Proceedings of the Annual Conference of the Cognitive Science Society vol:30 pages:757-762
Annual Conference of the Cognitive Science Society edition:30 location:Washington, DC date:23-26 july 2008
In this study we re-evaluated earlier findings that natural
language categories are represented by an exemplar
representation rather than by a prototype representation.
Using a restricted prototype model, an exemplar model and a
“flying” prototype model, we predicted typicality ratings for 11 natural language categories from 2 semantic domains,
animals and artifacts. We showed that exemplar models
outperform prototype models when the prototype is restricted
to the average exemplar, but the opposite pattern emerged
when the prototype was free to move.