Disentangling the contributions of shorter vs. longer lexical bundles to L2 oral fluency

Abstract

Shorter lexical bundles (LBs) have been the central point of focus in L2 oral fluency studies, with longer LBs often being neglected. The current study examined the extent to which longer LBs vs. shorter LBs relate to aspects of oral fluency. Data were collected from 50 undergraduate L2 English learners performing three speaking tasks. We analyzed speaking performances in terms of speed, breakdown, and repair fluency, while LB (2- to 5-word) usage was measured using a combined text-internal and text-external approach. Utilizing robust multiple regression, dominance analysis, and random forest techniques, our study found a marginal positive effect of longer LB use on speed fluency, a potential negative association with the frequency of mid- and end-clause pauses, and a strong negative association with the frequency of total repair. Furthermore, the analysis uncovers the significant impacts of shorter LBs (bigrams and trigrams) on various aspects of fluency. These insights underscore the pedagogical potential of longer LBs for enhancing oral fluency, while emphasizing the necessity for a comprehensive focus on different lengths and types of multiword sequences in EFL pedagogy. Our findings could inform more effective, data-driven language teaching strategies and materials. We discuss the findings in relation to L2 speech production models and provide important suggestions for future LB-fluency research.