Generalized perceptual adaptation to second-language speech: Variability, similarity, and intelligibilitya)

Area of Science:

• Psycholinguistics
• Auditory Perception
• Second Language Acquisition

Background:

• Perceptual learning for speech is crucial for understanding diverse accents.
• High-variability training typically aids generalization, but low-variability exposure can also be effective.
• A similarity-based account suggests talker similarity, not variability, drives generalization.

Purpose of the Study:

• To test the predictions of a similarity-based account of speech perceptual learning.
• To investigate the role of training-test talker similarity versus exposure variability in cross-talker generalization.
• To compare perceptual adaptation to second-language (L2) English speech following single- versus multiple-talker training.

Main Methods:

• A round-robin design was used with four L2 English talkers from different first-language (L1) backgrounds.
• Participants underwent single- or multiple-talker training with 60 L2 English sentences.
• Cross-talker/cross-accent generalization was assessed after training sessions.

Main Results:

• Cross-talker/cross-accent generalization occurred with both single- and multiple-talker training, but was not guaranteed.
• Generalization was not consistently better for identical training-test talker pairings compared to mismatched pairings.
• Acoustic analyses showed a dissociation between phonetic similarity and generalization; intelligibility during training correlated with adaptation and generalization.

Conclusions:

• Generalization in L2 speech perceptual learning is influenced by training-test talker similarity and training phase intelligibility.
• Exposure to talker variability may not be the sole determinant of cross-talker generalization.
• Effective L2 speech perceptual learning may require a combination of talker variability, training-test similarity, and high training intelligibility.