This project asks what it takes for a predicate to embed both declarative and interrogative clauses. The central issue here is why know embeds both types of clauses whereas the closely related believe does not (e.g. John knows / believes that Mary smokes} vs. John knows / *believes who smokes). The project addresses this issue by investigating a novel semantic theory of clause embedding. It is based on the hypothesis that the mechanism responsible for the exhaustive interpretation of an embedded interrogative also restricts the distribution of that same interrogative. Interrogatives under believe, on the one hand, lead to a contradiction and thus unacceptability. Interrogatives under factive predicates like know, on the other hand, do not give rise to such a contradiction making the sentence acceptable. Crucial evidence for the proposal comes from predicates like be certain whose embedding properties are seemingly affected by the linguistic context they appear in.