Analysing a Longitudinal Corpus of Learner English : Focus on Number-related Errors

This MA dissertation investigates the development of grammar accuracy among French-speaking learners of English. More precisely, grammatical errors related to number and produced by 20 students enrolled in a bachelor's degree program in modern languages and literatures at UCL during their first, second, and third academic year were thoroughly examined. A total of 60 argumentative essays were stored in the LONGDALE database, a learner corpus compilation project. The present master dissertation consists of a theoretical part and of a practical part. The theoretical part first sets out to describe the main grammatical approaches that have been favoured in SLA research and the evolution of methodologies used to examine morpheme acquisition in L2 English (including the morphemes third person –s and plural -s). It then aims to highlight important aspects related to Error Analysis (EA), including the evolution in the concept of learners’ errors and the different methodologies used to investigate them: Contrastive Analysis (CA), Traditional Error Analysis (TEA), and Computer-aided Error Analysis (CEA). The practical part of this study reports the results of the error analysis. The first section is aimed at determining which number-related errors are most common and thus, most problematic for learners. It then seeks to check the extent to which the results concerning the verb and the noun corroborate those obtained in a previous longitudinally-based study (Malghem 2015) and how number accuracy in the other categories (i.e. the pronoun, adjective, and determiner) evolves over time. The second and third sections of the practical part aim at determining whether the proficiency level that students initially have (i.e. before writing their first essay) has an impact on their linguistic development, and whether individuals with a low proficiency level at baseline display more linguistic variation than their high-level counterparts.