Clause connectives are linguistic forms that convey relations between clauses in a sentence. They are typically compared, whether within a language or across languages, on the basis of a shared syntactic category, such as coordination and subordination, or a semantic one, such as sequence, cause, or addition. While the inadequacy of the coordination/subordination distinction for the purposes of clause linkage typology has received much attention, the nature of the semantic distinctions has received comparatively little. In this dissertation, I investigate polysemous connectives--connectives that express a range of related semantic and pragmatic uses. I argue that adequate comparisons require categories such as cause and sequence to be defined relative to a finely articulated semantic-pragmatic framework able to capture both event-structural as well as discourse-structural distinctions. The argument is based on a detailed analysis of four polysemous connectives in Korean, -ese, -unikka, -taka, and -myense, which cover a wide range of meanings, from temporal relations such as sequence and simultaneity, to causal, conditional, and concessive relations.
The proposed semantic-pragmatic framework is developed through the analysis and comparison of Korean -ese and -unikka, which have been considered similar in both having sequential and causal meanings, but are nonetheless used very differently. For instance, causal -unikka is generally more flexible than causal -ese in the kinds of causal relations it can express, which include epistemic or speech act causal relations. However, -ese rather than -unikka is preferred for expressing causal links between successive real-world events. Independent analyses of each connective's entire polysemy network show that while -ese's different meanings arise from multiple ways of conceptualizing two events as a single complex event, -unikka's meanings stem from different aspects of the relationship between a subjective viewpoint and the content that is observed. Facilities for modeling both of these areas of meaning are combined into an integrated framework. The resulting analyses, in addition to expanding empirical coverage, explain why these connectives, with apparently similar meanings, are used in very different ways.
This framework is then applied to the comparative analysis of Korean -taka and -myense, which are typically used to express interruptive and simultaneous relations between events, respectively. The connective -taka also has successive, cotemporal, and causal uses, as well as a predictive conditional use, while -myense also has cotemporal, additive, and concessive uses. Differences in volitionality are observed to be crucial to the semantic characterization of the two connectives. The full range of facilities integrated into the framework is shown to be needed to account for the factors that condition both connectives' various senses and behaviors. For example, both connectives allow past tense to be marked in their dependent clauses. However, -taka with past tense marking expresses succession rather than interruption, while -myense with past tense marking expresses concession rather than simultaneity. Other issues include the greater flexibility of interpretation afforded by both connectives' nonvolitional uses, conditional -taka's ability to make only undesirable predictions, and the connectives' violable subject identity constraints.
The final chapter examines the syntax of these Korean connective constructions from the perspective of a multivariate approach to clause linkage typology. The connectives -ese, -unikka, -taka, and -myense, differentiated by their various meanings, are assessed according to 13 independent variables. A number of variables examine the ways in which the constructions limit the scope of markers in the main clause that indicate illocutionary force, tense and status, or negation. There are also variables that capture properties of the dependent clause, such as whether it allows for the marking of illocutionary force, or whether the dependent clause allows question words or for a constituent to be extracted. The investigation reveals a number of trends with respect to the syntactic correlates of sense distinctions. Certain sense distinctions, such as differences in aspectual alignment and volitionality, do not yield differences in syntax. However, there are other distinctions, such as the difference between content relations and grounding relations, that seem to correlate with significant differences. The dissertation concludes with discussion on the methodological implications of the study for a cross-linguistic typology.
