What we do (not) know about the mechanisms underlying adaptive speech perception: A computational framework and...

Area of Science:

• Cognitive Science
• Auditory Neuroscience
• Computational Linguistics

Background:

• Speech perception adaptivity is crucial for understanding unfamiliar talkers.
• Mechanisms driving this adaptation are not fully understood.
• Previous research proposed three distinct hypotheses: signal normalization, linguistic representation changes, or decision-making shifts.

Purpose of the Study:

• To develop and implement a computational framework for adaptive speech perception (ASP).
• To integrate and compare three leading hypotheses of speech adaptation within a unified model.
• To derive testable predictions from acoustic stimulus properties.

Main Methods:

• Developed a general computational framework implementing three distinct adaptive speech perception mechanisms.
• Used the framework to generate predictions for experimental paradigms.
• Analyzed existing experimental data using the framework's predictions.

Main Results:

• The developed ASP framework successfully integrated signal normalization, linguistic representation changes, and decision-making shifts.
• Current experimental paradigms, at their typical analysis level, fail to distinguish between these three proposed mechanisms.
• The study highlights limitations in current experimental designs for elucidating speech adaptation processes.

Conclusions:

• Existing experimental data and analysis methods are insufficient to differentiate leading hypotheses of speech adaptation.
• A shift in research practices, including experimental design and data analysis, is necessary.
• The study provides a framework and recommendations for future research into the mechanisms of adaptive speech perception.