Abstract Theories of imperatives differ in how they aim to derive the distributional and functional properties of this clause type. One point of divergence is how to capture the fact that imperative utterances convey the speaker’s endorsement for the course of events described. Condoravdi & Lauer (2017) observe that conditionals with imperative consequents (conditionalized imperatives, CIs) are infelicitous as motivations of advice against doing something and take this as evidence for an analysis of imperatives as encoding speaker endorsement. We investigate CIs in further contexts and argue that their account in terms of preferential conflicts fails to capture the more general infelicity of CIs as motivations for or against doing something. We develop an alternative in which imperatives do not directly encode speaker preferences, but express modalized propositions and impose restrictions on the discourse structure (along the lines of Kaufmann, 2012). We show how this carries over to conditionalized imperatives to derive the behavior of CIs, and conclude with a discussion of more general problems regarding an implementation of conditional preferential commitments, an issue that can be avoided on our account of imperatives.