This paper investigates various easily quantifiable performance features that might function as objective indicators of oral fluency. It would be advantageous if we could assemble a set of variables that functioned as good indicators of what expert judges, such as experienced native-speaker EFL teachers, are reacting to when subjectively assessing fluency. This would advance our knowledge of what constitutes fluency and especially what makes for perceived fluency differences among learners and how an individual learner improves in fluency over time.
To these ends a sample of the spoken performance of four advanced EFL learners was recorded at the start of six-months' residence in Britain and again shortly before departure. A panel of 10 native-speaker teachers of EFL subjectively rated the recordings for global fluency and generally agreed that the second set was more fluent than was the first, though for each subject one or two panel members dissented.
A battery of 12 readily quantifiable performance variables considered to be related to fluency was then assembled. Values per subject per recording were obtained, expressed as frequency rates or as proportions so that comparisons could be made between first and second renderings. For each variable, subjects' scores were compared between the two time points to ascertain in which features improvements were consistently manifested. For each variable t-tests were conducted between sample means at Week 2 and Week 23. Improvement of note at the 0.05 level of significance was found for three variables (one-tailed test), namely, speech rate, filled pauses per T-Unit, and percentage of T-Units followed by pause. Surprisingly, self-corrections did not prove a good indicator.
The implications of the study are that quantitative analysis can indeed help to identify fluency improvements in individual learners, and may have the potential to provide objective assessment of spoken fluency. Findings revealed two key areas of performance that seem to be important for fluency: (1) speech-pause relationships in performance and (2) frequency of occurrence of dysfluency markers such as filled pauses and repetitions (but not necessarily self-corrections).
However, even from this small-scale study it does seem that there is scope for individual variation among subjects in the precise areas in which fluency improvements may occur. Further research might be able to identify both “core” and “peripheral” fluency variables.
Quantitative analysis has applications both as a testing instrument and as a diagnostic tool to identify individual learner strengths and weaknesses among the components of fluency. Investigation of native-speaker performance might provide native-like target score ranges on each variable for learners to aim at.