Modification with De in Mandarin Chinese

Abstract

This dissertation looked at the de construction in Mandarin Chinese, in which the morpheme de connects two meaningful parts and conveys a modification relation between the two. The de construction can be used in both the nominal domain and the verbal domain. In nominal de constructions, the categories of the modifier vary. In verbal de constructions, the modifier can either precede de or follow it. There is no unifying account for all the variations of the de construction in the previous literature. This dissertation adopts a unified approach to de constructions in order to examine the degree to which analyses applied to nominal de phrases can be extended to verbal de phrases. To this end, I focused on the distribution, the headedness and the symmetry of the nominal and the verbal de constructions to see whether we can generalize the de constructions across the two domains. In Chapter 3, I investigated the frequencies of different grammatical categories in the pre- de and the post-de positions of the de phrases found in the Sinica Treebank, a parsed cor- pus of Chinese built by Chinese Knowledge and Information Processing (CKIP). The results showed that the nominal de constructions greatly outnumber the verbal de constructions. The features of the head determine the distribution of their modifiers. Only V and VP are found in the pre-de position in both the nominal and the verbal de constructions. In Chapter 4, I employed several tests to determine the position of the head in both the nominal de constructions and the verbal de constructions. From the test results, we conclude that that the nominal de constructions and the pre-verbal de constructions are headed by the post-de component while the post-verbal de constructions are headed by the pre-de component. In Chapter 5, I applied the de-reflection test to determine whether the syntactic structure iiiand the semantic composition of de constructions are symmetrical or not. The results indicate that the nominal and the verbal de constructions are both asymmetrical. How- ever, the conditions for passing in the nominal domain and in the verbal domain are different, which is an obstacle for generalizing the de constructions across the two domains. In this dissertation, I set out to see whether we could generalize the de constructions across the nominal and the verbal domain. Through the distribution test, the headedness test, and the de-reflection test, I could not find enough evidence for any generalization.