This dissertation provides an explicit syntactic and semantic account for a reasonably large sample of question constructions in Swedish. Within generative grammar, the existence of non-local dependencies as in constituent questions has been taken as evidence for the need to postulate transformational rules in the grammar of natural languages. Recently a number of linguists have proposed ways of handling such dependencies without transformations. Until now, these proposals have been based on English. In this study, we investigate the possibility of extending non-transformational approaches to languages like Swedish where question formation differs from English in a significant way. In Swedish, more than one constituent can be extracted from a clause. We discuss the consequences of this fact for transformational and non-transformational approaches to Swedish. It is shown that the non-transformational approaches need to be substantially modified in order to provide a syntactically and semantically adequate grammar for Swedish. The implications of these modifications are assessed from the point of view of choosing between grammars. The main part of the dissertation consists of an analysis of the semantics of constituent questions. We propose an extension to the semantics for questions in the framework of Montague grammar given by Hamblin and Karttunen. Most current approaches to questions take the entire question phrase to be the interrogative quantifier. We point out that these approaches are not adequate for questions where the interrogative phrase contains an anaphor bound from inside the sentence. In addition, these approaches cannot account for all readings of temporally ambiguous sentences. To allow the semantic rules to handle such cases as well, a more general approach to questions is proposed. On this approach, only the 'which' part of the question phrase constitutes the interrogative quantifier. This quantifier ranges not over individuals directly, as in the previous theories, but over functions that pick out sets of individuals. In simple questions, the result of the proposed analysis is tantamount to the results on earlier approaches. However, it is shown that only the proposed approach can generalize to more complex questions. The analysis proposed here is compared to current approaches to questions within transformational grammar. Finally, we discuss the relative merits of a structurally based and a semantically based approach to anaphoric relations.