Agreement Errors in Learner Corpora across CEFR: A Computer-Aided Error Analysis of Greek and Turkish EFL Learners Written Productions

This paper illustrates the use of learner corpus data (extracted from Cambridge Learner Corpus – CLC) to carry out an error analysis to investigate authentic learner errors and their respective frequencies in terms of types and tokens as well as contexts in which they regularly occur across four distinct proficiency levels, B1-B2; C1-C2, as defined by Common European Framework of Reference for Languages (henceforth CEFR) (Council of Europe, 2001). As a variety of learner corpora compiled by researchers become relatively accessible, it is possible to explore interlanguage errors and conduct error analysis (EA) on learner-generated texts. The necessity to cogitate over these authentic learner errors in designing foreign language learning programs and remedial teaching materials has been widely emphasized by many researchers (see e.g., Juozulynas, 1994; Mitton, 1996; Cowan, Choi, & Kim, 2003; Ndiaye & Vandeventer Faltin, 2003; Allerton et al., 2004). This study aims at conducting a corpus-based error analysis of agreement errors to reveal the related error categories between Greek and Turkish EFL learners, the distribution of agreement errors along the B1 - C2 proficiency range according to CEFR, and the distribution of agreement error types in respect of the L1 of the learners. The data analyzed in this study is extracted from the Cambridge Learner Corpus (CLC), the largest annotated test performance corpus which enables the investigation of the linguistic and rhetorical features of the learner performances in the above stated proficiency bands. The findings from this study reveal that, across B1-C2 proficiency levels and across different registers and genres, the most common agreement error categories by the frequency in which they occur are Verb Agreement (AGV), Noun Agreement (AGN), Anaphor Agreement (AGA), Determiner Agreement (AGD), Agreement Error (AG), and Quantifier Agreement (AGQ) errors. This study’s approach uses the techniques of computer corpus linguistics and follows the steps of the Error Analysis framework proposed by Corder (1971): identification, description, classification and explanation of errors