Contrasting exemplar and prototype models in a natural-science category domain

Area of Science:

• Cognitive psychology
• Category learning
• Geology

Background:

• Distinguishing between exemplar and prototype models of category learning is a long-standing challenge.
• Previous research primarily used artificial categories, limiting real-world applicability.

Purpose of the Study:

• To compare exemplar and prototype models of category learning within a naturalistic domain: geologic rock types.
• To investigate how humans learn and generalize categories using real-world stimuli.

Main Methods:

• Utilized a high-dimensional feature space derived from multidimensional scaling and dimension ratings of geologic rock stimuli.
• Compared predictions from exemplar and prototype models against human performance in category learning tasks.
• Employed conditions favoring prototype abstraction, including large categories and delayed transfer tests.

Main Results:

• Results favored the exemplar model over the prototype model in accounting for category learning of rock types.
• Participants demonstrated generalization from learned exemplars to novel items, rather than relying on rote memory.

Conclusions:

• Exemplar models provide a superior framework for understanding category learning in naturalistic domains like geology.
• Human category learning involves generalizing from specific remembered instances, not just abstract prototypes.