Go Figure: understanding figurative talk

We think and speak in figures. This is key to our creativity. We re-imagine one thing as another, pretend ourself to be another, do one thing in order to achieve another, or say one thing to mean another. This comes easily because of our abilities both to work out meaning in context and re-purpose words. Figures of speech are tools for this re-purposing. Whether we use metaphor, simile, irony, hyperbole, and litotes individually, or as compound figures, the uses are all rooted in literal meanings. These uses invite us to explore the context to find new meanings, new purposes, beyond the literal. Each employs different mechanisms to bridge the gap between what is said and meant. But the overall question that figures raise is what meaning is. Philosophers and linguists have focused on how to define figurative meaning. Is it in the same general line as literal meaning, or rather a special kind of meaning? This inquiry has taken the form of more specific questions:

• 1. The question of the output: Is figurative meaning a form of speaker-meaning (part of what a speaker inferably intends by her words), or not a matter of meaning at all?

• 2. The question of the input: What linguistic constructions are conducive to figurative meaning?

• 3. The question of the mechanisms: What kind of processes underlie our grasp of figurative meaning? Are they part of semantics or pragmatics, or are they more general cognitive mechanisms (e.g. signalling make-believe, expressing affect, developing imagery, play-acting)?

These questions have typically been approached as separate projects. This special issue aims to show how their alignment offers fertile ground. The majority of papers collected were initially presented at the ‘Go Figure’ workshop, held at the Institute of Philosophy, London, in June 2013. The contributions are grouped here based on two distinct families of figures (M-type and I-type):¹ with one group on metaphor and related figures, by Hills, Green, Camp, Magidor; the other group on irony and related figures, in particular understatement and overstatement, by Horn, Walton, Barker; and one final contribution on compound figures such as ironic metaphor that combine figures from each group, by Popa-Wyatt. Here I offer a rapid overview of the questions addressed, and how the papers move the field forward.

The title echoes Steve Yablo’s (2001) paper ‘Go Figure: A Path through Fictionalism’. In metaphysics we are fictionalists about a particular mode of discourse (e.g. arithmetic) when we think that it is best understood as propounding myths or fictions about one sort of entity (e.g. numbers) as a means of making serious assertions, conjectures, and arguments of entities of other less problematic sorts (e.g. concrete physical objects). There have been diverse ways of working out the fictionalist suggestion, yet most prove to be incoherent or self undermining on close examination. Yablo argues that an especially promising way of working out the fictionalist idea, relative reflexive fictionalism, is best understood as a kind of figuralism playing up the analogies between the myth mongering it sees as going on in mathematics and the myth mongering that manifestly goes on in the unselfconscious use of figures of speech. We should be fictionalists about some domain X, Yablo says, if ‘when we examine X-language in a calm and unprejudiced way, it turns out to have a lot in common with language that is fictional on its face. If one now acts which elements of everyday language are fictional on their face, the answer is the figurative elements’ (87).

In aesthetics we are fictionalists about a given cultural activity when we think it is best understood as a form of make believe, propounding myths or fictions by using actions and artifacts and existing natural objects and circumstances to prescribe what participants in that activity are called on to imagine. Ken Walton and David Hills are fictionalists about metaphor and kindred forms of figurative language (Walton (1990, 1993), Hills (1997)). They take very seriously Yablo’s idea that figurative elements of ordinary language are fictional on their face and put it to work in accounting for the expressive and suggestive powers of these figurative elements. They try to explain the metaphorical truth of utterances, the metaphorical applicability of terms, and the like in terms of what it takes in particular games of make believe to make something fictional, that is, to make it part of what we are called on to imagine there.

Here Hills elaborates and refines the fictionalist account of metaphor in his (1997). Precisely what needs to be actual in order to render a given thing fictional depends on the special rules, the special principles of generation, governing one or another particular make believe game. Humans are good at playing such games on a pickup basis, putting appropriate rules in place and governing their imaginings accordingly on the fly. The rules are a bit different in their details each time we play. But they usually provide somehow for each of four familiar broad modes of generation. We spell things out when we render things fictional by saying them in a spirit of stipulative pronouncement. We carry things over when we call attention to beliefs or assumptions familiar from prior contexts so as to render them fictional in the present one. We work things out when we add things to the content of a fiction-in-the-making by calling attention to ways of inferring them from things already fictional by other means. We act things out when we render things fictional by arranging to have them enacted before our eyes by appropriate props which have what it takes to assume various make believe roles, props whose actual properties and relations serve to represent fictional properties and relations belonging to the things they have come to portray.

