Variation in English world-wide : varieties and genres in a quantitative perspective
MetadataShow full item record
This dissertation takes a quantitative perspective on variation in English world-wide. It applies a feature-aggregational method to a corpus of 7,309 texts, representing ten national standard varieties of English and comprising a range of oral, written, and computer-mediated genres of communication. The study is designed as a testing ground for the adequacy of different models in World Englishes research. By virtue of drawing on data from the social networking service Twitter, the project also offers insights into the relationship between computer-mediated discourse and more established communicative genres. My analysis gives a unified account of different determinants of variation – the geographic and the generic-situational – and systematically compares their effects. I take inspiration both from variationist sociolinguistics and the text-linguistic view common in corpus research and demonstrate a method that brings these two traditions into productive conversation. Exploratory factor analysis (EFA) is used to develop ten latent dimensions of variation from 236 individual linguistic variables whose relative frequency is extracted for each corpus text. The resulting factorial structure relates systematically to both generic and geographic dimensions of variation. However, the former are established as significantly more powerful predictors of variation than the latter. Properties of genres and sub-genres at different levels of granularity are explored along the ten dimensions of variation, and relationships among varieties and groups of varieties are quantified. The dimensions developed through EFA are then applied to a new set of data, with a shift of focus from the multi-feature exploratory view to a contextualized analysis of a single case of variation. The competition between established and innovative syntactic options for because-complementation is modeled in a logistic regression analysis. Score estimates for the factors developed in EFA are included as predictors of variation in because-complementation. I demonstrate that this analytical choice improves both the goodness-of-fit and the theoretical insights generated by the model. This project emphasizes the importance of conversation and collaboration among different approaches to language variation and makes a case for methodological pluralism in the study of language variation.