taking as arguments the world of evaluation and contextual information determining what projections are relevant. More specifically, the extensional value of metaphorical propositions would be relativized to a world of evaluation and a projection parameter – that is, a parameter determining which cross-domain metaphorical projection is the relevant one. Accordingly, a property expressed metaphorically would be associated with a function from worlds and a projection parameter to extensional value (sets of things). So, the proposition expressed by metaphorical utterances of ‘The French teacher is a sergeant’ is, on this relativistic view, true if the way the teacher is counts as a (contextually relevant) way of being metaphorically a sergeant – that is, if the relevant cross-domain projection of what it is to be a sergeant characterises the teacher accurately. In some context, the relevant way of being metaphorically a teacher may mainly amount to being harsh with discipline and punishments – so that these are the sorts of features more prominent in the projection from the source domain. However, in another context it may be that the emphasis is on the teacher being very authoritative and fond of giving orders – whereas being harsh with punishments plays a less dominant role (so that the weight of these features in the projection from the source domain is attenuated). In this partcularist approach, distinctive metaphorical skills (i.e. mapping skills) are required for using in one’s thought and speech the propositions expressed by metaphors. In order to know, on each occasion of use, what counts as according with the way the proposition describes the world to 38 Camp (2006a : 11). be (what counts as having the property attributed), one would need to be able to fix appropriately the projection parameter in the circumstances of evaluation – something that requires exercising suitable metaphorical skills. Note that what I have sketched is a particularist-relativist characterisation of metaphorical propositions. One can perhaps coin new terms intended to express such propositions, but, according to this particularist-relativistic view, metaphorical skills will still be needed to grasp them – to know how the world is being characterised by those propositions. For my specific dialectical purposes here, it is enough to show that this particularistrelativist model of metaphor is an open possibility and that it vindicates the (strong) Indispensability Thesis – since my target, the Dispensability Argument attacks the very possibility of such a thesis. Nevertheless, a few words can be said to motivate this proposal. It is often noted that many metaphorical utterances are difficult to paraphrase satisfactorily, despite seeming to express some content. On the other hand, the sorts of skills involved in metaphorical discourse (i.e. projection skills) are highly context-sensitive. So, perhaps the reason why metaphors are often especially difficult to paraphrase is because it is difficult to capture this context-sensitivity without resorting to the kinds of skills characteristic of metaphors. In the next Section, I show how the Dispensability Argument can be resisted if the particularist view of metaphor sketched above is adopted. VI. The Indispensability of Metaphor Revisited It has to be stressed that what I am proposing is not that in metaphorical thought one resorts to some metaphorical capacity in order to identify what property is being attributed in the target domain, so that once that the identification is accomplished, one can proceed to think of things as having that property – even if one stops exercising the metaphorical capacity that lead initially to the identification. This would be perfectly compatible with the Dispensability Argument. On the view I am defending here, in contrast, the suitable responsive sensibility – the capacity to appreciate salient projections – has to be exercised all along our thinking about the property. These are just the same points that were made in the particularist views described above. The only peculiarity now is that the required skill is the capacity to make relevant metaphorical projections between different conceptual domains. In this way, my proposal does not reduce metaphorical speech to other sort of context-sensitive discourse, as some contextualists try to do: 39 in metaphorical speech, it is essential that the ability whose exercise allows us to understand the metaphor involves making relevant mappings across domains. This would not happen in other cases of context-sensitivity, such as loose talk or quantifier domain restriction – or, in general, in the sort of occasion-sensitive but non-metaphorical language examined by particularists like Travis. Admittedly, I have not presented arguments to the effect that the context-sensitive mechanisms involved in metaphor cannot be reduced to the mechanisms involved in other kinds of contextsensitive discourse. I am only suggesting that it is possible that metaphor involves its own peculiar context-sensitive mechanisms.40 In order to make my proposal clearer by way of contrast, I will present an example of the view I am not defending. Let us suppose that when we say, metaphorically, that somebody is cold, the property we are actually attributing to that person is some peculiar combination of unfriendliness and lack of emotional expressiveness. Now, imagine that we do not have a word in our language to refer – literally – to that property. All the same, once we locate that peculiar combination of unfriendliness and lack of emotional expressiveness – once we are able to tell what sort of person is a cold person –, we can coin a new term that can be used to express such a property in a literal way – say, ‘emo-cold’.41 Then, we will be able to think and talk (resorting to the new 39 Wilson and Carston (2006), Wearing (2006). 