Abstract

This dissertation demonstrates how advancements in corpus approaches to linguistic inquiry can be used to improve the methodological rigour, reliability, and general usefulness of findings in various areas of Second Language Acquisition (SLA) research. Although these studies primarily focus on improvements in areas where corpus approaches are already commonplace, this dissertation also demonstrates how corpus methods can be usefully applied to new areas. Through the use of these methods, the presented studies highlight issues learners face when attempting to gain proficiency in second language (L2) English.
Study 1 investigated the usefulness of transitional probability as a way of improving the extraction of formulaic sequences (e.g., on the other hand) from large scale corpora. Since current methods of identification often lead to lists of overlapping structures that lack psycholinguistic validity and pedagogical usefulness (Liu, 2012; Nekrasova, 2009; Simpson-Vlach & Ellis, 2010), this study evaluated the effectiveness of a new statistical measure in this area, transitional probability, as a way of improving the psycholinguistic status of corpus derived formulaic sequences. Using a sequence completion task, results revealed that corpus derived formulaic sequences with higher transitional probabilities were more accurately completed by first language (L1) and L2 English users, leading to the conclusion that these sequences are more likely to be stored as prefabricated units.
Study 2 used a corpus approach to investigate the relationship between L1 background and the lexical choices made by L2 English writers. Looking specifically at L2 English writers of L1 Arabic, Chinese, and French backgrounds, a corpus of 150 argumentative essays written as part of an English for Academic Purposes program at a large English-medium university in North America was used to identify production tendencies in the use of linking adverbials by each L1 group. Results revealed important L1 differences for the use of specific linking adverbials and broader functional categories.
Study 3 investigated lexical dimensions of L2 English speech associated with differences in perceived linguistic ability as judged by naïve L1 English raters. Using a corpus of transcribed speech samples from 97 L2 English users across two tasks (194 speech samples), naïve L1 English raters evaluated each sample for perceived comprehensibility and nativeness. Variables associated with factors related to dimensions of lexical density, sophistication, and diversity were targeted for potential correlations with L1 rater judgements of each construct. Results indicated important linguistic measures significantly correlated with each construct as well as task-based differences.