The contribution of learner corpora to the substantiation of fluency levels

The study reported in this paper examines whether, and to what extent, corpus data on three features contributing to (dis)fluency, namely unfilled pauses, restarts and false starts produced by 50 French-speaking learners of English, support the fluency descriptor scale of the Common European Framework of Reference for Languages (CEFR; Council of Europe 2001). The analysis shows that, except for restarts, there is no statistical difference in the mean frequency of pauses and false starts across the CEFR fluency levels represented in the data. Statistical measures did, however, show significant differences between the group of advanced learners and a comparable group of native speakers for all three categories of (dis)fluency phenomena. An investigation into the immediate environment of the three features further revealed that, rather than being used in isolation, they tend to form recurrent clusters that bear the potential of providing convincing evidence for the delineation of fluency levels.