Visual prototype formation with discontinuous representation of dimensions of variability.

The prototype-distance model (Posner, 1969) predicts that when a series of similar visual stimuli are experienced, a prototype is abstracted at the point in the multidimensional similarity structure which represents the greatest similarity to all stimuli, whether the elements of the prototype have actually been experienced or not. The attribute-frequency model (Neumann, 1974) predicts the prototype as a pattern composed of the most frequently experienced elements on each dimension of variability. In three experiments, it was determined that: (1) Under some conditions, a prototype is formed of unexperienced values, and, under other conditions, the best recognized stimuli are those incorporating the most frequent values; (2) the present form of the prototype-distance model cannot account for best recognized stimuli being other than the central tendency; and, (3) the attribute-frequency model can, in principle, account for either finding by incorporating additional assumptions about the specificity with which values on dimensions of variability are encoded.