Background <p>Learners’ self-evaluations of pronunciation are closely related to their emotional and motivational experiences in second language learning. However, relatively little is known about how L2 motivational selves, foreign language anxiety, and speaking self-efficacy are jointly associated with learners’ self-perceived pronunciation accuracy.</p> Methods <p>This cross-sectional study surveyed first-year English and English-related majors at two universities in western China. Classroom-based foreign language anxiety, ideal and ought-to L2 selves, speaking self-efficacy, and self-perceived pronunciation accuracy were measured using an online questionnaire. Confirmatory factor analysis and structural equation modelling were conducted in R. Alternative full and partial mediation models were compared, indirect effects were tested with bootstrapping, and the hypothesized moderation effect of speaking self-efficacy was examined using latent product-indicator models.</p> Results <p>The final structural model showed that the ideal L2 self was negatively associated with foreign language anxiety, whereas the ought-to L2 self was positively associated with anxiety. Foreign language anxiety was negatively associated with self-perceived pronunciation accuracy and statistically mediated the associations between both L2 selves and self-perceived pronunciation accuracy. Speaking self-efficacy was positively associated with self-perceived pronunciation accuracy. However, speaking self-efficacy did not significantly moderate the association between foreign language anxiety and self-perceived pronunciation accuracy. Additional model comparisons showed that direct paths from the L2 selves to self-perceived pronunciation accuracy were not significant.</p> Conclusions <p>The findings suggest that Chinese EFL learners’ self-perceived pronunciation accuracy is associated with both classroom-based anxiety and speaking self-efficacy, and that anxiety may represent an important statistical link between L2 self-guides and pronunciation-related self-evaluations. The results also indicate that general speaking self-efficacy may be more directly associated with self-perceived pronunciation accuracy than with weakening the FLA–SPP association. Future research should incorporate pronunciation-specific anxiety and efficacy measures, objective pronunciation assessments, and longitudinal designs.</p>