Functional Variation In Interlanguage English Oral Discourse:A Multi-dimensional Study

A multi-dimensional (MDA) study has been conducted on the one-million-word LINDSEI (Louvain International Database of Spoken English Interlanguage) corpus to investigate functional variation in oral discourse of495English learners with eleven different L1. The study finds that interlanguage oral conversation is only marginally similar to counterpart native registers, with deviance observed in the communicative functions actually realised. Interlanguage discourse exhibits distinct patterns of variation, which is described by a five-dimension model:Involved Topic Development, Elaborated and Cohesive On-line Production vs. Stacking of Idea Units, Descriptive Account of Facts and Events vs. Presentation of Personal Sentiment, Personal Narratives, and Explicit, Cooperative Speech vs. Implicit, Spontaneous Speech. Participant L1is not seen of clear correlation with performance variation. For comparison with native texts, the corpus is first investigated of dimension score distribution and linguistic feature co-occurrence in relation to the oral genres in Biber’s (1988) MD model. Four of the six dimensions in the1988model are used. Frequency means of individual features are compared. The results are further investigated using L1as category. The mainstay of the study attempts to describe the functional variation specific to interlanguage oral discourse. A five-factor multi-dimensional model is established out of two candidates after several rounds of piloting tests. The factor model is constituted by27linguistic features of the original tag-set. Dimensions are systematically interpreted according to a practice guideline stipulated as amendment to the MDA framework. Throughout the study, texts are programmatically computed for their dimension scores; ANOVA tests are conducted with Post-Hoc analysis to explore potential L1relevance. The results imply room for improvement in register appropriateness of oral discourse for most learners in the LINDSEI pool. The research sheds light on adaptation of the MDA framework to second language research, with refinement in the MDA methods as well as data presentation. Implications on pedagogy and L1transfer studies have been discussed.The data sheets and source codes for this project can be downloaded at http://scholar.renyan.me. Readers are advised to have at least Microsoft·Windows(?) VistaTM with VBA-enabled Microsoft*Office(?02007installed before accessing the files.