Asking questions

Kyle Rawlins (Johns Hopkins University)

A classic, widely assumed idea about questions is that they are requests for information. This idea is not much older than the long list of question types that break this mold, sometimes termed non-canonical questions. These include a fairly wide variety of move types that are heavily used in naturalistic data such as: rhetorical questions (where all agents may know the answer); biased questions / questions with marked forms including rising intonation and tags (where information may be requested, but the question also provides some guidance towards an answer via a marked form); a very wide range of questions marked with diverse particles cross-linguistically; exam questions (where the addressee is asked to demonstrate knowledge of an answer); self-addressed/speculative questions; and many more. This talk develops a view about what it is to ask questions under which few if any of these can be said to be non-canonical. The idea is that questions are moves that open coordination on the resolution of an issue. In some cases successful coordination may just take the addressee providing information to an ignorant speaker, and this case corresponds to classic requests-for-information. On the coordination view, I suggest that supposedly non-canonical questions fall out as coordination strategies that are optimal under certain circumstances when there are not full asymmetries in agent ignorance (among other contextual factors). This work builds on a wide range of research where a coordination view is arguably implicit, e.g. in the literature on Questions Under Discussion, as well as more recently the Farkas & Bruce-based Table models of discourse; it isn’t intended to offer an alternative to these ideas, but rather a reorientation under which it becomes clearer both what it is to ask a question, and where there should be such diversity in the kinds of questions we see in natural language.