Beyond exemplars and prototypes as memory representations of natural concepts: A clustering approach

Categorization in well-known natural concepts is studied using a special version of the Varying Abstraction Framework (Vanpaemel, W., & Storms, G. (2006). A varying abstraction framework for categorization. Manuscript submitted for publication; Vanpaemel, W., Storms, G., & Ons, B. (2005). A varying abstraction model for categorization. In B. Bara, L. Barsalou, & M. Bucciarelli (Eds.), Proceedings of the 27th annual conference of the Cognitive Science Society (pp. 2277–2282). Mahwah, NJ: Lawrence Erlbaum). This framework assumes a continuum between highly abstract memory representations (prototypes) and highly detailed representations of concept members (exemplars). Comparison stimuli for categorization are obtained by taking for each category the centroids of a set of clusters, produced by K-means clustering, effectively producing the Generalized Context Model (GCM; Nosofsky, R. M. (1986) Attention, similarity, and the identification-categorization relationship. Journal of Experimental Psychology: General. 115, 39–57) and the Single-Prototype Model as extreme cases. The clustering version of the Varying Abstraction Framework was fit on a set of novel, to-be-classified fruits and vegetables ( Smits, Storms, Rosseel, & De Boeck, 2002) and on a new set of novel, to-be-classified carnivores and herbivores. Better fit values were clearly obtained for a model based on intermediately abstract representations, indicating a strategy where people compare the novel stimuli to a set of multiple prototypes. This sheds a new light on the prototype versus exemplar discussion that has dominated the literature over the past 25 years.