When games make significant use of enactment, the playing of them often requires signals by means of which individual players express their defeasible personal working understandings as to the rules of the game, what’s fictional in the game so far, and how what’s fictional under the rules came to be fictional under them in the first place. Such signals mingle role concepts and role player concepts in various characteristic patterns. The simplest and most familiar of these patterns pairs the expression of a role player concept (e.g. the concept of Juliet) and an expression of a role concept (e.g. the concept of the sun) by means of an appropriate form of the verb ‘to be’—e.g. ‘Juliet is the sun’—so as to signal that in the opinion of the signaler, Juliet has what it takes for the role of the sun to fall to her in the game at hand. In contexts where listeners can be credited with a good independent feel for the rules of the game and for significant parts of what’s supposed to be fictional in the game, such a signal will be richly and elusively informative about Juliet herself. Hills’s thought is that an ‘A is B’ metaphor like ‘Juliet is the sun’ can be viewed as a make believe signal of precisely this sort, put to predominantly prop-characterizing use in the context of a game Romeo called into being in the first place for predominantly prop-characterizing reasons. In the case at hand, the fictional content Juliet helps to enact is a sunrise: ‘Juliet’s swimming into view serves to render it fictional that the obscurity and sterility of night is replaced by the lucidity and fecundity of day, the previously impressive light of the moon (Rosaline) becomes pale and insignificant in the presence of the new and greater light of the sun, … and so on’. This is one way to implement and think through Walton’s suggestion that verbal metaphor and metaphor-like manoeuvres in nonverbal media are a matter of prop-oriented make believe.

Hills sees precedents for his way of thinking about metaphorical interpretation in Aristotle’s discussion of ‘bringing before the eyes’ in the Rhetoric. And he responds to ambitious attacks on fictionalism from Camp (2009) and Wearing (2012), contending that such critics neglect the distinction between make believe signals on the one hand and direct make believe stipulations (spellings out) on the other, and have overly simple views of the roles of similarity (or analogy) and attention (and inattention) in successful make believe activity.

A different role for imagination in metaphor, in the sense of imagery, is explored by Mitchell Green. Imagery is a key concept in Davidsonian non-cognitivist accounts of metaphor. It’s the experience of seeing one thing as another that enables us to think of something or someone in a new light. This experience is critical for Green in eliciting affect and empathy. In contrast to Davidsonians, who deny that metaphors can have any meaning, Green wants to maintain a tie with cognitivism by allowing metaphorical meaning. To do this, he tries to establish a middle ground between cognitivism and non-cognitivism.

His strategy is to distinguish between metaphors that ‘demand’ imagery to be understood—‘image-demanding metaphors’ (IDMs), and ones that merely permit it—‘image-permitting metaphors’ (IPMs). IPMs are everyday conversational metaphors where the metaphorical content is put in the service of illocutionary acts such as assertions, questions, orders, and the like. Imagery here plays a minimal role, merely helping the hearer grasp the speaker-meaning. Green follows other cognitivists in explaining IPMs in terms of ‘semantic content, speaker meaning, and satisfaction conditions’. However, he differs in that he requires that metaphorical content not be reduced to expressing a proposition. He calls his position ‘narrow cognitivism’.

Green’s interest, however, is with IDMs. He argues that IDMs are special because they are characteristically used in the service of ‘self-expression’. In expressing oneself, one shows how one thinks or feels, or what an experience is like, thereby enabling others to empathise. Self-expression, however, is not just manifesting what is inside. Instead, it’s an ‘expressive behaviour’ that is ‘designed’ through evolution or culture to both show and signal what a mental state feels like. One can thus express a state, even when one lacks that state, because all is required is engaging in a behaviour that is expressive of that state. This is how IDMs work for Green. They function as expressive signalling that is designed to display the speaker’s mental states though images. This does not require that she intend to show how she thinks or feels, let alone reflexively intend to do so. Thus, IDMs are not a matter of speaker-meaning. They are instead vehicles for prompting insight and reflective imagination by encouraging hearers to vividly keep in their mind images prompted by the literal meaning of the words uttered. Such expressive-supporting imagery is what makes IDMs excellent tools for eliciting affect and experiential responses in hearers. By imagining themselves in the speaker’s emotional situation, they can empathise with the speaker and assess her affective state for its aptness to the situation to which it is a response.

