Metaphor and contextual coherence: it’s a match!

ABSTRACT

Many sentences can be interpreted both as a metaphor and as a literal claim, depending on the context. The aim of this paper is to show that there are discourse-based systematic constraints on the identification of an utterance as metaphorical, literal, or both (as in the case of twice-apt metaphors), from a normative point of view. We claim that the key is contextual coherence. In order to substantiate this claim, we introduce a novel notion of context as a rich and heterogeneous body of information, including previous discourse, elements coming from the surroundings of the utterance, background information, and Questions Under Discussion (QUD) issued from these three sources. We then define contextual coherence as a relation between what we call the minimal paraphrase of the metaphor and the context, and argue that for an interpretation to be coherent two conditions must be met. First, the minimal paraphrase must address some question in the QUD stack. Second, it must be externally consistent, i.e. consistent with the available contextual information. Finally, we argue that an approach based on contextual coherence is better suited to deal with twice-true and twice-apt metaphors than traditional approaches based on semantic deviance or pragmatic lack of fit.

KEYWORDS:

• twice-true metaphor
• twice-apt metaphor
• coherence
• Questions Under Discussion
• context
• interpretation

Acknowledgements

This is joint work to which the three listed authors have equally contributed. Names are listed in alphabetical order. The present work springs from discussions within a reading group on contextualism set up at the Institut Jean Nicod (CNRS, EHESS, ENS-PSL University) back in 2015. Palle Leth participated in the group along with the three authors. Those discussions led the group to consider how metaphorical interpretation relates to establishing rhetorical relations. The authors thank Palle Leth for his contributions to the discussions from which the present work emerged. Previous versions of this paper have been presented in 2021 at the Seminario del Departamento de Lógica, Historia y Filosofía de la Ciencia of UNED, Spain, the workshop Metaphor and Ambiguity Analysis at the University of Tübingen, Germany, and the Workshop on Context: Semantics, Pragmatics and Cognition at the University of Genoa, Italy. We thank the organisers and participants for their feedback.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Cf. (Bach 2003, 5), quoted in the DWDS Corpus, originally in German: Ich muss raus, ich kann hier nicht atmen.

2 We thank a reviewer and the journal editors for pressing us to clarify this point.

3 For an overview of the tradition of cognitive pragmatics, cf. e.g. (Mazzone 2021). For a glimpse of what formal pragmatics is about, cf. (Roberts 2012).

4 The trouble of explaining twice-apt metaphor is especially acute for those pragmatic accounts that assume that speakers cannot be taken to mean what they are saying literally (cf. Grice 1989; Searle 1979). This assumption in fact generates the implicature in Grice’s original account. However, not all pragmatic accounts that appeal to Gricean speaker meaning make that assumption (cf. e.g. Camp 2008). Where we differ from these latter accounts is with respect to the explanation we give of the phenomenon in question, notably by relying on our particular conception of context as well as of coherence (see Sections 3 and 4 further below).

5 In absence of a conversation or discourse, the matter would be undecided. More elements of the original context for this example are offered in Section 5 below.

6 Cohen (1976) spoke of ‘twice-true’ metaphors because he aimed to disprove the adage that ‘a metaphor taken literally is false’. By contrast, we understand the term in a way that undermines the assumption that metaphors taken literally are nonsensical, i.e. lack clear truth conditions. This would be the case, e.g. if one assumed that metaphors necessarily involve some sort of category mistake (and assumed that category mistakes are meaningless). What is important for our discussion is that there are utterances deemed metaphorical the literal interpretation of which is meaningful too. Whether or not the truth conditions that the literal reading yields are fulfilled, i.e. whether the sentence taken literally is true in a given context of utterance, does not matter for present purposes. In light of this, what we mean by ‘twice-true metaphors’ would more aptly be captured by something like ‘doubly meaningful utterances’. We will stick, however, to the established terminology.

