Variation in Metonymy: A Corpus-based Cognitive Linguistic Approach.

This dissertation presents an investigation of variation in metonymy in the framework of Cognitive Linguistics, focusing specially on variation from an onomasiological perspective, i.e. a change of naming/designation with respect to a particular concept. The principal objective is to provide a systematic way of disclosing how certain bodily and/or cultural-social factors exert an influence on the metonymic conceptualization. A secondary objective is methodological, i.e., to demonstrate the importance of including corpus data in metonymy research as well as to devise quantitative statistical methods to explore and visualize the variations. Three case studies are conducted to achieve the objectives.The comparative study compares metonymical expressions for PERSON between English and Chinese. This study is based on Chinese metonymy dictionaries and the Historical Thesaurus of English of OED. Descriptive statistical techniques (e.g. Fisher s exact test) are employed to do the quantification and comparison. It attempts to reveal culturally relevant cross-language variation in metonymy.The diachronic study examines the metonymies for FEMALES in the history of Chinese from the 11th century BC to the early 20th century. Metonymic expressions for FEMALES collected from dictionaries are confronted with the Historical Chinese Corpus for their distributions in different historical periods and genres. Statistical techniques (e.g. multidimensional scaling, Poisson regression) are employed to explore and visualize the historical and stylistic changes in metonymic patterns in the multivariate diachronic data. Significant variations are then inspected against their cultural and social background to detect the potential factors behind the changes in metonymic conceptualization of FEMALES.The lectal study investigates the factors that influence the alternation of metonymic vs. literal designations for the concept GOVERNMENT between Mainland Chinese and Taiwan Chinese. This study is based on a self-built corpus of newspaper articles and online forum postings from the two language varieties. The corpus is tagged for a set of factors, after which the multivariate analysis (specifically, mixed-effects logistic regression) is applied. It aims to show that the alternation is not a question of free variation, but signals a specific lectal stratification of the linguistic community.The dissertation contributes to current metonymy research from the following two perspectives. First, theoretically it shows that the usage of metonymy is not an isolated linguistic or cognitive phenomenon but is culturally, historically and socially contextualized. Second, methodologically it responds to the call of Cognitive Linguistics to adopt a usage-based empirical methodology.status: publishe