I offer a semantics and pragmatics for imperatives, developed in the framework of a dynamic pragmatics in the vein of Portner (2004, 2007, 2018, 2018b) and Roberts (1996/2012, 2004, 2012b, 2018); see also Roberts (2017). Context on this approach constitutes the scoreboard (Lewis 1979) of the language game in play; it is articulated into several types of information shared by the interlocutors, including a Common Ground (CG), a stack of questions under discussion (QUD), and the publicly evident goals and plans to which they are individually and jointly committed (G; cf. Portner’s (2004, 2007) ToDo lists). Each utterance is a move which functions to update the scoreboard. The three central types of illocutionary acts—assertions, interrogations, and directions—are canonically made by uttering linguistic tokens differing in semantic type: propositions (sets of possible worlds) vs. questions (sets of sets of propositions) vs. directed properties (functions from worlds to sets of entities, indexed to an addressee). Semantic content is compositionally derived and static; and semantic type is determined by clause type—indicative, interrogative, imperative—as indicated in a syntactic representation of logical form, e.g. LF. But illocutionary force is determined purely pragmatically as a function of the state of play in the language game, as reflected in the dynamic scoreboard: Only utterances (Bar-Hillel 1971: ordered pairs of a linguistic constituent and a context of utterance) have illocutionary force; and only consideration of the CG, QUD and interlocutors’ evident goals and intentions reflected in G permits one to infer the intended force of an utterance, i.e. the way in which the speaker proposes to update the context with its contextually resolved semantic content.
I will briefly sketch how this framework affords us an account of the semantics and pragmatics of imperatives which has the virtues of the best previous accounts (in particular those of Portner 2004, 2007, 2011b; Kaufmann 2006 (as Schwager), 2012; Charlow 2011, 2014; and Condoravdi & Lauer 2012), while avoiding problems that arise in those accounts. And I think it makes sense of the intuitions underlying Barker’s (2012) proposal that imperatives denote actions; here, they do not denote actions, but (at least the non-expressive imperatives) propose them. The semantic content of an imperative used to issue a directive is understood to constitute a conditional goal for addition to the public goals to which the targeted addressee is committed. Imperatives’ variable flavor, as in Kaufmann (2012), is captured by the deontic character of this commitment to goals, along with a presupposed Kratzerian modal base and ordering source used to capture their conditional character. Understanding imperatives in this way naturally helps to illuminate classic problems like Ross’ paradox (1941) and the problem of free choice permission (Kamp 1973).