Vocabulary Use and Development in a Corpus of Japanese Learner Blogs

Abstract

This study investigates issues related to vocabulary use and development via the construction and analysis of a corpus compiled from the writing on blogs of lower/intermediate-level Japanese learners of English studying at a university in Japan. Using a hybrid corpus-based/corpus-driven approach, as well as quantitative and qualitative means of analysis, this research reveals the high frequency words used by this group of learners in a specific context and examines both how their use of these words is different from that of native or otherwise more proficient speakers and how it may have changed over time. The main product of this research is the Japanese Learner English Blog Corpus (JLEBC), a collection at the time of this writing of approximately 15,000 blog entries totaling over 3 million running words of text written over a two-year period by 725 learners. In its design, which also incorporates data obtained from interviews and questionnaires, it addresses several gaps in current learner corpus research related to proficiency, medium/genre, longitudinal data, learner control, size, and task/learner variables. A multifaceted analysis of this corpus suggests first that these learners used a large number of high frequency words either substantially more or less than might be expected in writing by native or otherwise more proficient users of English, for reasons possibly related to collocation confusion, grammatical issues, and the influence (or lack thereof) of topic and/or learning environment, among others. Second, these learners also used high frequency words corresponding to English-based loanwords in Japanese at a disproportionate rate, albeit not to the degree estimated in previous studies. Third, the blog entries by different groups of these learners show evidence of improvement in use of words over time most notably in cases where opportunities for output were numerous and interest in English writing by the learners involved was strong. However, this evidence of improvement is limited overall, at least as measured by tools designed to focus on lexical range and variation. The implications of these results are discussed, a main one of which is that quantitative analysis involving learner corpora can only examine development in use of words in limited ways.