This talk starts from the following two assumptions: First, that sentences are associated not only semantic content but also conventional discourse effects – specifications of how the sentence affects the discourse structure against which it is uttered (see Condoravdi and Lauer 2012). Second, that sentence forms in a language divide into unmarked forms, whose conventional discourse effects are determined purely by their semantics, and marked forms, whose conventional discourse effects are partly determined by their semantics and partly by their special form (see Farkas and Roelofsen 2017). The hypothesis to be explored in this talk is that the conventional discourse effects of some marked sentence forms signal departures from the prototypical speech act associated with that form. In the case of interrogatives, the marked forms to be discussed signal that the assumption of Speaker ignorance is absent (non-neutral questions), or that the assumption of future Addressee engagement with the question is absent (`deliberative’ questions). In the case of declaratives, the marked form to be discussed signals that the assumption of Addressee competence is absent (tentative assertions). Special attention will be given to non-intrusive questions in Romanian (marked by oare), and tentative questions and assertions in German (marked by wohl, discussed most recently in Eckardt 2017).