How speech-to-text technology affects pronunciation gains and self-confidence in EFL learners

Abstract

Pronunciation training is a crucial aspect of second language learning. However, instructional approaches often struggle to balance accuracy-focused feedback with maintaining learner confidence. This study investigates the effects of speech-to-text technology on both pronunciation gains and self-confidence among Japanese EFL learners. A 15-week quasi-experimental study was conducted with 77 university students, comparing two groups: (1) a speech-to-text group that used automated pronunciation feedback within a flipped learning model; and (2) a flipped-only group that received the same explicit instruction, but without speech-to-text technology. Results indicated that speech-to-text training significantly improved phoneme-level pronunciation but also increased self-consciousness, lowering students’ speaking confidence. In contrast, the flipped-only group reported greater confidence gains despite demonstrating less measurable pronunciation improvement. Additionally, while overall speaking proficiency showed a near-significant trend of improvement in the speech-to-text group, listen-and-repeat pronunciation practice alone was insufficient to drive strong statistical significance. These findings highlight a confidence-accuracy tradeoff in pronunciation instruction, and that educators must carefully consider how to integrate speech-to-text technology while fostering a supportive learning environment that sustains student motivation and confidence.