This thesis has three primary goals. The first goal is to provide a treatment of bounded discontinuous constituency and word order in general and semi-free word order in particular. Bounded discontinuous constituency is taken to mean the kind of category-bounded discontinuity typically found in the German Mittelfield and in Dutch "cross-serial" dependency constructions. It does not mean unbounded dependencies such as wh-movement. The second goal is to provide an alternative account of cross-linguistic variation in word order (particularly with respect to the West Germanic languages) to the Principles and Parameters approach of Government and Binding Theory (GB). The third goal is to formalise the account in a single homogeneous logical formalism which is not based on rewrite rules or other formal language theoretic machinery. A secondary goal is to formulate generalisations about natural languages and families of natural languages which are primarily motivated by empirical considerations rather than theory internal ones. With these goals in mind, the notion of a word order domain is introduced. Informally, a word order domain is the domain over which category bounded discontinuous constituency and partially free word order can occur. The structure of the theory of syntax and word order then consists of two parts. First, a level of unordered functor-argument structure is posited which is similar to D-structure in GB. Second, a level of word order domain structure is posited which is similar to phonological form (PF) in GB. The theory rejects the conventional assumption made by almost all modern syntactic theory that word order is determined by the order of the leaves of phrase structure trees. Functor-argument structure is responsible for subcategorisation, selection of heads by modifiers and semantic compositionality.