7 Binkley (1974) had already drawn attention to these cases some two years earlier, without however proposing a term to designate that class.

8 Cf. e.g. (Black 1954; Beardsley 1962).

9 This is the umbrella term that Stern (2000, 3–4) chooses to designate a broad family of views. The various individual views thus subsumed do not necessarily employ the terminology of ‘deviance’.

10 As compared to proposals that appeal to pragmatic lack of fit (cf. Grice 1989; Searle 1979, and especially Camp 2008), in turn, we differ with respect to our conception of context and, consequently, of coherence (cf. also footnote 4).

11 Relevance Theorists form one prominent group within this camp. But truth-conditional pragmatists and contextualists more generally are certainly drawn to such a view, if they have not outright endorsed it.

12 The distinctive, multipropositionalist model for metaphorical interpretation that Heise (2022) defends capitalises on an idea that Ofra Magidor (2020, §3.2.2) recently floated with respect to category mistakes: ‘in category mistakes such as “Two is green”, each word succeeds in picking out a content but these contents fail to compose together to form a unified proposition’. Rather, on the multipropositionalist model in question, the respective contents provide the propositional templates from which to build the bifurcated propositions that enable the comparison metaphorical interpretation sets up.

13 Cf. (Leezenberg 2001). Incidentally, this is in line with more recent developments within Relevance Theory, cf. (Carston 2010; Carston and Wearing 2011; Wearing 2014).

14 In support of this position, cf. e.g. (Searle 1979; Bergmann 1982; Stern 2000; Bezuidenhout 2001; Carston 2002; Camp 2006; Wearing 2006; González de Prado Salas 2016). Davidson (1978) and Reimer (2001), among others, argue against this position.

15 Stotts (2021) notes that some metaphors might have paraphrases that fully capture their expressive power and distinguishes them from more complex metaphors whose paraphrase would be a command to associate two things in one’s mind. We deem it plausible to think that most metaphors are in-between: they can be paraphrased roughly, but not exhaustively.

16 There are contexts where the same sentence could be given a metaphorical interpretation. Imagine that it is uttered by a professor who complains to one of her colleagues about the budget cuts the university has recently suffered and the consequent low quality of the research done at the university. She says things like: ‘We have become an educational centre, but this is not what the university was supposed to be. We are a research centre. But nobody is doing research anymore, there is no money for it. Where is the university now? I can’t see it’. The child example is inspired by an example by Ryle (1949).

17 Cf. e.g. (Lascarides and Stone 2006; Stone, Stojnic, and Lepore 2013; Hunter, Asher, and Lascarides 2018).

18 Why should we require that perceptual information be gathered in propositional form? The requirement about perceptual information being gathered in the form of propositions is a methodological choice which facilitates handling this information in the framework we develop below. Granted, not all perceptual information is readily available in propositional form. However, we make the assumption that perceptual information that counts in the interpretation of an utterance can, if necessary, be formulated in propositional form by the interpreter, in particular in case she is asked to explain or justify her interpretation. (We thank one of the journal’s reviewers for raising this issue.)

19 We thank one of the journal’s reviewers for pushing us to clarify our position with regard to this point.

20 The QUD framework was originally proposed by von Stutterheim and Klein (1989) and van Kuppevelt (1995) as a general approach to the analysis of discourse structure, in particular regarding topicality. Van Kuppevelt’s hypothesis is that ‘a discourse derives its structural coherence from an internal, mostly hierarchical topic-comment structure.’ The relation of subquestionhood is introduced as a technical notion allowing to obtain a formal model of the topic-comment discourse relation. Later elaborations can be found among others in (Ginzburg 1996, 2012; Roberts 2012; van Rooij 2013). Details about how questions get introduced and removed from the QUD stack can be found in Ginzburg (2012, 4.3).

21 We write ‘typically’ because information about the speaker or the genre might motivate a rejection of this requirement. As an example, think of exquisite corpses. Because of what we know about these texts, we don’t take the contributors to intentionally address a unified topic.

