Anxiety represents “a palpable but transitory emotional state or condition characterized by feelings of tension and apprehension and heightened autonomic nervous system activity” (Spielberger, 1972, p. 24). In the field of second/foreign language (L2) learning and testing, a remarkably extensive body of research has hitherto explored the contribution of this affective factor to learning and performance variations on the part of learners. Moreover, as delineated by MacIntyre and Gardner (1991), the bulk of this research could be considered as adopting one of three perspectives on the nature of anxiety: the trait, state, and situation-specific perspectives. However, to date, few research efforts have been invested into investigating how the anxieties within these different perspectives might interact with one another and how they might independently and jointly influence L2 learning or test performance. In response to the paucity of relevant research, this study aimed to explore the interactions between four different forms of anxiety, i.e., trait anxiety, state anxiety, language anxiety, and test anxiety, under these three perspectives, their respective and collective impact on the performance on the speaking section of a large-scale standardized English proficiency test in Taiwan, that is, the General English Proficiency Test (GEPT), and the potential moderating role of gender in such impact. By conducting this academic project, the researcher intended to achieve a manifold purpose: (1) to tease apart the effects of different anxiety reactions on L2 speaking test performance, (2) to bring to light the patterns of interaction between these anxiety reactions, and (3) to obtain further validity evidence for the interpretations of GEPT scores.

   Toward this end, the researcher recruited 251 English learners from Taiwanese universities to respond to two sets of GEPT Intermediate Level - Speaking Test (GEPTI-S) and four anxiety scales adopted or adapted from relevant research, namely, test anxiety inventory, state anxiety inventory, English classroom anxiety scale, and test anxiety scale. Capitalizing on path-analytical techniques, he then constructed and evaluated several path models to capture the interrelationships between the four anxieties and GEPTI-S performance and, as such, arrived at three primary findings (Figure 1). First, trait anxiety and language anxiety constituted the direct sources of state anxiety, whereas test anxiety mainly affected state anxiety indirectly. Second, trait anxiety and language anxiety had a statistically significant impact on the speaking test performance. Third, the impact of anxiety on performance did not vary with respect to gender.

   These findings hold implications for L2 assessment theories and instructional practices. Theoretically, they substantiated the theoretical frameworks of L2 use and speaking performance that incorporate anxiety as a crucial component (e.g., Bachman & Palmer, 2010). That is, by demonstrating that trait anxiety and language anxiety indeed played a part in test completion, they lent empirical support for the inclusion of anxiety and, more broadly, affective factors in the aforementioned frameworks to more adequately characterize the construct of L2 (speaking) performance. Pedagogically, these findings highlighted the importance of developing anxiety-coping strategies applicable to the L2 testing context. As this study has found that trait anxiety and language anxiety weaken performance, L2 practitioners might consider implementing lessons that offer training on anxiety-coping strategies that may help L2 learners grapple with those anxiety reactions occurring during the test-taking process, such as emotion-focused coping strategies (Kando & Yang, 2004), in attempts to both allow the test-takers to perform at their best and support a stronger validity argument for the test-score interpretations.