The Development of Collocations as Constructions in L2 Writing

Cross-sectional and longitudinal learner corpus studies utilizing phraseological, frequency, and association strength approaches to phraseological unit identification have shown how the use of phraseological units varies across proficiency levels and develops over time. However, these methods suffer from several limitations, such as a reliance on native speaker intuition, a limited focus on contiguous word sequences, and a neglect of part of speech information in association strength calculation. This study seeks to address these limitations by defining lexical collocations as constructions (henceforth “collconstructions”) within the framework of Construction Grammar and investigating their cross-sectional variation and longitudinal development in two corpora of L2 writing. The cross-sectional corpus consisted of beginner and intermediate EFL learner texts assessed for overall writing proficiency, while the longitudinal corpus contained freewrites produced by ESL learners over the course of one year. Contiguous and non-contiguous adjective-noun, verb-noun, and adverb-adjective collconstruction tokens were extracted from each learner text in the two learner corpora. Each learner text was assessed for multiple constructional and collostructional indices of collconstruction production. Constructional indices included type frequencies, token frequencies, and normalized entropy scores for each collconstruction category. Collostructional indices consisted of proportion scores for different categories of adjective-noun, adverb-adjective, and verb-noun collconstruction types and tokens based on covarying collexeme scores calculated using frequency information from an academic reference corpus. Variation across proficiency levels was evaluated both qualitatively and quantitatively. The qualitative analysis consisted of examining variation in the production of specific functional collconstruction subcategories from a Usage-based Second Language Acquisition perspective. The quantitative analysis consisted of the calculation of an ordinal logistic regression in order to determine whether any indices of collconstruction production were predictive of L2 writing quality. Longitudinal development at the group level was investigated through the use of linear mixed effects models. Development for individual learners was examined from a Dynamic Systems Theory perspective that focuses on the role of variability in language development as well as interconnected development for multiple indices of collconstruction production. This study has important implications for future research on L2 phraseology research and second language acquisition research as well as phraseology pedagogy