Green argues that in spite of important differences between IDMs and IPMs, they nevertheless share a similar cognitive profile. Though IDMs are fundamentally self-expressive, this does not prevent them to serve as vehicle for illocutionary acts, to lie, or to banter, in the same way IPMs do. However, for Green, metaphorical banter is not a dispute about a determinate metaphorical content that one speaker asserts, and the other denies. Instead it concerns the aptness of the speaker’s mental states in the situation to which the metaphor is a response. For example, a rich evocative metaphor like Groening’s ‘Love is a snowmobile racing across the tundra and then suddenly it slips over, pinning you underneath. At night, the ice weasels come’, may allow one to respond: ‘No, it’s a ride in a Goth amusement park at the end of which you get dropped into a vat of boiling oil’. Here the disagreement does not require that the two speakers coordinate on a speaker-meant propositional content. Rather, they disagree about the aptness of the emotional response, and the imagery evoked by the situation described. This manoeuvre enables Green to establish a ‘wide cognitivism’ which accommodates both the expressive nature of IDMs and the cognitive profile characteristic of IPMs.

In her paper, Elisabeth Camp stresses the role of seeing as in metaphor interpretation, and the various non-propositional effects it affords. Like Green, she too wants to maintain a tie with cognitivism, but unlike him, she believes that speaker-meaning-based accounts have the resources to accommodate metaphors’ non-cognitive, non-propositional import. Here she focuses on metaphorical insults such as ‘Chairman, you are a Bolshevik!’. She argues that what makes such metaphors powerful rhetorical weapons to offend is their ability to cultivate a denigrating way of seeing and thinking about someone (or something). This negative way of thinking of someone makes hearers complicit, thereby making it harder for them to resist and reject ways of thinking they wouldn’t otherwise inhabit. Yet this is not the work of some ‘weird imaginative brainwashing’. Camp shows instead that a combination of familiar mechanisms—perspectives, presupposition, and pragmatics—which are pervasive in literal communication is also critical to explaining the complicity and anti-deniability of metaphorical insults.

Perspectives are, for Camp, dynamic tools for thinking. They bind together many features in a holistic way of thinking about a subject, so that prominent features stick out in our minds, while less salient ones fade into the background. Metaphors use such perspectives to frame one thing in terms of something else. But the thing that is doing the framing and the thing being framed, also involve complex conceptual structures, which Camp calls ‘characterizations’. Characterizations are mental structures that are informationally rich, often affectively-laden and experientially vivid, which raise to prominence properties that are fitting rather than actually instantiated. Characterizations structure the properties they bind together along two dimensions—prominence and centrality—so the overall significance of basic features in a characterisation depends upon the structure of the whole.

What work do characterizations and perspectives do in metaphor? In metaphors of the form a is F, Camp argues, we employ one characterisation of F in order to frame or structure our thinking of the subject a. This structuring seeks to find matches for prominent and central F-features within the characterization of a, and raise the prominence and centrality of those matched a-features. For example, ‘George is a tail wagging lapdog of privilege’ prompts the hearer to find the most prominent features in the characterisation of ‘tail-wagging lapdog of privilege’—e.g. growing up in swanky surroundings—and introduce this matching property as part of what is conversationally warranted about George. This yields a restructured characterisation of George which makes us see him in a new light, say, where privilege-features are foregrounded, and other less central features are backgrounded.

How does this help explain the complicity and anti-deniability of metaphorical insults? Camp shows that perspectives have several features that make them apt for the job. First, they are cognitively irresistible: they structure our thoughts in a way that ‘intrude on us unbidden’. We reconfigure our thinking about the subject by moulding our mind to the speaker’s denigrating perspective, and once we’ve seen the subject in that objectionable way, it’s hard to un-see. We feel tainted in doing something we shouldn’t, and even if we were to repudiate specific features of the perspective, the overarching organisational structure remains untouched.

