In his talk last Wednesday at the LOGOS Seminar, "On Metaphorical Content", Gergö Somodi gave an argument that puts pressure on anti-Davidsonian theories of metaphorical content, and suggested a possible way out, to be further researched and elaborated. Here I will present my interpretation of the paper, and I will indicate that, if I understood it correctly, the research project is indeed worth pursuing.
Following more or less the Austinian terminology that Gergö was using, and more or less the interpretation of Davidsonian views he was assuming, on a Davidsonian view the only illocution made with an utterance of a metaphorical sentence such as 'Juliet is the Sun' is the one whose content is the literal necessary falsehood that Juliet is identical with our star. It is true that the utterance conveys to audiences other, more sensible ideas, such as the claim that Juliet gives warm and light to the speaker; but this is no illocution, it is only a causal effect of the literal illocution on which the latter has as little rational influence as if that idea had been produced in the audience by "a bump in the head" (Davidson, sic). On a Davidsonian account, then, grasping that Juliet gives warm and light to the speaker is merely a perlocutionary effect.
Anti-Davidsonians like Elizabeth Camp argue instead that the more sensible idea is also an illocution (in addition perhaps to the illocution of the literal meaning) of the utterance, perhaps conveyed in the indirect way that indirect speech acts or conversational implicatures are conveyed, or perhaps more in the way that context-dependent meanings are conveyed. Now, the problem that Gergö raised for these views (as Genoveva helped me to appreciate) goes as follows: writers like Camp accept that an essential part of the mechanism through which the alleged metaphorical illocution is conveyed to audiences has audiences "noticing resemblances, seeing things, entertaining pictures"; but all of these are perlocutionary effects; how can an illocution be conveyed by essential perlocutionary means?
My initial resistance to this way of setting the problem was as follows: if perlocutions are defined the way Gergö proposed (intention-irrelevant causal effects of utterances), then it is not clear that the categories illocution/perlocution are incompatible. For understanding the literal semantic content of a context-dependent utterance, such as Kaplan's 'That is a picture of the greatest philosopher of the twentieth century.', might well be an intention-irrelevant (in the sense indicated by Gergö) causal effect of an utterance. But then the anti-Davidsonian is safe, because "noticing resemblances, seeing things, entertaining pictures" might be perlocutions but also illocutions, and there is no problem with the view. This would be clearer if we used a less question-begging description of the way we interpret metaphors, such as, in our example, "thinking of the target-domain of persons in terms of the commonly believed properties of the source-domain of stars", instead of speaking of "noticing, seeing, picturing".
Alternatively, we can define "perlocution" in such a way that the categories of illocutions and perlocutions are really incompatible, as on the Strawsonian definition that what is distinctive of perlocutions is that they cannot be produced by Gricean communicative intentions. But then it is question-begging to say that the way we interpret metaphors, through "noticing resemblances, seeing things, entertaining pictures", is a perlocution. The anti-Davidsonian would say that this is no perlocution, in the Strawsonian sense, but rather something that can indeed be achieved by means of communicative intentions, what, again, would be clearer if we described it in less question-begging terms such as "thinking of the target-domain of persons in terms of the commonly believed properties of the source-domain of stars".
In the course of the discussion (particularly the exchange with Josep), I came to think that this is also what Gergö wants to suggest, and that his point was rather that defending it requires a better clarification of what is usually meant by perlocution, so that we can see that effects that in some sense can be called 'perlocutionary' can contribute to properly illocutionary effects. With this I agree, in fact Gergö's characterization of a perlocution as an intentionally-irrelevant causal effect of an utterance is what many writers on these topics seem to have in mind (see for instance chapter 2 of Alston's Illocutionary Acts and Sentence-Meaning).