22 Material incompatibility designates a particular sort of mutual exclusiveness exemplified by the way in which, e.g., colour predicates such as, say, ‘blue’ and ‘green’ may not simultaneously apply to a clearly delineated patch of fabric.

23 This kind of reasoning is, of course, fallible, in the sense that it can lead to a wrong interpretation of what the speaker means. Actual speakers can fail to be contextually coherent. And background information can turn out to be false.

24 Here our account differs from other theories that use QUD to explain context-dependence, such as the one found in Schoubye and Stokke (2016). For the identification of the figure of speech, QUD are sometimes insufficient. Other arguments against QUD being the sole foundation to work out discourse coherence are offered by Hunter and Abrusán (2017).

25 In practice, the contextual dynamics can be more complex. An otherwise coherent interpretation which is in conflict with a piece of background information might provide the grounds to reject a piece of background information. To deal with these cases we might need to consider degrees of coherence, a topic that goes beyond the aim of this paper.

26 We have already argued that this assumption is not always in place. Sometimes what seems to be a category mistake is indeed one. For instance, it is also a piece of background information that some people’s utterances, e.g. young children’s, can simply be internally inconsistent, as seen in example (3). This is also the kind of possibility that one should seriously take into account in scenario 2 for example (2), which challenges our background assumption that (grown-up) speakers are competent and cooperative. Even if cases that defy this assumption may seem rare, there are such situations, and the corresponding awareness partly guides our interpretations.

27 We borrow the term ‘robust’ as well as the idea that contexts of interpretation vary in robustness from Stotts (2021).

28 Note that not every interpreter would have access to such background information: Consider a young teenager, a native English speaker, who is being given the excerpt of the letter. Although she could read the letter and even work out a minimal paraphrase of sentence (6), she might have little background information about the sexual mores and moral taboos in continental Europe during the XVI–XVII centuries.

29 A reviewer’s concerns lead us to note here that we do not claim that this somewhat educated reader’s interpretation would be the same as, or a better interpretation than, the one that Katya Berger Andreadakis, the addressee of this letter would work out. Although we cannot possibly provide such an interpretation ourselves, we can suppose that it might differ from ours given that their correspondence is initiated by Katya, who sends a postcard to John Berger featuring a reproduction of a painting by Titian, to which she writes ‘What do I think about Titian? In one word on a postcard: flesh’. She has access to specific perceptual information we do not have access to, since we do not know which painting by Titian was reproduced on the postcard. The flesh she refers to, is it an animal’s? Is it human? Plus, she ends the letter previous to the one in which our example (6) appears with a piece of day-dreaming where she sees herself ‘naked on a canvas in the exhibition’, with ‘a dog at [her] side’. Could this contribute to produce a more specific reading, one where dogs are the ambassadors of her desire? We do not know this, and yet, this does not run against our working out of example (6) above.

30 Decisions about how to interpret an utterance are defeasible. Forthcoming discourse can lead to revising our decisions, both in oral and written settings. Granted, access to forthcoming discourse is radically different in oral vs. written settings, as already suggested in Section 3 above. One can flip the pages of a book to search for hints on how to interpret the current utterance, one cannot do the same in oral exchanges. However, in both written and oral exchanges, information obtained as we move along may lead us to modify our prior decisions on how to interpret an utterance.

31 See also (Picazo 2022) for a discussion of the problems of introducing questions to the QUD via accommodation.

32 We thank an anonymous reviewer for pointing this out.

Additional information

Funding

Andreas Heise declares that his work was supported by two department-wide grants from the Agence Nationale de la Recherche (ANR-17-EURE-0017 FrontCog and ANR-10-IDEX-0001-02 PSL). Claudia Picazo declares she has received support by Ministerio de Ciencia e Innovación, Government of Spain (Grants No. PID2019-105728GB-I00 and PID2021-123938NB-I00) (MINECO/FEDER, EU).