Relationships between measures of phraseological complexity and writing quality in a CEFR assessment context

The present study contributes to understandings of the relationships between bigram and collocation complexity and writing quality by analysing a corpus of student placement tests in the UAE. At the heart of the study lies the need to understand the relationships between bigram and collocation diversity and sophistication. In developing such an understanding, the study extends work in this area by examining an underrepresented CEFR grading context. Using correlation analysis, findings indicate that several bigram and collocation measures correlate positively and negatively with essay grades. When built into regression modelling, the 3 predictors of: number of bigram types, Mean MI bigram type, and the number of non-collocation noun + noun bigram types emerge as significant measures that predict grade variation. The implications of these findings for assessment practices in ELF contexts are discussed.