TOEFL iBT integrated speaking tests: a comparison of test-takers' performance in terms of complexity, accuracy, and fluency

This study compares three integrated tasks of the TOEFL iBT speaking subtest in terms of complexity, accuracy, and fluency. To this end, a group of TOEFL iBT Iranian candidates took a simulated TOEFL iBT some days prior to their real exam. The collected oral responses were first transcribed and then quantified using software such as ‘Syllable Counter’ and ‘Coh-Metrix3’ for fluency and complexity, respectively. For accuracy, however, the responses were tallied manually. The results revealed the responses to the three speaking tasks were significantly different in terms of fluency. The difference in the accuracy index also turned significant, though the pairwise comparisons showed some inconsistencies. As for the selected complexity measures, lexical diversity, the mean number of modifiers per NP, and latent semantic analysis all showed significant differences between tasks 2 and 3 on the one hand and task 4 on the other. Left-embeddedness, however, revealed no significant difference among the three tasks. The results may support the influential role of prompting texts in such integrated speaking tasks