40 See Gibbs and Tendahl (2006). 41 If you think that metaphors express rough and open-ended contents, this new term would be used to express literally such rough and open-ended contents (some literal utterances seem to call for an open-ended interpretation (see Camp 2006a). term ‘emo-cold’) about that property without engaging in metaphorical speech or thought. The metaphor will have been successfully decoded.42 Note that this would be possible even if we needed to engage in metaphorical mappings (between the domains of physical temperature and emotional features) in order to single out the property we now refer to as ‘emo-coldness’. The view I am arguing for here is different. According to my proposal, as a general rule we do not use metaphorical projections just to single out some property, which can be then talked and thought about without returning to such projections. In genuine metaphorical expressions, we have to resort continually to the relevant metaphorical projections in order to see how those expressions would be correctly applied in different circumstances of use. So, granting that thinking of emotional features in terms of temperature is a genuine metaphor – and not a lexicalized one –, if we want to carry on recognizing what it is for a person to be cold in each new context of use, we will have to engage in metaphorical mappings, because these mappings are the only way we have of knowing what, on a given occasion, is for a person to be cold. Perhaps we can talk about the property of being emotionally cold by using linguistic expressions in a way that is not superficially metaphorical (e.g., by coining new terms); nevertheless, according to the proposal I am considering, such use of language would be underpinned by the exercise of metaphorical mapping abilities. On this view, therefore, it is not in general possible to find some property that is identical – in every context – to the property attributed by the metaphor, and that – once it is identified – we know how to express and think about without the help of metaphorical skills. The reason for this is that, in accordance with the particularist insight, there can always arise questions as to what counts as having that property on new occasions of use – in alternative contexts, in extensions of the metaphor. And, in the case of metaphors, these questions have to be answered by making a relevant metaphorical projection. The view I am presenting, therefore, is committed to the claim that – at least in some cases – the contents expressed by active metaphors cannot be expressed by subsequent 42 I do not go on to say that a decoded metaphor is a dead metaphor because the metaphor may also have a non-propositional component – for instance, an imagistic component. dead metaphors (by expressions that can be understood without resorting to metaphorical mechanisms). To be sure, even if one does not exercise metaphorical skills, one can just introduce a new term designed to denote whatever property is expressed by utterances of some metaphor. Imagine that, after identifying in an initial context the property expressed by a metaphor, I stop resorting to the relevant metaphorical abilities but, still, I coin a term (‘M’) intended to refer to such a property (or perhaps I refer to it by means of the identifying description ‘The property expressed by that metaphorical utterance’). However, as I am not exercising the required (metaphorical) skills anymore, I will not be able to figure out what counts as instantiating the property on new occasions of use. Insofar as I will not know to what sorts of things ‘M’ (or the identifying description) is correctly applied, I will not know what property I am trying to talk or think about. Resorting to the term ‘M’ (or to the identifying description) will no put me in a position to think and talk, in a full sense, of things as having the relevant property – unless I exercise the required metaphorical skills. As I see it, there is an intuitive difference between the way in which one can think and talk about a given property when one knows what it is for something to have that property and when one does not. In order to buttress this point, let me present a parallel situation. Imagine that I do not know the (literal) meaning of the term ‘dilettante’ (and I am not familiar with the concept it expresses, either), but I hear Mary use the term so as to characterise someone. It will still be possible for me to refer to the property expressed by Mary’s use of ‘dilettante’, by means of the identifying description: ‘The property expressed by Mary’s use of “dilettante”’. However, it seems that appealing to this description is not enough