Impacts of self-compiled corpus-based instruction on Ethiopian EFL students’ academic writing sub-skills development: A focus on using AntConc software tools

Abstract

This study aimed to investigate the effect of self-compiled corpus-based instruction (CBI) on English as a Foreign Language (EFL) learners' academic writing development, with a focus on specific writing sub-skills and their perceptions of learning through the target corpora. A class of 51 participants from the second-year Software Engineering Department at Woldia University, Ethiopia, was conveniently selected for a 40-hour intervention in 2025. Using AntConc software, a self-compiled corpus of 200 IELTS Writing Task 1 responses was analysed. A mixed-methods approach with an interrupted time series quasi-experimental design was employed. Quantitative data were collected using test scores, and qualitative data were gathered through participant interviews. Data analysis involved paired samples t-tests and one-way repeated measures (RM) ANOVA for quantitative results, and thematic analysis for qualitative findings. The paired samples t-test revealed a significant effect of CBI on each writing sub-skill (p < .01), except for task achievement. Similarly, the one-way RM ANOVA revealed that CBI had a significant effect on developing participants' writing sub-skill quality, with F(6, 50) = 243.659, eta squared = .970, p < .001 for lexical resources; F(6, 50) = 149.459, η² = .952, p < .001 for grammar accuracy and range; and F(3.34, 50) = 322.107, η² = .866, p < .001 for coherence and cohesion. Its effect on task achievement was comparatively less pronounced (F(4.247, 50) = 18.433, η² = .222). Thematic analysis of participant interviews revealed that participants were particularly interested in CBI, especially for learning grammar and vocabulary in both autonomous and collaborative contexts, due to its innovative nature and authenticity. However, challenges remained in learning task achievement, suggesting the need for further investigation. While the findings underscore the overall effectiveness of CBI, future research should explore its application across different academic writing genres.