Distribution-dependent representations in auditory category learning and generalization.

Area of Science:

• Auditory Sciences
• Cognitive Neuroscience
• Perception

Background:

• Understanding auditory generalization is key to deciphering acquired category representations.
• The dual learning system (DLS) framework distinguishes explicit (rule-based) and implicit (information integration) learning systems.
• However, the nature of representations and transferability across contexts remains unclear.

Purpose of the Study:

• To investigate differences in information integration (II) and rule-based (RB) auditory category representations.
• To examine how acoustic and perceptual novelty/variability impact generalization success.
• To elucidate the distinct mechanisms underlying generalization in II and RB learning.

Main Methods:

• Three experiments were conducted to test generalization of II and RB auditory categories.
• Learners were trained on categories and then tested on novel sound sets and contexts.
• Representational similarity modeling was used to analyze generalization strategies.

Main Results:

• Both II and RB learners successfully generalized categories after training.
• II learners' generalization suffered with untrained perceptual areas and distributed samples.
• RB learners' generalization was robust to perceptual changes but sensitive to sound dispersity.

Conclusions:

• Distinct auditory category representations emerge from II and RB learning.
• Generalization mechanisms flexibly adapt to novel perceptual variability and context.
• Findings support dissociated representations and novel generalization strategies for perceptual constancy.