Effects of emotion regulation training on EFL learners’ speaking anxiety, academic engagement, and oral performance: An explanatory sequential mixed-methods design

Speaking anxiety is widely associated with reduced oral participation and performance, but the key unresolved issue is more specific: existing EFL research is dominated by correlational evidence or by interventions that change broad instructional conditions rather than directly teaching learners how to regulate emotion during speaking. It remains unclear whether an explicit, theory-derived classroom program can be associated with concurrent changes in speaking anxiety, academic engagement, and oral performance. The present study addressed this gap through an 8-week emotion regulation training program. Gross’s process model specified the trainable strategies (e.g., attentional deployment and cognitive reappraisal), while Control–Value Theory (CVT) provided the primary explanatory framework: regulation was expected to help learners experience greater control over valued speaking tasks, thereby making disruptive anxiety more manageable and supporting engagement and performance. An explanatory sequential mixed-methods design was used. The quantitative phase employed a quasi-experimental pretest-posttest non-equivalent control-group design with 80 Chinese undergraduate EFL learners from four intact General English classes. Two classes were assigned to the control condition ( n = 37) and two to the experimental condition ( n = 43). The experimental classes received emotion regulation training integrated into regular speaking instruction; the control classes received regular speaking instruction only. Student-level MANCOVA and follow-up ANCOVAs adjusted posttest outcomes for corresponding pretest scores. Because treatment assignment occurred at the class level and only four classes were available, these analyses are interpreted as baseline-adjusted student-level comparisons rather than definitive cluster-adjusted causal estimates. In the qualitative phase, 12 students representing strong, moderate, and limited improvement profiles were interviewed, and the data were analyzed using reflexive thematic analysis. The baseline-adjusted student-level analyses favored the experimental condition on all three outcomes. The multivariate effect was Wilks’ Λ = 0.68, F(3, 73) = 11.51, p < .001, partial η² = 0.32, 95% CI [0.14, 0.46]. Follow-up analyses showed lower speaking anxiety, F(1, 75) = 15.82, p < .001, partial η² = 0.17, 95% CI [0.05, 0.33]; higher academic engagement, F(1, 75) = 8.97, p = .004, partial η² = 0.11, 95% CI [0.01, 0.25]; and higher oral performance, F(1, 75) = 13.24, p = .001, partial η² = 0.15, 95% CI [0.03, 0.30]. Qualitative findings yielded six themes and, importantly, showed uneven response: some learners transferred strategies flexibly, whereas others continued to struggle during graded, spontaneous, or highly evaluative speaking. The findings are consistent with the possibility that explicit emotion regulation training can support emotional management, engagement, and oral performance in university EFL speaking classes. Gross’s process model and CVT jointly offer a coherent interpretation of the classroom activities and learners’ experiences. However, the four-class cluster structure prevents strong separation of the training effect from class- or instructor-level influences, and the qualitative strand explains participants’ experiences rather than establishing causal mechanisms.