Submission Summary

Direct link:

Interactions between linguistic complexity and fluency in a corpus of first and second language speech

Mon, March 26, 9:10 to 10:40am, Sheraton Grand Chicago, River Exhibit B

Session Submission Type: Poster


Interactions between linguistic complexity and fluency are explored in this study using a speech corpus with recordings in L1 Japanese and L2 English. Results show across both L1 and L2 that clause depth and lexical variety are closely related to the duration of silent pauses but not filled pauses.


Complexity, accuracy, and fluency are often taken as the fundamental areas in which second language proficiency develops and is measurable (Skehan 1998, inter alia). But far from being modular areas of communicative performance, these areas may interact with each other (as when a speaker reduces accuracy to increase fluency; cf., Skehan 2009). Furthermore, second language (L2) performance in any of these areas may be modulated by speakers' first language (L1) speech behavior (cf., Segalowitz 2010). The present study explores these possibilities by looking at the relationship between linguistic complexity and fluency in the L1 and L2 speech of native speakers of Japanese via analysis of a crosslinguistic corpus.

The corpus (Rose 2013) consists of speech recordings by native Japanese speakers doing three elicitation tasks (reading aloud, picture description, topic narrative) in both their L1 and L2 (English). The corpus is fully transcribed, including time intervals and shallow syntactic analysis (clause boundaries). Meta-data include independent estimates of L2 proficiency (based on standardized tests). Linguistic complexity was estimated using two measures: mean clause depth (i.e., embeddedness) and lexical variety (using type-token ratio). Fluency was estimated also using two measures: mean length of silent and filled pauses (e.g., uh/um).

Results (using linear regression modeling) show that speakers use longer silent and filled pauses in L2 than in L1, but that this does not differ across proficiency levels: Even high proficiency speakers still have longer pauses. More importantly, as clause depth increases or lexical variety decreases, the length of silent pauses decreases. But these complexity measures have no effect on filled pause duration.

Thus, linguistic complexity is in a trade-off relationship with limited aspects of fluency; here, silent pause duration only. This suggests that co-attention to silent pauses and lexical complexity may be warranted, while filled pauses may be unrelated.