Discourse-level information structure and the challenge of metadiscursives

In Information Structure: Towards an integrated formal theory of pragmatics (2012), Craige Roberts offers a modeling apparatus designed for use in giving a unified analysis of diverse pragmatic phenomena. As part of her account, Roberts offers precise conditions on what an utterance must be like in order for it to count as felicitous in any particular context. A key condition is that, in order to be felicitous, an utterance must count as an answer to what Roberts calls “the question under discussion.” I demonstrate that this account generates some false predictions regarding the class of utterances I will call “metadiscursives,” as well as that of epistemic reports. I will then consider several possible alterations to Roberts’ view that would preserve the spirit of her project and render the account invulnerable to my objections.