Classifier constructions of Turkish

This dissertation addresses three interrelated issues concerning classifier constructions in Turkish: (i) semantics of bare noun phrases, (ii) the function of classifiers and syntactic constituency of classifier constructions, and (iii) syntax and semantics of number marking. Regarding the first issue, it is argued that Turkish bare noun phrases denote properties of entities but are inherently ambiguous with respect to whether they take kind or object entities as their domains. With respect to the second issue, it is argued that classifiers serve to restrict the ambiguous domain to object individuals only. Syntactically, a measure phrase is proposed which hosts classifiers and measure words in its head position and numerals in its specifier position. The resulting constituent, a property-denoting µP, is then applied to nouns predicatively or attributively through adjunction. It is demonstrated in particular that µP only combines with number-neutral bare nouns. As for the third issue, it is demonstrated that number specification in Turkish is relevant for DPs only, and is expressed in the φP that projects above the DP. Sub-DPs are shown to lack number specification. Number markers on bare nouns are uninterpretable agreement reflexes which must be checked against interpretable counterparts in the φP or in some other functional head. As they lack number specification, sub-DPs always receive a number-neutral reading, and thus are restricted to contexts which can handle such interpretations. DPs, on the other hand, can never receive number-neutral readings. Overall, the proposed number system is shown to have a broader empirical coverage crosslinguistically.