Typicality in logically defined categories: exemplar-similarity versus rule instantiation

Area of Science:

• Cognitive Psychology
• Computational Neuroscience

Background:

• Investigating how humans represent and categorize information is crucial for understanding cognition.
• Distinguishing between rule-based and exemplar-based models offers insight into category learning mechanisms.

Purpose of the Study:

• To contrast rule-instantiation and similarity-to-exemplars models in predicting category judgments.
• To examine the impact of exemplar frequency on typicality and classification of learned categories.

Main Methods:

• Two experiments were conducted using logically defined categories (unidimensional and disjunctive rules).
• Frequency of exemplar presentation during learning was manipulated.
• Subjects' typicality judgments and speeded classification reaction times were measured.

Main Results:

• Exemplar frequency significantly impacted goodness-of-example judgments and classification speed.
• Neither model solely predicted the observed results based on rule instantiation alone.
• A mixed exemplar model incorporating similarity components predicted goodness judgments effectively.

Conclusions:

• Category representation for logically defined concepts relies heavily on stored exemplars.
• Exemplar frequency plays a critical role in category learning and representation.
• Similarity-based processes are integral to how exemplars contribute to category knowledge.