Longitudinal L2 Development of the English Article in Individual Learners

We investigate the accuracy development of the English article by learners of English as a second language. The study focuses on individual learners, tracking their learning trajectories through their writings in the EF-Cambridge Open Language Database (EFCAMDAT), an open access learner corpus. We draw from 17,859 writings by 1,280 learners and ask whether article accuracy in individual learners fluctuates randomly or whether learners can be clustered according to their developmental trajectories. In particular, we apply kmeans clustering to automatically cluster in a bottom up fashion learners with similar learning curves. We follow learners for a period covering one CEFR level. Given the relatively short learning window, the majority of learners follow a horizontal line. Nevertheless, we also identify groups of learners showing a power-function and U-shaped curve. Crucially, these groups are ‘hidden’ when the aggregate of learners is considered, a finding highlighting the importance of individual level analysis.