Speeded classification in a probabilistic category structure: contrasting exemplar-retrieval, decision-boundary, and...

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Machine Learning

Background:

• Understanding how humans learn and classify categories is crucial for cognitive modeling.
• Existing models like exemplar, decision-boundary, and prototype models offer different explanations for category learning.
• Investigating the impact of feedback type on classification performance can differentiate between these models.

Purpose of the Study:

• To experimentally distinguish between exemplar-retrieval, decision-boundary, and prototype models of perceptual classification.
• To assess the effect of probabilistic versus deterministic category feedback on classification accuracy and response times.

Main Methods:

• Conducted speeded perceptual classification experiments.
• Manipulated feedback type (probabilistic vs. deterministic) for individual stimuli across conditions.
• Analyzed accuracy and response times as a function of feedback type.

Main Results:

• Subjects showed lower accuracy and longer response times when classifying stimuli with probabilistic feedback compared to deterministic feedback.
• These performance differences persisted despite an ideal observer using an identical linear decision boundary across conditions.
• Results align with exemplar model predictions.

Conclusions:

• The findings challenge the predictive power of prototype and decision-boundary models.
• Exemplar-retrieval models provide a better account of perceptual classification under varying feedback conditions.
• Feedback reliability significantly influences category learning and classification strategies.