Analyzing the Development of a Longitudinal Error-Tagged English Learner Corpus: Exploring Accuracy Development in Iranian EFL Learners' Writing

Abstract

Error-tagged learner corpora are helpful resources in language teaching, providing authentic samples of learners' errors. This longitudinal study aims to investigate accuracy development in three subsequent writing performances of Iranian EFL learners across beginner, intermediate, and advanced proficiency levels. This investigation involved developing and annotating the first error–tagged English written learner corpus for Iranian EFL learners, followed by analyzing the errors. The current corpus includes 35747 tokens from 219 written texts, manually transcribed and annotated based on the latest version of the Louvain Error Tagging Manual. A total of 6917 errors were identified. The developmental patterns of all error categories were detected using potential occasion analysis, specifically focusing on the most frequent error types (i.e., articles, noun numbers, and personal pronoun errors). The results indicated that grammar, lexical, and word redundant/missing/order errors increase significantly as proficiency levels increase. Conversely, form, lexico-grammatical, and punctuation errors exhibited a U-shaped trend, rising from beginner to intermediate levels and declining from intermediate to advanced levels.Additionally, the accuracy of article and noun number usage improved from beginner to intermediate levels but showed little or no change from intermediate to advanced levels, suggesting that higher proficiency levels did not lead to much improvement in this area. However, there was a significant decrease in personal pronoun accuracy from beginner to intermediate levels, followed by a slight increase from intermediate to advanced levels. This study reveals error patterns across different proficiency levels, offering guidance for teachers to adapt their writing instruction methods and enhance learners' writing accuracy.