Interpretation of Optimal Signals

According to the optimal assertions approach of Benz and van Rooij (2007), conversational implicatures can be calculated based on the assumption that a given signal was optimal, i.e. that it was the sender’s best choice if she assumes, purely hypothetically, a particular naive receiver interpretation behavior. This paper embeds the optimal assertions approach in a general signaling game setting and derives the notion of an optimal signal via a serious of iterated best responses (c.f. Jäger, 2007). Subsequently, we will compare three different ways of interpreting such optimal signals. It turns out that under a natural assumption of expressibility (i) the optimal assertions approach, (ii) iterated best response and (iii) strong bidirectional optimality theory (Blutner, 1998, 2000) all prove equivalent. We then proceed to show that, if we take the iterated best response sequence one step further, we can account for M-implicatures (Horn’s division of pragmatic labor) standardly in terms of signaling games. Often we express more with the use of our words than what those words mean literally. For example, if you were to say that this observation is not particularly new, I would clearly get the hint and understand that you meant to say that it is more than just not particularly new, indeed a working standard in linguistic pragmatics. Such conversational implicatures were first studied by Grice (1989) and still concern the community in various ways. In particular, recent years saw an increasing interest in game-theoretical models of conversational implicature calculation, and this study belongs to this line of research. It provides a formal comparison of selected previous approaches ∗Thanks to Anton Benz, Reinhard Blutner, Tikitu de Jager, Gerhard Jäger, Robert van Rooij and an anonymous referee for very helpful comments and discussions. “KNAW ̇Franke ̇ConvImplic ̇Revised” — 2007/10/17 — 15:02 — page 2 — #2