This thesis addresses theoretical aspects of Cantonese and Mandarin Verb-Object Compounds (VOCs). While most studies in the past approached the topic by discussing the wordhood status of VOCs and the validity of various criteria in determining their status, this research explores key aspects of the syntactic and morphological behaviours of Cantonese and Mandarin VOCs.

There are three main objectives for this thesis. The first objective is to show that although, syntactically, a VOC can be interrupted in certain contexts in which the verbal constituent acts as a predicate and the nominal constituent is a syntactic object, semantically, and in some morphological aspects, the VOC is still like a single lexical item with unitary lexical semantics. This study defends the view that VOCs exhibit both syntactic and morphological properties. Based on this, a novel interpretation for the discontinuous VO compounds as idiom-like expressions is provided.

Taking into consideration the fact that most studies of VOCs were done on Mandarin and nothing in detail has been attempted on Cantonese VOCs, the second aim of this thesis is to look at the Cantonese VOCs in detail, together with the Mandarin VOCs, and to make a comparison between them with regard to the lexical, syntactic, and phonological aspects. It is found that Cantonese and Mandarin VOCs reveal both similarities and differences in these two aspects.

Based on these descriptions of VOCs we then offer a revised formal representation of VOCs within the syntactic framework of Lexical-Functional Grammar (LFG). A critical review of two studies (C.-R. Huang 1990, Her 1997) that have dealt with the representation of VO constructions in LFG and the proposal of making a distinction between Verb-Object compounds and Verb-Object phrases through the argument structure and the use of constraining equation are presented.

The LFG framework is shown to be a plausible syntactic model to account for the kind of lexical discontinuity observed within Cantonese and Mandarin VOCs. The results of the research developed here could have practical applications and implications for the field of computer technology for the development of large-scale computational lexicons