In the literature, little research has hitherto been conducted to examine the implementation of integrated speaking test tasks. This study, in response, set out to compare the anxiety induced by a reading-to-speak task and the anxiety produced by a speaking-only task and to explore students' experiences of taking the reading-to-speak task. Forty-seven Taiwanese EFL learners completed a reading-to-speak task and a speaking-only task, filled out a state anxiety inventory twice, and responded to an open-ended questionnaire. The paired t-test revealed that the reading-to-speak task produced a statistically comparable degree of anxiety as the speaking-only task, possibly owing to the additional anxiety inflicted by its reading component. The qualitative analysis disclosed that students preferred the reading-to-speak task to the speaking-only task, even though it provoked additional anxiety, featured additional challenges, and entailed different response strategies. In light of these findings, implications are proposed for oral assessment research and EFL oral assessment practices. (Contains 2 tables.)