The structure of relative clauses in Turkish: The strategy for election between -dik and -an morphemes

As one of the most apparent examples of natural language's property of infinite recursiveness, Relative Clauses have been intensely attracting the attention of linguists working in the field of Generative Grammar since the dawn of the theoretical approach. Directed by the general postulates of the theory at hand, linguists have considered the participial Relative Clauses like those found in Turkish to be equal to their English finite counterparts and have submitted them to wh-movement analysis. In this thesis we came to the conclusion that the application of the wh-movement approach developed in Generative Grammar when handling nominalized Relative Clauses has hindered the proper description and analysis of those constructions. Therefore, while aiming at providing a new explanation for the choice between the two strategies for producing Relative Clauses in Turkish, we tried to develop a new analysis that pivots around the process of nominalization. With this aim in mind, we inspected the data and the findings that we had collected from previous analyses on the subject and we determined that the difference between the two strategies is not in the grammatical function of the relativized element, but it the semantic status of the logical subject that stays inside the clause. Based on that observation we singled out two starting points for our analysis: the notion of subject and nominalization as a subordination strategy. As a result of the detailed discussion on the notion of subject, we determined that one should separate what is usually known as 'subject' into at least four, independent of each other, grammatical functions. After we introduced those four different grammatical functions and we established their places in the X' schema, we turned out attention towards nominalizations in order to pin down which of the four possible grammatical functions can be found in such constructions. To be able to determine the structure of the nominalizations at hand, we used The Mixed Extended Projections Hypothesis and mapped the verbal and nominal functional projections in the two Turkish Relative Clauses. As a result, we concluded that only the [-DIğ+I] type of relativization strategy projects up to a DP layer. On the basis of this conclusion and the application of the Split DP Hypothesis we determined that the difference between the two relativization strategies in Turkish can be summarized as follows: When the subject of predication and the logical subject are one and the same element, the language switches to the [-DIğ+I] strategy for producing Relative Clauses, which allows that element to stay in the clause and check its features. We believe that the theoretical discussion conducted in our thesis paves the way for future research in the fields of phrase structure, sentence structure and mixed categories.