A Corpus-based Contrastive Analysis of Spoken and Written Learner Corpora:

1. Introduction The purpose of this research is to investigate the variability of interlanguage that has been claimed in previous study by means of a corpus-based quantitative analysis. It aims to observe the style shifting of various grammatical features and word formation errors by tagging errors in 297 learners' data. Various studies have been undertaken to describe and explain the process of second language (L2) acquisition. Tarone (1983), for example, claims that interlanguage capability continuum of learners diverges with respect to the degree of attention to language form, so-called 'careful' style to 'vernacular' style. Additionally Tarone (1985) showed that learner's performance varies depending on the dissimilarity of task, whether a written grammar test, oral interview and oral narrative. Ellis (1987) has confirmed the style shifting in the L2 learners' use of the past tense. The variability of accuracy in past tense morphemes was observed in his study, when different amounts of planning time were set for a single narrative discourse task. Through his examination Ellis (1987) concluded that " so-called 'natural' order may not be a stable phenomenon " (p.1). In this study I compared the features of variability in L2 written and spoken corpus data based on the hypothesis that processing mode of learners affects their performance in L2. This research focussed on the analysis of the same task that used different production mode. The similarities and differences of learners' errors in each mode were examined mainly from the perspective of grammatical and word formation features. In addition, English proficiency level of learners is another factor added to this study. Although this addition was only possible for the spoken corpus data, we can still observe the performance of learners at different proficiency levels.