Why are metaphorical insults hard to combat and repudiate? This has to do, Camp argues, with the fact that perspectives are merely presupposed. The objectionable perspective under which the speaker frames the subject is presented as taken-for-granted, so that hearers feel compelled to accept it without dispute. Thus, if they deny a metaphorical insult, use it in a conditional, or report someone else’s insulting metaphor, the objectionable perspective remains standing. By way of presupposing an objectionable perspective, the speaker makes us complicit in legitimating her metaphor without challenging it, thereby allowing the conversational game to proceed on her objectionable terms. This does not mean that hearers are ‘powerless victims’. Camp argues that hearers may resist the pull of metaphorical insults in various ways. For example, they can rebuff (albeit disingenuously) the speaker’s invitation to inhabit an objectionable perspective by insisting on a flat-footed literal interpretation. Or they may reject the aptness of the speaker’s perspective, and re-frame the subject under a different characterisation by offering another metaphor in response. What enables hearers to respond in this way is pragmatics. Camp argues that though perspectives are presupposed, they may nevertheless underwrite primary illocutionary acts such as assertions, questions, orders, and the like, thus giving metaphorical insults a rhetorical punch. Overall, by combining perspective, presupposition, and pragmatics, Camp shows how non-cognitive, non-propositional aspects of thinking can be vindicated as part of cognitivism about metaphor.

If Hills, Green, and Camp have focused on the question of output in metaphor interpretation, as well as touching on the question of the underlying mechanisms, Ofra Magidor focuses on the question of input in figurative language. In particular, she considers what theoretical implications arise from figurative speech using category mistakes as input. She argues that the view that category mistakes such as ‘Green ideas sleep furiously’ are meaningless is inconsistent with prominent contemporary theories on which figurative language is meaningful. This raises the following dilemma. Either the meaninglessness view of category mistakes is correct, or the meaningfulness view of figurative speech must be rejected. Conversely, either the meaningfulness view of figurative speech is correct, or the meaninglessness view of category mistakes must be rejected. This dilemma has been overlooked by contemporary theories of figurative speech. This is because they explain the meaningfulness of figures as something that is recovered at the level of speaker-meaning, or other form of meaningful import, so there is nothing inconsistent in the claim that a sentence can be figuratively meaningful while being literally meaningless. Magidor, however, maintains that such theories are committed to the assumption that the input for figures must be meaningful, so this makes them incompatible with the view that category mistakes are meaningless.

For example, Davidsonian non-cognitivist accounts which explain metaphorical import in terms of imagery and affective non-propositional effects, require that the sentence used metaphorically be literally meaningful as a whole, since it is the grasp of such literal meaning that causes hearers to draw metaphorical associations. Equally, Gricean cognitivist accounts which explain figurative meaning in terms of implicature-derivation, face a similar problem because working out the implicature requires that what is said be meaningful. Both strategies, Magidor contends, are incompatible with the meaningless view.

A way out of this dilemma is to drop the requirement that the input-sentence for figurative interpretation be meaningful as a whole. This makes room for a local interpretation taking as input a word or expression, and thus undercutting the need of a prior literal interpretation of the whole sentence. This strategy has been implemented in several ways. One option is semantic. For example, Stern (2000) holds that metaphorical meaning is semantically constrained through a lexical (unpronounced) operator, Mthat, at the level of logical form, which takes as input the expression’s literal meaning and delivers metaphorical meaning as output. This directly contributes to what a speaker metaphorically asserts, without having to ask the question of what she’s literally saying and whether or not she’s saying something meaningful. Magidor, however, thinks this strategy is not general enough. She argues that Stern’s account cannot explain complex metaphors such as ‘She dropkicked the idea noisily off the nearest cliff’. This is because for Stern, Mthat should operate on the whole verb-phrase ‘dropkicked the idea noisily off the nearest cliff’, and since he requires that Mthat operate on meaningful phrases, it follows that the entire verb-phrase must be meaningful. But this verb-phrase is part of what generates a category mistake, so it counts as meaningless.

Magidor argues that the same problem arises for pragmatic accounts of figurative meaning, as championed by Recanati (2004) and Bezuidenhout (2001). Since they require that the literal meaning of the figurative expression be pragmatically fine-tuned, prior to a literal interpretation of the whole sentence, they also predict that in a complex metaphor like the above the entire verb-phrase must receive a pragmatically modulated meaning. Since they require that the figurative expression be meaningful, it follows that the entire verb-phrase must be meaningful. Thus, pragmatic views also remain incompatible with the meaningless view. Similar problems arise for complex metonymy such as ‘The large impatient ham sandwich left without paying’, and fictional discourse more generally. Though Magidor does not offer a solution to how this incompatibility should be resolved, she is of the belief that this is an important dilemma that prominent theories of figurative language should not ignore.

