The syntax and pragmatics of embedded yes/no questions

The paper investigates the distribution of English if and whether as complementizers for polar questions in extensional contexts, and specifically the restriction of if-questions to negative contexts. I will criticize the recent analysis by Adger and Quer (2001) and suggest that markedness of embedded ifquestions in certain contexts arises through the conspiracy of several factors. I follow Bolinger (1979) in assuming that the question “if S” presupposes an (ir)relevance asymmetry between the proposition denoted by S and its negation. I argue that this leads to competition between embedded if-questions and that-clauses in the relevant constellations, which in turn excludes the if-question as the less optimal variant. 1 Embedded questions In English, like in many other languages, questions can occur as subordinate clauses under a matrix verb. The matrix verb can be one of questioning, doubting or wondering, like in example (1). The embedded clause here denotes a question and reports that the subject poses or could pose the question, mentally or explicitly, to himself or to others. (1) Susan wondered/asked/did not know who was the murderer. Susan wondered/asked/did not know whether Jones was the murderer. This kind of example, also called extensional question embedding, contrasts semantically with examples where the subject holds a certain mental attitude towards the correct answers of the embedded question, like in (2): (2) Elsie knew/told me/discovered who took the book. Elsie knew/told/disclosed whether Jones took the book. These predicates have also been called intensional question embeddings. Finer classifications of question embedding verbs have been proposed in the literature (Ginzburg 1996, Lahiri 2002 Ginzburg and Sag 2000), starting from seminal work by Karttunen (1978). The present paper is focused on embedded yes/no questions in intensional contexts, and more specifically on the distribution of the complementizers if and whether in such embeddings. I will argue that these complementizers carry different presuppositions and that intensional question embeddings compete with embedded that clauses. It will be shown that the distribution of if and whether in positive and negated question embedding contexts can be derived from pragmatic factors. This counters a recent proposal by Adger and Quer (2001) who offer a feature-based analyses of the