This study explored the relationships among language proficiency, two selected test-taker characteristics (i.e., topical knowledge and anxiety), and integrated speaking test performance. Data collection capitalized on three sets of instruments: three integrated tasks derived from TOEFL-iBT preparation materials, the state anxiety inventory created by Spielberger, and three topical knowledge tests constructed and validated by the researchers on a group of 421 English as a Foreign Language (EFL) learners. For the main study, another 352 EFL learners answered the topical knowledge tests, took two integrated tasks (i.e., one practice task and one focal task), and responded to the state anxiety inventory. Furthermore, half took the focal task on one topic while the other half dealt with another topic. Path analyses revealed that in integrated speaking assessment, language proficiency did not represent a primary source of anxiety, whereas the impact of topical knowledge on anxiety varied, depending on the topics used. In addition, language proficiency and the two selected test-taker characteristics all exerted a significant influence on integrated speaking test performance. These findings disclosed the antecedents of anxiety in integrated speaking assessment, supported the inclusion of test-taker characteristics on second/foreign language (L2) theoretical models, and highlighted the validity threat posed by test-taker characteristics to test-score interpretations based on integrated speaking test performance.