This talk addresses the much-discussed issue of capturing what is common and what is not across declaratives and interrogatives by focusing on assertions and questioning acts performed when uttering such sentences. I will show that the properties of canonical assertions and questions can be derived from the basic discourse effects of uttering declaratives and interrogative sentences proposed in Farkas and Roelofsen 2017. I will also claim that non-canonical assertions and questions should be treated as departures from the canonical cases. Such non-canonical speech acts arise either because of the speech context overriding default assumptions (such as in the case of 'quiz' questions or rhetorical questions in English) or because of special markers that trigger special discourse effects (such as biased questions in English or oare-marked questions in Romanian). In the case of special conventional discourse effects markers, a new problem that arises is how to predict what type of sentences they combine with, a problem that parallels the issue of predicting complement selection for predicates. The ideal solution is to derive the distribution of such markers from the discourse effects they trigger.