to put me in a position to think of someone as being a dilettante, at least in a full sense: I still do not know what sort of person counts as being a dilettante. Clearly, there is a significant difference between the role that the concept of being a dilettante can play in Mary’s thought and speech and the role it can play in mine. Mary can think and talk about someone being a dilettante in a way that is not available to me. On the view I am presenting, something analogous would happen to someone who, lacking the relevant metaphorical skills, tried to refer to the property expressed by some metaphor by coining a new term or by resorting to the description ‘The property expressed by that metaphorical utterance’. She would not be able to think and talk about the property in the same way as someone who possesses the relevant metaphorical skills. The strong Indispensability Thesis is concerned with this sort of difference. A further possible objection is that I have, at best, managed to show that thinking and talking about certain properties requires the intervention of some mechanisms or skills (projections between domains), which are also required for the understanding of some metaphor. But, it may be argued, this would not show that the metaphor itself is necessary for thinking and talking about that property.43 However, in my proposal the skills required for thinking about the property would be the sorts of skills that define what metaphor is (i.e. a projection between different conceptual domains), so exercising such skills would amount to exercising metaphorical skills. One could try to reply that this sort of projection skill is not actually unique to metaphor: for instance, one might argue that such projection skills ultimately amount to the ability to recognize relevant similarities, and that they can be exercised in similes or comparisons, without the involvement of metaphors. Although there is a long tradition that accounts for metaphor in terms of similarity,44 such views are heavily contested in the contemporary literature. 45 In particular, in the framework I am presenting, the skills involved in metaphor and simile are clearly different. In similes, we would compare the properties of two different topics. In metaphors, on the other hand, we would resort to a projection from a source domain in order to determine what is to count as instantiating a certain property in the target domain. This projection does not need to involve the recognition of 43 Grant makes this point in several places (e.g. 2013: 134). 44 For instance, Fogelin (1988). 45 See, among others, Glucksberg and Haught (2006), Camp (2006a), Searle (1979), Lakoff and Johnson (1980), Lycan (1999). similarities or shared properties (for instance, it may be based on correlations). Think, for example, of the metaphor ‘John is very cold: he is a block of ice’: although John shares many properties with cold things, none of them seems to be relevant for the understanding of the metaphor.46 In particular, an emotionally cold person may be physically warmer than an affectionate one. Of course, we can say ‘John is like a block of ice’. But, arguably, the understanding of this simile presupposes the understanding of the associated metaphor – by determining what it is to be a cold person, the metaphor creates the relevant similarity. I do not pretend to have provided conclusive arguments against the view that metaphors can be explained in terms of similes and comparisons47 – the comments above only aim at offering some motivation for remaining sceptical towards such views. At any rate, even if it were granted that the relevant analogical skills are not distinctive of metaphor, the discussion in this paper would still be interesting, because a version of the Dispensability Argument can be directed against the (strong) indispensability of similes or analogical skills. At the very least, the considerations in this paper would offer a reply to such an argument. In addition, as I have tried to show above, it is plausible to think that there are certain mapping skills which are idiosyncratic of metaphor – and that such skills may be indispensable for entertaining and expressing some propositions. VII. Conclusions In this paper, I have offered reasons for resisting the Dispensability Argument, according to which once we identify the property expressed by some metaphor we can think and talk about it without further appeal to metaphors (in particular, we can coin a new term that will be used literally to refer to that property). Although I have focused on this particular argument, my defence of the indispensability of metaphors can also, I think, provide new replies to related challenges hinging on 46 See Searle (1979). 47 For a defense of this sort of view, see Grant (2013: 115-124). Due to space limitations, I am not able to deal with Grant’s points in detail. the assumption that, if metaphors expressed propositional contents, we should be able to somehow specify such contents. 