Towards more valid scoring criteria for integrated reading-writing and listening-writing summary tasks

Despite the increased use of integrated tasks in high-stakes academic writing assessment, research on rating criteria which reflect unique construct summary skills is comparatively rare. Using a mixed-method approach expert judgement, text analysis, and statistical this study examines features that discriminate summaries produced by 150 candidates at five levels proficiency reading-writing (R-W) listening-writing (L-W) tasks. The judgement revealed wide range discriminated R-W L-W responses. When responses were coded these features, significant differences obtained seven including relevance ideas, paraphrasing skills, accuracy source information, style, language control, coherence cohesion, task fulfilment across task. same did not yield levels. findings have important implications for clarifying different modalities, indicating possibility expanding categories with some potential translating identified into automated systems. results indicate need developing descriptors can more effectively