The second group of contributions explores I-figures or irony-related figures that have seen relatively little attention, e.g. litotes (understatement), hyperbole (overstatement), embedded irony. Laurence Horn focuses on litotes, in particular what makes a good input for a litotic interpretation. He shows how double negations are an excellent input for litotes. This is because they can be coerced either semantically or pragmatically into contrariety, and contrariety yields an unexcluded middle that allows for something to be in-between. For example, someone who is ‘not unhappy’ may not be happy either, and one who is ‘not not friends’ with another may not be friends with her either. How is this possible if not not-p reduces to p? The key, Horn suggests, is recognizing in litotes a corollary of MaxContrary—i.e. the tendency for contradictory (wide-scope) sentential negation ¬p to strengthen (at least) pragmatically to a contrary ©p. When not not-p conveys ¬©p, the negation of a virtual contrary, the middle between p and not-p is no longer excluded. This makes double negation very expressive, because by negating a contrary it yields the perception of a weakened force through the recovery of an unexcluded middle. This enables one to understate what one means by saying less. For example, saying that someone is ‘not unintelligent’ conveys that he’s not of an average intelligence, but decently above the average. This is a weaker affirmation than what the un-doubly-negated counterpart would express, and it works as a better tool to yield understatement because it is socially and epistemically preferred despite its linguistic complexity.

Kendall Walton, on the other hand, focuses on the relationship between understatement and overstatement, and how they relate to irony. He argues that though understatement and overstatement are similar in some respects, they should not be analyzed in a similar manner. To start with, they have some surface similarities. Whether I say ‘There are a couple of cops out there’ to understate that there are surprisingly quite a lot of policemen, or I say ‘There are hundreds of cops out there’ to exaggerate how many there are, what I mean in both cases is roughly the same, say, there are significantly more than two and significantly fewer than two hundred. In neither case do I mean what I literally say, or what Walton calls ‘explicit content’ (EC). What I mean instead is that there are quite a few policemen, or what he calls ‘assertive content’ (AC). These two notions of explicit content and assertive content also serve to explain how understatement and overstatement differ. In understating, one is representing by what one says a quantity as being smaller than what one asserts it to be: EC is less than AC. In overstating, it’s the reverse: one is representing, by what one says, a quantity as being larger than what one asserts it to be: EC is larger than AC. This explains the intuitive difference between the two figures: understating is saying less than what one means, and overstating is saying more than what one means.

But this difference, Walton notes, threatens to evaporate once we recognise that understatement and overstatement are not just the reverse of one another, but they do recruit, respectively, elements of one another. Thus, understating how large a quantity is, is overstating or exaggerating how small it is, and vice versa overstating how large a quantity is, is understating how small it is. The question arises then as to whether there is understatement and overstatement simpliciter. To answer this, Walton introduces a third notion of ‘salient contrast’ (SC), which is roughly what the speaker is especially concerned to indicate is not the case in a context. He argues that whether an utterance is best understood as understatement, or overstatement, depends on how small or how big the distance between EC and SC is, compared to the distance between AC and SC. In understating, SC and EC coincide or overlap, or at least the distance between them is smaller than it is between AC and SC. In other words, the speaker voices, by what she literally says, the salient contrast, or at least she understates the gap between what she says and what she means to deny. Understatement is in this regard akin to irony. In irony too the speaker means to deny what she literally says, or at least to minimise the gap between what she says and what she indicates is not the case. Overstatement is different: the distance between EC and SC is greater than that between AC and SC. This is because the speaker needs to exaggerate the gap between what she says and what she means especially to deny, so that hearers wouldn’t take her to mean what she says, but would rather look for a smaller quantity as being what she asserts. With these three notions of explicit content, assertive content, and salient contrast, Walton offers a principled explanation of how understatement and overstatement are related, and how they differ.