48 For instance, the sort of view I have sketched explains why in many cases it is difficult to offer satisfactory literal paraphrases of metaphorical utterances that, nonetheless, seem to express contents. It has to be stressed that, even if I have been successful, I will not have shown that metaphors are actually indispensable for entertaining and expressing certain propositional contents, but only that such a possibility cannot be ruled out a priori – as the Dispensability Argument purports to do. There are some assumptions in my arguments that I have not proved. First, I have taken for granted that metaphors can express contents; moreover, it is an open question whether metaphorical discourse cannot be successfully characterised in terms of rules; finally, it also an open matter whether the alleged context-sensitivity of metaphorical discourse cannot be dealt with by appealing to the mechanisms that allow us to understand other context-sensitive, but nonmetaphorical, utterances. Answering these questions is an ambitious task that lies beyond the scope of this paper.49 Nova Institute of Philosophy (IFILNOVA), Universidade Nova de Lisboa Av. de Berna, 26, 48 49 For instance, Lepore and Stone (2010). For their comments and feedback, I thank Daniel Whiting, several anonymous referees and audiences at UNED, the University of Santiago de Compostela, the University of Connecticut and the European Society for Philosophy and Psychology 2013 conference at Granada. This work has been supported by the Portuguese Foundation for Science and Technology Project UID/FIL/00183/2013 (“Values in Argumentative Discourse”) and the Spanish Government research project FFI2014-57258- P (“Normative inferences and interferences in scientific research”). Lisbon 1069-061 Portugal · [email protected] References Bezuidenhout, A. (2001). “Metaphor and what is said: A defense of a direct expression view of metaphor.” Midwest Studies in Philosophy 25: 156-186. Camp, E. (2006a). “Metaphor and that certain ‘je ne sais quoi’.” Philosophical studies 129: 1-25. Camp, E. (2006b). “Contextualism, metaphor, and what is said.” Mind & language 21: 280-309. Carston, R. (2002). Thoughts & Utterances: The Pragmatics of Explicit Communication. Oxford: Blackwell. Carston, R. (2010). “Lexical pragmatics, ad hoc concepts and metaphor: from a relevance theory perspective.” Italian Journal of Linguistics 22: 153-180. Davidson, D. (1978). “What metaphors mean.” Critical inquiry 5: 31-47. Dancy, J. (2004). Ethics without principles. Oxford: Clarendon Press. Fogelin, R. (1988). Figuratively Speaking. New Haven: Yale University Press. Gentner, D. and D.R. Gentner. (1983). “Flowing Waters or Teeming Crowds: Mental Models of Electricity.” Cognition 7: 155-170. Gibbs Jr., R.W. (2011). “Evaluating conceptual metaphor theory.” Discourse Processes 48: 529-562. Gibbs Jr, R.W., and M. Tendahl. (2006). “Cognitive effort and effects in metaphor comprehension: Relevance theory and psycholinguistics.” Mind & Language 21: 379-403. Glucksberg, S. and C. Haught. (2006). “On the relation between metaphor and simile: When comparison fails.” Mind & Language 21: 360-378 Grant, J. (2010). “The dispensability of metaphor.” British Journal of Aesthetics 50: 255-272. Grant, J. (20139. The Critical Imagination. Oxford: Oxford University Press. Guttenplan, S. (2005). Objects of Metaphor. Oxford: Clarendon Press. Kölbel, M. (2015). “Relativism 2: Semantic Content.” Philosophy Compass 10: 52-67. Kripke, S.A. (1982). Wittgenstein on rules and private language: an elementary exposition. Cambridge, MA: Harvard University Press. Lakoff, G.J. (2008). ‘The neural theory of metaphor.’ In R.W. Gibbs (ed.), The Cambridge handbook of metaphor and thought. Cambridge: Cambridge University Press, pp. 17-38. Lakoff, G.J. and M. Johnson. (1980). Metaphors we Live by. Chicago: University of Chicago. Lakoff, G.J. and M. Turner. (1989). More than cool reason: A field guide to poetic metaphor. Chicago: University of Chicago Press. Lepore, E. and M. Stone. (2010). “Against Metaphorical Meaning.” Topoi 29: 165-180. Lycan, W. (1999). Philosophy of Language. A Contemporary Introduction. London: Routledge. MacFarlane, J. (2009). “Nonindexical contextualism.” Synthese 166: 231-250. MacFarlane, J. (2014). Assessment sensitivity: Relative truth and its applications. Oxford: Oxford University Press. McDowell, J.H. (1998). Mind, value, and reality. Cambridge, MA: Harvard University Press. Predelli, S. (2005). “Painted Leaves, Context, and Semantic Analysis.” Linguistics and Philosophy 28: 351-374. Recanati, F. (2008). “Moderate Relativism.” In M. García-Carpintero and M. Kölbel (eds.), Relative Truth. Oxford: Oxford University Press, pp. 41-62. Reimer, M. (2001). “Davidson on Metaphor.” Midwest studies in philosophy 25: 142-155. Searle, J. (1969). Speech Acts: An Essay in the Philosophy of Language. Cambridge: Cambridge University Press. Searle, J. (1979). Expression and Meaning. Cambridge: Cambridge University Press. Stern, J. (2006). “Metaphor, literal, literalism.” Mind & language 21: 243-279. Travis, C. (2008). Occasion-sensitivity: Selected essays. Oxford: Oxford University Press. Vicente, A. and F. Martínez-Manrique. (2010). “On relevance theory’s atomistic commitments.” In B. Soria and E. Romero (eds.), Explicit Communication: Essays on Robyn Carston’s Pragmatics. London: Palgrave, pp. 42-58. Wearing, C. (2006). “Metaphor and what is said.” Mind & language 21: 310-332. Wilson, D. and R. Carston. (2006). “Metaphor, relevance and the ‘emergent property’issue.” Mind & Language 21: 404-433.