Abstract
Speaking a language fluently is commonly regarded as the ultimate goal of mastering a new language. As a result, test-takers’ fluency is frequently evaluated as part of speaking proficiency tests. Recently, Iwashita, Brown, McNamara, & O’Hagan (2008) reported the relationship between detailed features of speaking performance and holistic scores of around 200 subjects’ performance on the TOEFL iBT prototype speaking tasks. The results showed that vocabulary and fluency had the strongest impact on distinguishing overall speaking proficiency levels. Among the fluency measures, unfilled pauses, total pause time and speech rate showed a significant relationship with proficiency levels. Inspired by Iwashita et al. (2008), this study aims to investigate EFL test-takers’ fluency on simulated TOEFL iBT speaking tasks. To see whether Iwashita’s results would also apply to EFL learners who mainly receive formal on-campus instructions, we used English learners in Taiwan and propose the following questions: 1. How do TOEFL iBT speaking test-takers who have different levels of Delivery, Language Use and Topic Development differ in terms of temporal measures of fluency? 2. What are the relationships among the temporal measures of fluency used in this study? The participants were 100 Taiwanese EFL learners who either had already taken the TOEFL iBT or had planned to do so by the year 2010. To make sure that these participants were representative of TOEFL iBT test-takers from Taiwan, they had to spend most of their lives in Taiwan and did not spend more than six months abroad. The participants took one simulated TOEFL iBT speaking test in a one-on-one audio-recorded setting. Their performances were rated by two EFL teachers based on the TOEFL iBT rating scales and were grouped into three groups in each task following three speaking constructs – Delivery, Language Use, and Topic Development. The analyses of the speaking data followed Iwashita et al (2008), including the number of filled pauses, unfilled pauses and repair, total pausing time, speech rate, and mean length of run (MLR). One more variable was added – phonation-time ratio (PTR), which was found to be a useful indicator in identifying group differences in Kormos & Denes (2004) and Towell, Hawkins, & Bazergui (1996). The results indicated that the number of unfilled pauses, total pausing time, speech rate, MLR and PTR successfully differentiated test-takers from three groups based on Delivery and Language Use scores throughout all six task types, and in Task 1, 2, 3, 5 based on Topic Development score. Based on Topic Development score, speech rate was the only temporal measure which showed significant results in Task 4, and speech rate as well as MLR showed significant results in Task 6. On the other hand, the number of filled pauses and repair showed limited results in identifying group differences. Also, the five statistically significant temporal measures highly correlated with one another. The number of unfilled pauses, total pausing time and PTR have strong relationships with each other, with the total pausing time being the opposite of the other two. Speech rate and MLR have strong relationships with each other, while speech rate, unfilled pauses and total pausing time have moderately strong negative relationship with each other. Finally, filled pauses and repair seemed to have no obvious relationship with the rest of the temporal measures. Previous research on fluency has tended to focus on ESL context or use narrative tasks in non-testing contexts. Not much has been done with monologic tasks in a testing context. Thus, results from this study provided further information on EFL learners’ fluency in a testing context. Additionally, the results from this study could assist test-takers to pay more attention to the significant features attributing to language flow and do better in their actual TOEFL iBT performance.