Noun-adjective collocations in L2 Portuguese: Exploring the effects of proficiency and native language in a large-scale learner corpus

Abstract We examined learners’ use and development of multiword expressions in Portuguese, an understudied L2. We analyzed essays from a large pool of learners across five proficiency levels (i.e., beginner, elementary, intermediate, upper intermediate, advanced), representing a variety of typologically diverse L1s (N = 679). Specifically, we investigated learners’ use of multiword expressions in terms of five different measures of collocability: phrase frequency, mutual information, log dice and two measures of Delta P, as functions of two key learner-related variables: proficiency levels, and the word-order typology of learners’ L1s. We also examined the accuracy in learners’ use of multiword expressions. The results of our mixed-effects models revealed a complex picture. While highly proficient learners use multiword expressions more accurately, they do not always produce more frequent and idiomatic outputs. Typological similarity between learners’ L1s and Portuguese do not consistently result in more idiomatic multiword expression use. Keywords: Multiword expressions; learner corpora; collocations; association measures; Portuguese, usage-based approaches