Comparisons of L2 Learners’ Written Performance by Two Types of Error Analysis

Abstract

This study compares the results of linguistic errors of Korean EFL college learners’ corpus with two types of error analysis. 86 female college students participated in the study, and wrote a narrative essay through the Criterion®. After their essays were rated by e-rater®, they were divided into three proficiency groups according to writing scores and then linguistic errors were classified, based on Ferris’ (2002) error classification. All linguistic errors were analyzed by two different frames such as conventional analysis (i.e., the number of each error type and its percentage) and T-unit analysis (i.e., the number of each error type was divided by T-unit). Through the comparison of two error analysis, morphological errors were the most frequent, followed by lexical errors across writing proficiency levels. Syntactic errors were the least frequent among these three main types of errors. The same results arose from two different error ratio measures, but error distribution across writing proficiency levels differed. In particular, error ratio measures with T-units yielded more accurate results when analyzing error distribution across writing proficiency levels. The findings suggest that error analysis through this new ratio measure presents more verifiable results than through conventional analysis.