Exploring academic writing through corpus linguistics: When discipline tells only part of the story

The primary goal of this dissertation is to investigate the linguistic characteristics of registers published in academic journals, taking into account the varied realizations of research articles in fundamentally diverse disciplines. That is, the study seeks to go beyond the traditional and often one-dimensional analysis of a generically-defined research article to first distinguish between different types of articles (or registers) within and across disciplines, and then to describe those registers according to their characteristic linguistic and non-linguistic features. A secondary goal is to explore the various approaches to studying linguistic variation available to researchers through corpus linguistics methodologies, and to illustrate how corpus analytical approaches to linguistic variation can offer multiple, complementary perspectives about language use.;The research in this dissertation is based on a corpus of 270 research articles from six disciplines: philosophy, history, political science, applied linguistics, biology, and physics. Research articles within these disciplines are further categorized by journal register: theoretical, qualitative, and quantitative research articles. An analysis of both the non-linguistic and linguistic characteristics of the corpus is undertaken in the study. In the analysis of the non-linguistic features of these texts, a framework for describing the situational characteristics is first proposed and then applied to each text in the corpus. The linguistic analysis relies on quantitative and qualitative analyses of data extracted through several specialized computer programs, and includes a grammatical survey of the distributions of core grammatical features, a description of syntactic features which 'elaborate' and 'compress' discourse, and a statistical analysis that identifies co-occurrence patterns of 70 linguistic features.;The results of these analyses show that linguistic variation does not occur along a single parameter, but rather across multiple situational parameters. On one hand, some linguistic features are shown to vary along disciplinary lines, and more generally along the parameters of traditional discipline groupings (such as the humanities, social sciences, and hard sciences). On the other hand, these studies also highlight variation that occurs irrespective of discipline, instead seeming to follow along parameters related to situational characteristics such as the purpose of the research, the nature of the evidence used in the research, and so on. The implications of these results are discussed in terms of our understanding of disciplinary practices, the multidimensional nature of linguistic variation in academic writing, and the methodological considerations required for corpus studies of academic writing.