Toward a Transparent Construct of Reading-to-Write Tasks: The Interface Between Discourse Features and Proficiency

Abstract

As a growing number of testing programs use integrated writing tasks, more validation research is needed to inform stakeholders about score use and interpretation. The current study investigates the relationship between writing proficiency and discourse features in an integrated reading–writing task. At a Middle Eastern university, 136 undergraduate students completed a reading-based writing task. The essays were holistically scored by two raters and then classified into three proficiency levels. In addition, the essays were analyzed for a number of discourse features, including fluency, lexical sophistication, syntactic complexity, grammatical accuracy, verbatim source use, and direct and indirect source use. A one-way analysis of variance was employed to look into the relationship between writing proficiency and the discourse features of interest. The results yielded significant differences across proficiency levels for a number of discourse features. Nonetheless, follow-up comparisons indicated that the differences were greater between the lowest level and the two upper levels. As for the upper levels, no statistically significant differences were found between these two levels for most of the discourse features. The implications of the study suggest that the selected discourse features play a major role at lower levels, whereas other textual features, such as cohesion, content, and organization, are more critical at higher level writing. The results also support the need in a construct of integrated writing for the inclusion of reading proficiency and knowledge about discourse synthesis.

Notes

¹By no means an exhaustive list of discourse features.

²The term T-unit was first introduced by Hunt (1965) and refers to the shortest unit that can stand alone as a sentence.

³Brown–Forsythe formula and Dunnett's C statistics do not assume equality of variance, thus they were used in this study where the equality of variance assumption is violated to verify the ANOVA results.