Abstract
Prototypes play a central role in theories of visual sciences, for instance, for the “recognition by prototypes” framework. Typically, prototypes are defined as results of principal components or averages of given exemplars. Both approaches seem promising as they are based on clear models that allow specific inferences, but data from 2 empirical studies contradict both assumptions. In study 1, we let 41 participants sketch the “prototypical face they have in mind”. By comparing the drawings with anthropometric face data, we found systematic deviations from the average face. Participants displaced cardinal features, most pronouncedly the eyes much higher, and drew the size of features very differently. In study 2, 107 participants were exposed to 2 average faces (a female and a male one) which were covered after 30 seconds; then they had to draw both faces from memory. Again systematic distortions from the average faces were found which were compatible with results from study 1. Both studies indicate that prototypical faces and even sketches depicting average faces do not seem to be based on principal components and averages but follow an abstracted configuration of a face.