Understanding writing quality change: A longitudinal study of repeaters of a high-stakes standardized English proficiency test

This study examined the writing score and writing feature changes of 562 repeat test takers who took the Canadian English Language Proficiency Index Program-General (CELPIP-General) test at least three times, with a short (30-40 day) interval between the first and second attempts and a longer (90-180 day) interval between the first and third attempts. Analysis was conducted to uncover whether changes occurred at different testing durations (short vs. long) and whether the observed changes varied across repeater's initial proficiency groups (low, mid, high). The writing scores measured by CELPIP bands showed great stability over the 6-month period, but the trends of development differed by proficiency group. Low proficiency test takers were more likely to have faster observable score gains, compared to the medium proficiency group, whereas high proficiency repeaters may not maintain their score levels at later attempts. Writing quality was analyzed using natural language processing (NLP) tools. Results suggested that for all proficiency groups, lexical features were more likely to improve over the 6-month period, with some measures showing improvement at 1 month; features in cohesion and syntactic sophistication, however, did not change significantly.