From meaning to sound: how word learning shapes non-native speech perception.

Introduction: Adult learners often struggle to perceive and acquire unfamiliar speech sounds in a second language, especially at the initial stages of learning. Traditional perceptual training methods, such as discrimination tasks, tend to be less effective with beginners, as they rely on low-level acoustic judgments and lack meaningful context. This study investigates whether training with cross-situational word learning (CSWL), a meaning-based learning paradigm, can improve the perceptual discrimination of non-native vowel contrasts. Methods: Thirty-seven native speakers of Hungarian were trained on eight European Portuguese pseudowords through a single CSWL session involving alternating passive and active learning blocks, feedback, and exposure to multiple native voices. Participants completed identification and discrimination tasks before and after training. Non-native word learning and vowel discrimination were measured before and after training, by means of identification and discrimination tasks, respectively. Results: Learners achieved above-chance word identification, indicating successful lexical learning. However, improvement in vowel discrimination was contrast-specific: participants improved in three of six contrasts, while performance remained low for the most difficult contrast. Learners also showed lower identification accuracy for pseudowords containing this contrast, and individual discrimination ability was associated with word learning success. Discussion: These findings highlight that while meaning-based training through CSWL can support early lexical and phonological learning, perceptual challenges remain for difficult contrasts. The study advances our understanding of how word learning and sound perception interact during second language acquisition and demonstrates the potential of lexically grounded approaches for perceptual training at the onset of learning.