Fluency in Second Language Testing: Insights From Different Disciplines

ABSTRACT

This article considers the construct of fluency in second language assessment and how it might be informed by research in applied linguistics. It briefly describes the way fluency is conceptualized in four language tests, as embodied in their respective assessment criteria, to show where the field is at present. The article then takes into account recent insights from applied linguistics, and from the fields of psycholinguistics, discourse analysis, and sociolinguistics. The article questions the current conceptualization of fluency in language testing, in which it is defined as a concept that should be sought in the ear of the beholder, and where disfluency is only seen as a deficit. Future research should focus on finding ways to ensure that the measures used in testing reflect the ability to talk fluently and efficiently, rather than measures that only reflect listeners’ impressions about such ability, and to show that these can be marked appropriately and reliably.

Acknowledgments

I thank all reviewers for the valuable comments on previous versions of this article. I especially thank the special issue editor Gad Lim for his great help in finalizing the article.

Notes

¹ This section on the way fluency is assessed in four current language tests is a slightly altered version of a section that appears in De Jong (2016a).

² This table also appears in De Jong (2016a). It presents almost the same variables as Kormos (2006, p. 163) had in her Table 8.2. I have added the measure pruned speech rate and subdivided her measure “number of disfluencies per minute” into “number of repetitions” and “number of repairs.” Finally, her measure “Space” has been left out here.

³ Speech rate is a mathematical combination of the speed measure articulation rate plus the breakdown measures number and duration of silent pauses. Mean length of utterance, if it is measured as the mean length of uninterrupted speech (in seconds) between silent pauses, is almost equivalent to the measure number of pauses per minute. If mean length of utterance is measured as the mean number of syllables between silent pauses, then it becomes a measure that combines articulation rate with several silent pauses per minute: the faster the speech, the more syllables in any stretch of speech and the fewer pauses, the longer these stretches of speech become. The following three formulae show these mathematical relations formally:(1) Speech rate = ((total duration – number of pauses*mean length of pauses) * articulation rate) / total duration(2) Mean length of utterance (in seconds) = (total duration – number of pauses*mean length of pauses) / (number of pauses + 1)(3) Mean length of utterance (in syllables) = MLUseconds * articulation rateIn other words, if total duration, number of pauses, mean length of pauses, and articulation rate are known, the measures speech rate and mean length of utterance (in seconds and in syllables) are also known, because they are mathematical combinations of these underlying measures. For the measure pruned speech rate, one can similarly formulate a mathematical formula using these same measures plus a measure of the number of disfluent syllables (e.g., the number of filled pauses, repetitions, and repairs).

⁴ In interviews in testing situations, however, interviewers usually allow for longer silences than is normal in natural conversation. This can be seen as test-talk (Fulcher, 1996), which may (negatively) impact on the perceived fluency of the interviewee. For the sake of fair judgments, it may be important to point out to raters that pauses in such interviews may be longer than those occurring in natural conversation.