<![CDATA[Pair collaboration might contribute to students’ writing quality, but that effect might differ across students with different levels of writing self-efficacy. This study aimed to explain how different kinds of writing self-efficacy-based pair formations affect the students’ writing quality in the aspect of content, organization, and language use. The participants in this study were college students majoring in English language teaching (N= 135). This experimental study consisted of three phases, where in each phase, different combinations of comparison were investigated through counter-balanced experimental design. The main inferential statistical analysis employed in this study were including Two-way repeated measure ANOVA, Friedman’s ANOVA and Wilcoxon Signed Rank Test to provide evidences on the simple effect of the comparisons. The findings showed that: first, students who write collaboratively in a homogeneous pair produce a significantly better writing quality than those who write individually regardless of their writing self-efficacy level. Second, high levels of self-efficacy students writing individually produce better writing quality than heterogeneous (high-low) pairs. Third, homogeneous pairs of high self-efficacy students outperformed heterogeneous pairs in terms of their writing quality. These findings implied that heterogeneous pairs are recommended for learning purposes. However, homogenous pairs are recommended for assessment to minimize evaluation errors. Consequently, these results advocate for a differentiated pedagogical approach where instructors strategically shift from heterogeneous "growth-oriented" scaffolding during writing instructions to homogeneous "performance-oriented" grouping for summative tasks, ensuring that individual accountability is preserved without stifling the social benefits of peer collaboration.]]>
