This work is a study of certain Turkish syntactic constructions, such as the reflexive, the passive, and the causative, and their interaction with one another. In particular, it is a well-known fact that the passive and the -In reflexive constructions cannot occur in the complement clause of the Turkish causative. However, no unified analysis within Turkish grammar has heretofore been presented to account for this fact. I argue that the structures I propose for the -In reflexive and the structure of passive share a common feature; thus, these structures can be ruled out from the causative complement clause by a single generalization. In the course of positing the structures for the above construction, various theoretical issues are discussed: for example, the necessity of the notion of coreference in reflexive structures vs. multiattached structures which do not mark coreference, and the lexical vs. syntactic nature of the causative construction.

This study was conducted within the framework of Relational Grammar (RG) (Perlmutter and Postal; 1974, 1977, 1983, 1984) which takes grammatical relations and multiple syntactic levels to be primitives of the theory. The structures posited for the constructions in Turkish have also been proposed for languages unrelated to Turkish; thus, the structures not only account for data specific to Turkish, they are representations of possible constructions in human language.

Many of the assumptions and hypotheses of Relational Grammar are upheld by the grammar of Turkish. For example, Chapter Five presents three syntactic pieces of Turkish-internal evidence for the Unaccusative Hypothesis. There are two constructions in Turkish, however, which violate one of two major assumptions within the theory: the 1 Advancement Exclusiveness Law or the advancement analysis of passive. Turkish provides no evidence as to which of the assumptions should be abandoned. The constructions which violate one of these assumptions include the impersonal passive of an initially unaccusative predicate and the impersonal passive of a personal passive. Alternative analyses of the apparent impersonal passive of superficially intransitive predicates, which do not counterexemplify either of the major assumptions of RG, are considered. I show, however, that these alternative analyses, which claim that Turkish does not possess true impersonal passives, cannot be maintained in light of the complications they create elsewhere in Turkish grammar. Thus, the alternative analyses are inferior to the analyses I propose; consequently, one of the two aforementioned assumptions of RG must be abandoned as a universal.