Interrogatives: Interpretation and resolution

Even the most mundane interrogatives have highly underspecified meanings. Imagine we are hiking on a trail you know well, and we run out of food and water. If I then ask you, "Where can we buy supplies?", you're likely to name one or a few close places where you think we can get what we need. In contrast, if we're at my house scheming to start a company, your answer is more likely to consist of an exhaustive list of (affordable, permitted) vendors you know about. The interrogative is the same in both cases, but our understanding of the question asked and, in turn, what counts as a felicitous answer, is dramatically different. I'll illustrate this underspecification, mostly using examples collected via a two-player, online game designed to study the language people use during collaborative search tasks. These data provide a glimpse into the ways in which the underspecification of interrogatives (and the replies they receive) is resolved. We'll see that it's a complex optimization problem: the interrogative both shapes, and is shaped by, the discourse goals and the nature of the reply. This optimization problem is hard. I can't offer a solution, but I think I can point the way to one, by building upon the decision- theoretic approach to pragmatics pioneered by David Lewis, Arthur Merin, Robert van Rooij, Sophia Malamud, and many others. The most illuminating cases are those in which the reply seems to provide more information than was requested. These can be unexceptional ("Do you know what time it is?" "It's 2 o'clock"), they can be shocking ("Is Smith happy at his new job?" "Yes, and he hasn't been to jail yet"), and they can be remarkably devious.... (This talk is based, in part, on ongoing work with Jesse Harris, Maribel Romero, and other members of the SUBTLE project.)

.pdf version