Rising declaratives (RDs) pose a crucial test case for theories of the pipeline from syntactic categories to semantic categories to illocutionary acts: they’re syntactically declarative, but they comprise askings. This could be because they have a special semantics—they denote questions despite being syntactically declarative. And it could be because they comprise special context updates—they withhold speaker commitment despite denoting a proposition. It’s hard to find hard-and-fast evidence to distinguish between these in root contexts. But RDs can embed under rogative verbs, suggesting that they denote questions (Farkas & Roelofsen 2017). This paper revisits embedded RDs with closer scrutiny, and comes to the opposite conclusion. It turns out that embedded RDs tell us more about the semantics of quotative complements to verbs of speech, and about the lexical semantics of rogative verbs of speech, than they tell us about the semantics of RDs.
RDs can embed under antirogative manner-of-speech verbs. And they can’t embed under rogative verbs that aren’t verbs of speech. So their distribution does not actually track the rogativity of the embedding verb. A true generalization is that RDs can only be embedded under verbs of speech, and only as quotations. Quotative complements to verbs of speech do not have the same syntactic or semantic representation as non-quotative complements, and do not compose with verbs of speech by supplying them with the kinds of denotations that are supplied by their non-quotative complements; quotative speech reports have a different thematic structure than non-quotative speech reports. Therefore rogative verbs of speech cannot impose a restriction on their quotative complements that the quoted sentence denote a question. Rather, quotative complements involve demonstrative reference (Davidson 1979) to a performance which combines with verbs of speech via a demonstration relation between events and performances (Clark & Gerrig 1990, Davidson 2015). The only restriction that applies to quotative complements to the verb ask is that their utterance comprise an asking. It is not necessary that an asking be an utterance of a sentence that denotes a question. The most crucial desideratum that must be met by any theory of RDs on which they denote propositions, of which there are many, is providing a well-defined sense in which utterances of RDs perform askings despite RDs not denoting questions. So there is no argument from embedded RDs that they denote questions; the facts are all predicted by theories on which they do not.