摘要
考试评分的主观性使评分者偏差成为口语考试关注的焦点。本研究以我校上海大学667位考生为样本，运用项目反应理论的Rasch模型和一致性及偏差分析方法，详细分析了英语计算机口语考试中6位评分者宽严度差异、自身一致性和在考生及考试项目维度上的偏差。研究结果显示，英语计算机口语考试中评分者存在显著宽严差异；评分者虽自身一致性较好，但3位评分者有集中趋势；评分者与考生的偏差呈时间漂移性；两位评分者在两个考试项目维度上发生显著偏差。
Rater bias, the consequence of raters＇ subjectivity, becomes the major concern in the Oral Proficiency Interview. This study, with a sample of 667 examinees taking part in the Computerized EFL Oral Proficiency Interview （COPI） in Shanghai University, aims to analyze in detail the 6 raters＇ severity/leniency, self-consistency and bias on the facets of examinees and items by adopting the Rasch Model in the Item Response Theory （IRT）. The results suggest that in COPI great discrepancies exist in the raters＇ severity/leniency, and that these raters are consistent in rating, though three of them show central tendency. The rater bias on the examinee facet is prominent with regard to time drift and two raters bias against the first two testing items.
出处
《外语界》
CSSCI
北大核心
2010年第5期87-95,共9页
Foreign Language World
关键词
计算机口语考试
评分者偏差
RASCH模型
Computerized Oral Proficiency Interview （COPI）
rater bias
Rasch Model
