Linguistic features that impact essay scores: a corpus linguistic analysis of ESL writing in three proficiency levels

Abstract

This study, which is a part of a larger project on learner corpora, used a corpus linguistic approach to analyzing ESL writing produced by freshmen college students from five private schools in Metro Manila, Philippines, in their first week of classes. The data for the study includes (1) essay scores used to classify texts by levels of proficiency, (2) computer tagging results, and (3) MANOVA, ANOVA, and regression analyses. The analyses revealed that most of the general text characteristics (e.g. fluency, unique words, word per sentence, number of sentence, number of paragraphs) and lexical (except conjuncts), clause-level, and grammatical features all have a steady increase across proficiency levels. This implies that the (1) writers of high score essays (level 2 and 3) are likely to employ more of these structures, and (2) The employment of these structures impact the raters’ assignment of scores on the essays. The findings of the present study lead to tentative conclusions and implications for curricular enhancement and research.