How project-based learning improves EFL speaking in Chinese higher vocational colleges: a serial mediation study

English-speaking competence remains a persistent challenge for Chinese EFL learners, and project-based learning has been advocated as a remedy, though existing evidence yields mixed effects and rarely specifies the psychological mechanisms or sub-skill profiles driving the variation. This explanatory sequential mixed-methods quasi-experimental study examined a 12-week project-based learning intervention with a 6-week delayed posttest in Chinese higher vocational college students (N = 179; PBL n = 93, CTL n = 86), testing autonomous motivation and collaborative engagement as serial mediators of English-speaking competence. Speaking was rated on a four-dimensional rubric covering fluency, accuracy, complexity, and pragmatic-interactional competence, with inter-rater intraclass coefficients of 0.76 to 0.87; serial mediation was estimated using PROCESS Model 6, and inductive thematic analysis of post-intervention interviews probed the operating mechanisms within project groups. Composite speaking showed a large condition advantage at the immediate posttest (d = 1.07), concentrated on fluency (d = 0.82) and pragmatic-interactional competence (d = 0.68), modest on complexity (d = 0.48), and not statistically detectable on accuracy. Retention 6 weeks later was differential, with motivation and engagement retaining 88 and 78% of their immediate gains while composite speaking competence retained 93%. The serial indirect path through engagement was significant whereas the motivation-only path was not, indicating that engagement carries the effect and motivation reaches speaking only through engagement. Qualitative themes located the active mechanism in group accountability while revealing compressing dynamics including free-riding, conversational dominance, and L1 slippage under time pressure. These findings imply that the active ingredient of project-based learning is the psychological condition it creates for learners rather than the task choreography, and that instructional design should center on supporting learners' needs for autonomy, competence and relatedness, and specifically on the group accountability structures through which motivation becomes speaking.