Stephen Barker’s paper focuses on irony, in particular embedded irony, and what theoretical implications they raise for the dominant semantic paradigm. He argues that irony embedding in compound sentences, like belief-reports or conditionals, challenges a core semantic thesis which he calls the propositional content principle (PCP). PCP says that what is common to a sentence in self-standing illocutionary acts like assertion, and its being embedded in compound sentences, is a propositional act—i.e. an act of uttering a sentence encoding a truth-conditional content. This is because irony is a pragmatic content which is neither propositional nor truth-conditional. In self-standing irony the speaker is ridiculing someone who might believe what she literally says, conversationally implicating that she believes an inverted content. Yet irony embeds, and it’s reasonable to assume that whatever the speaker is doing in self-standing irony, is also doing in embedded irony. Take the following exchange about George who is extremely dim: Max—‘I get the feeling that George is a real genius’, to which Sam replies—‘So do I. If he is, then we ought to allow his genius to shine, by getting him to lick those stamps’. Both Max and Sam engage in pretence, ridiculing someone who might think/assert that George is smart. But when Sam uses irony in the antecedent, the implicatures that arise in self-standing irony should also be commitments of the whole conditional. But clearly they are not. In making a conditional, Sam is not committed to George being dim, and ridiculing anyone who might think he’s a genius. He’s merely putting forward these claims conditionally. Irony is used in the antecedent, and consequent, but neither of them is performed in a self-standing act. How is this problem to be solved?

Barker argues that PCP-semantics cannot ultimately explain embedded irony. The problem is the separation between force (illocutionary act) and content (propositional act)—where force is the pragmatic operation applying to a proposition. According to PCP, only the propositional act embeds, not the force or the implicatures that a self-standing act might otherwise carry. To solve this problem, Barker suggests giving up the separation between force and content, and treat them as part of a single package. He proposes an expressivist speech-act theoretic approach, which he calls language agency (LA), and which defines illocutionary acts, truth-aptness, belief/thought, ironic interpretation, and embedding, without propositions, and so without the force/content distinction.

For example, assertion is defined, in a Brandomian vein, as involving an intention to defend a mental state—where defence is understood as a disposition to provide reasons for possessing the state. This enables defining the truth-aptness of assertions as something that is not inherited from the truth-aptness of beliefs. Assertions are truth-apt, for Barker, not because they express a truth-apt belief, which bestows its truth-aptness on the utterance, but rather because they are acts produced with the purpose of defending the mental state expressed. Other kinds of illocutionary acts are only made with the purpose of manifesting states, but do not involve defence, so for Barker, they are non-truth-apt. Irony is non-truth-apt because making an ironic utterance does not come with the purpose of defending the states expressed, but rather of giving a ridiculing portrayal of cognitive states.

What of embedded irony? If irony is non-truth-apt, can it embed without having to carve off a propositional (truth-conditional) content from the ironic act? To explain this, Barker introduces the notion of a proto-act, which is the most basic building block in the structure of a speech-act. To make a proto-act is to produce an utterance, presenting oneself as having certain expressive aims—depending on the kind of illocutionary act—while at the same time communicating that one lacks those commitments. The notion of proto-act can thus help explain what goes on in embedding. When we produce the antecedent of a conditional, we are not performing a fully-fledged act, say, an assertion. Rather, we engage in the behaviour characteristic of someone making an assertion—that is, someone who is disposed to defend a mental state—while indicating that we lack characteristic assertoric commitments. Our overall goal is with the entire conditional, not just the antecedent, or the consequent. Similarly, when Sam uses irony in a conditional antecedent (and/or  consequent), what embeds is not a fully-fledged ironic act, but a proto-ironic act. Sam presents intentions to parody by engaging in a behaviour characteristic of an ironic speaker who intends to ridicule someone thinking that George is a genius, and respectively that his genius would shine by making him lick stamps. But Sam isn’t committed to any of these claims. His point is instead to show that characteristic ironic commitments in the antecedent entail characteristic ironic commitments in the consequent.

This solution to embedding of irony via proto-acts undercuts the need of having to box irony into truth-conditional content to comply with PCP. This is because what embeds is a proto-ironic act, and not a fully-fledged irony. Proto-ironic acts can thus function as a conditional antecedent, and/or consequent, or any embedded sentence for that matter. Moreover, since proto-acts don’t divide into content and force, but contain them both as part of a unitary act, Barker is able to explain how characteristic ironic commitments may be confined to the embedded sentence, without however being undertaken with the whole compound sentence. Barker extends this idea to show that objects of beliefs can be speech-act types, and that logical connectors can be speech-act operators, suggesting that none of the problems that arise for the semantic paradigm arise here.

