Tracing phraseological complexity development in a longitudinal learner corpus with verb + object co-occurrences

Paquot, Magali [UCL] Naets, Hubert [UCL] Gries, Stefan [UCL]

This presentation aims to report on an investigation of the development of verb + object co-occurrences in French EFL learner writing from the Longitudinal Database of Learner English (LONGDALE; Meunier, 2013). The study makes use of mean mutual information (MI) scores to examine the phraseological complexity of foreign language texts: this association measure has been shown to promote the relatively less frequent and more semantically complex word pairs in learner production, thus tapping into the sophistication of word combinations (cf. Paquot, 2017). The focus is placed on verb + object structures as they have typically not been investigated in learner corpus studies that adopted a positional model of statistical co-occurrence (e.g. Granger & Bestgen, 2014) but are considered a major stumbling block for EFL learners. Building on the current state of the art, the study addresses the following research questions: - RQ1: To what extent can syntactic co-occurrences, and verb + direct object structures more particularly, be used to trace phraseological development in a longitudinal corpus of EFL learner writing? - RQ2: What are the effects of proficiency vs. time spent learning English on collocation strength in learner writing development? To answer the RQs, the study replicates the methods used in Paquot (2017) to extract and analyze syntactic co-occurrences from reference and learner corpora on the LONGDALE corpus. Results suggest that syntactic co-occurrences, and verb + object relations more particularly, will only be useful to trace phraseological complexity development in longitudinal learner data if and only if topic/prompt is controlled or used as a predictor. Time spent learning English does not have an effect on collocation strength per se; what matters more is foreign language proficiency and whether learners improve from one year to the next.