This dissertation explores issues with regards to the syntax of complement functions in Cantonese, a variety of Chinese, from the theoretical perspective of Lexical-Functional Grammar (LFG). The central question we are concerned with is the extent to which LFG is applicable to Cantonese based on the examination of complement functions.

Syntactic properties of various Cantonese complement functions have been presented. The following functions have been identified: OBJ, OBJ_q, OBL_BEN/GO, OBL_AG, OBL_LOC, OBL_TH, COMP, and XCOMP. Each of them differs from one another in terms of the structural configuration, categorial realization, and passivizability. Considerable focus has been placed on the function COMP. Based on previous studies which introduce the concept of mixed languages as the motivation of retaining the COMP function, evidence has been shown that Cantonese is a mixed language. Upon this, evidence has also been adduced to support the idea of introducing degrees of OBJ/COMP mixedness in the description of natural languages.

The work also addresses the limitations concerning the Lexical Mapping Theory (LMT) arising from some typical constructions in Cantonese predicates like passives, ditransitives, and locative inversions. On this account, a revised version of LMT with a rearranged Thematic Hierarchy and two morpholexical rules for passivization and locative inversions have been proposed for the analysis of Cantonese.

Furthermore, an additional pair of argument structure feature [+ p] '(non)propositional' has been proposed for the LMT in the light of distinguishing COMP and XCOMP from other argument functions. This move is thought to result in a more elaborate argument structure theory with three pairs of binary features for argument function classification and a new markedness hierarchy of argument functions. This introduction of a new feature represents an initial step towards maximizing the predictive power of LMT.