Par auteur > Onofri Andrea

This paper discusses and criticizes an important aspect of Lewis's theory of conventions (Lewis 1969, 1975):

The Symmetry Principle: If a declarative sentence m has meaning p in the language of a population P, then generally: (i) A member of P utters m only if she believes that p; and (ii) if a member of P hears an utterance of m, then she forms the belief that p.

According to this principle, linguistic conventions are “symmetric” in two different respects. First, if a communicative exchange follows the conventions in question, then the speaker expresses a belief and the hearer forms the same type of attitude (a belief) in response to the speaker's utterance. Second, the speaker and the hearer's respective beliefs have the same content as the uttered sentence.

In this paper, I criticize both of these claims: I argue that the conventional response to a declarative sentence m need not be a belief, and even if it is, it need not have the same content as m.

Concerning the type of attitude that's formed by the hearer in a conventional linguistic exchange, there are several cases where the receiver doesn't form a belief while still conforming to the relevant linguistic conventions. Consider for instance a hearer who understands the assertion, while suspending judgment about its truth; a hearer who wonders whether the assertion is true and begins inquiring as a result; or a hearer who imagines what would happen if the assertion were true. I'll argue that all these cases display the central features of a conventional linguistic response, yet none of them involves belief formation on the part of the hearer.

I'll then move on to the second part of my argument: A declarative sentence m can have meaning p in the language of a population P even though, upon hearing an utterance of m, the members of P tend to form an attitude towards a proposition that's different from p. My argument will be framed in game-theoretic terms, following Lewis's own approach. Lewis (1969) famously discussed a signaling game where the players can settle into different optimal equilibria or signaling systems. While the signaling systems of the Lewis signaling game do seem to conform to the Symmetry Principle, a different picture emerges when we move to more complex games. Even a slight adjustment of some parameters can move us from the Lewis signaling game to a very different game, where the conditions of the Symmetry Principle don't apply. For instance, Skyrms discusses various games where there is an “information bottleneck,” because the number of available signals is smaller than the number of possible states (Skyrms 2010, 15–16, ch. 9). In one of Skyrms's games, there are three equiprobable states, two signals available to the sender, and two responses available to the receiver (ibid., pp. 15-16, 107-09). Analyzing this game, I'll argue that the optimal signaling systems in this game do not behave as the Symmetry Principle predicts: More specifically, the receiver's conventional response to the signals fails to match the content of the signals themselves.

References:

Lewis, D. (1969). Convention: A philosophical study. Wiley.
Lewis, D. (1975). “Languages and language”. In Martinich and Sosa (eds.), The Philosophy of Language (pp. 682-700), sixth ed., Oxford University Press.
Skyrms, B. (2010). Signals: Evolution, learning, and information. Oxford University Press.