Cross-talker generalization in the perception of nonnative speech: A large-scale replication

Area of Science:

• Cognitive Science
• Auditory Perception
• Speech Processing

Background:

• Speech perception relies on generalizing across different talkers.
• Previous research suggested exposure variability is crucial for cross-talker generalization.
• Bradlow & Bent (2008) found generalization only after multi-talker exposure.

Purpose of the Study:

• To replicate and investigate the role of exposure variability in cross-talker generalization in speech perception.
• To determine if single-talker exposure can elicit cross-talker generalization.
• To explore the mechanisms underlying cross-talker generalization.

Main Methods:

• Two large-scale speech perception experiments (N=640) with 20 exposure-test talker combinations.
• Bayesian data analysis, including meta-analyses and replication tests.
• Open science practices: data, stimuli, and code shared via OSF.

Main Results:

• Cross-talker generalization was observed after both multi-talker and single-talker exposure.
• Generalization varied depending on the specific talker combinations.
• Exposure variability is not necessary, but may indirectly influence generalization.

Conclusions:

• Single-talker exposure can lead to effective cross-talker generalization in speech perception.
• Similarity-based inferences provide a unifying explanation for generalization findings.
• Rethinking the necessity of exposure variability for learning and generalization in speech perception.