Multi-Word Expressions in Second Language Writing: A Large-Scale Longitudinal Learner Corpus Study

In the present study, we sought to advance the field of learner corpus research by tracking the development of phrasal vocabulary in essays produced at two different points in time. To this aim, we employed a large pool of second language (L2) learners (N = 175) from three proficiency levels—beginner, elementary, and intermediate—and focused on an underrepresented L2 (Italian). Employing mixed-effects models, a flexible and powerful tool for corpus data analysis, we analyzed learner combinations in terms of five different measures: phrase frequency, mutual information, lexical gravity, delta Pforward, and delta Pbackward . Our findings suggest a complex picture, in which higher proficiency and greater exposure to the L2 do not result in more idiomatic and targetlike output, and may, in fact, result in greater reliance on low frequency combinations whose constituent words are non-associated or mutually attracted.