| 研究実績の概要 |
We completed data collection. In total 347 people signed up to participate in the study and 200 completed the procedures. Results of a post-evaluation questionnaire indicated that Japanese L1 students and teachers had significantly less confidence in their ability to rate English L2 speech than other rater groups. Also, a qualitative analysis of reasons for confidence found that Japanese L1 raters were mostly confident based on their effort and ability, whereas non-Japanese L1 raters were confident due to their training and experience as teachers as well as their exposure to different English varieties. Next, an analysis of different vocabulary measures and human ratings found that human ratings are heavily influenced by the number of words produced, so basic measures of lexical density correspond more closely to human ratings than measures that control for text length. Also, different rater groups show similar correlation patterns with lexical density measures, although severity of judgment may vary between groups. Finally, looking at factors affecting ratings, a RASCH analysis followed by Linear Mixed Effects modelling indicated that English L1 raters gave significantly higher ratings than non-English L1 raters. Also, Japan-based raters gave higher scores than Canada-based raters, but there were no differences between student and teacher raters. In addition, scores were lower when the rater and the speaker had the same L1, but higher when the rater lived in a country where the speaker L1 is primarily spoken.
|
