Predicting inappropriate source use from scores of language use, source comprehension, and organizational features: A study using generalized linear models

This paper examines the extent to which inappropriate source use – verbatim source use and patchwriting – can be predicted by scores of other textual features that are commonly evaluated in second/foreign language (L2) integrated writing assessment. 246 advanced-level English as a Foreign Language (EFL) test-takers enrolled in a Chinese higher education institution provided integrated essays that required both summary and argumentation. All the collected essays were rated by two experienced raters and checked for interrater reliability by way of generalizability theory. Then, a series of generalized linear models was compared to identify the best-fitting model that explained the relationship between the independent variables and inappropriate source use. Results indicated that the zero-inflated beta-binomial provided the best fit to the data, with approximately 43.67 % of the data estimated to be an extra zero. Next, parameter estimates of this model included (1) non-significant effects of language use and source comprehension on inappropriate source use and (2) a significantly negative effect of organizational features on the dependent variable. This suggests that focusing on organizational features, operationalized herein as organization, coherence, development of ideas, and authorial voice, can help L2 test-takers reduce reliance on inappropriate source use. Implications for research and practice are discussed.