The last contribution by myself, features a relatively little studied phenomenon of compound figures, namely when two figures combine together to make a more complex figure. I focus here on ironic metaphor—e.g. ‘What delicate lacework’ said about a doctor’s messy piece of handwriting—and in particular on the question of the order in which the two figures are interpreted. Stern (2000) and Bezuidenhout (2001), who previously considered this problem, have argued that metaphor comes prior to irony in the structure of what is communicated. This is a thesis about the logical priority of metaphor over irony, which I call Logical-MPT. It says that we first derive the metaphorical content, which in turn launches the ironic content, so that the latter logically depends on the former. This content-dependence is key for Stern and Bezuidenhout to argue that metaphor and irony are markedly different types of content—metaphor is truth-conditional, irony is non-truth-conditional.

While the basic line of argument is correct, I show that a Strong-Logical-MPT in fact holds—i.e. in all cases of ironic metaphor, metaphor must be logically prior to irony—including compounds with highly conventionalised metaphors, for which Stern predicts an indeterminate order of interpretation. A more decisive argument for Logical-MPT comes from the impossibility of an irony-first order of interpretation, and the main reason, as I see it, is that it cannot model the behaviour of the ironic attitude in the compound. We might expect that a metaphor like ‘He’s a towering figure’—said ironically of an ineffectual politician—would deliver the same content, regardless of whether metaphor comes first, or irony comes first. But there is a critical difference that is overlooked by content-based explanations, which has to do with the role of ironic attitude. In particular, if irony were interpreted first—that the man is a diminutive figure—and then re-interpreted metaphorically—he is an unimpressive politician, there is no way we can preserve the ironic attitude as part of what is meant with the whole compound. The attitude merely targets the literal claim, and is not subject to metaphorical re-interpretation. Instead, if metaphor is interpreted first, then the whole compound can retain the ironic attitude as part of its overall communicative goal: the attitude is about the metaphorical claim, so that the compound counts as primarily ironic. It amounts to giving a ridiculing portrayal of someone making a metaphorical-act, and showing how inappropriate it is in the context.

To explain the role of the attitude in ironic metaphor, I develop a complementary speech-act based explanation of Logical-MPT and argue that is to be preferred to a content-based explanation. To create this explanation I draw on Barker’s (2004) expressivist speech-act theory, in which speech-acts build on other speech-acts, here to combine in a more complex speech-act. In particular, I show how Barker’s general ideas help explain why metaphor can be an assertive-act, and irony a ridiculing-act, and how metaphorical-acts and ironic-acts can build one on the other. While an ironic-act can build on a metaphorical-act, a metaphorical-act cannot build on an ironic-act. This restriction on how they can be composed establishes Logical-MPT via a different route. Metaphor has priority, because metaphorical-acts provide the basis for primary illocutionary-acts, and ironic-acts require and build on primary illocutionary-acts. So irony must build on metaphor and not the other way round.

This explanation also accounts for the role of metaphor in the compound. In making an ironic metaphor, the speaker is not undertaking metaphorical commitments, but is merely using the metaphor for ironic purposes. This is, I claim, because the metaphor is not a fully-fledged metaphoric act, but is instead performed as a proto-metaphoric-act—i.e. an act by which the speaker engages in the behaviour characteristic of someone making a metaphor, say, someone who intends to defend a similarity between the doctor’s handwriting and lacework (that the handwriting is beautiful, shows skill, etc.), while indicating that she lacks such metaphorical commitments. Since the speaker’s overall point is to be ironic, this proto-metaphoric-act is thus nested inside an ironic-act. The speaker’s communicative intention is with the latter, ironic-act, and metaphor is merely instrumental to achieving that intention. This offers a neat way of building complex speech-acts out of more simple speech-acts.

To conclude, together these papers illustrate the importance and fruitfulness of detailed investigation of figurative talk. The variety of figures of speech considered here and the breadth of theoretical positions provide a valuable opportunity for reflection and analysis on fundamental issues in philosophy of language. In particular, figures raise questions about what a good theory of meaning and communication should make room for. Figures of speech highlight the richness and flexibility of human communication, paving the way towards a better understanding of how language and mind work. We thus hope that the collection of papers presented here will impact future research not only on figurative talk but on this much wider issue.