A Latent Curve Model Approach to Studying L2 N-Gram Development

Current quantitative methods in second language (L2) acquisition have proven useful in examining how phraseological unit production changes over time. However, these methods are limited in that they do not allow for the analysis of individual differences in those changes. This study demonstrates the potential for Latent Curve Modeling, a type of Structural Equation Modeling, to address questions about productive phraseological knowledge development. It examines growth in multiple indices of bigram and trigram use (frequency, association strength, proportion) in the spoken output of L2 speakers over the course of a 4-month study. Results for unconditional latent curve models indicate that spoken bigram and trigram proportions increased for the entire group over the study period. Conditional latent curve models showed that growth in bigram frequency and bigram proportion was predicted by proficiency, with less proficient writers experiencing greater growth. These models also demonstrated that conversation dyad predicted growth in spoken bigram frequency in that L2 speakers with L2 conversation partners, as compared to first language (L1) partners, produced more high-frequency bigrams over time. These results have implications for research on L2 productive phraseological knowledge development specifically and longitudinal L2 research in general.