Integrative and metric properties of abstracted information as a function of category discriminability, instance variability, and experience

Investigated the abstraction of prototypical information as a function of instance variability, category discriminability, and category size. Following 3 study-sort trials involving schematic face stimuli, 72 undergraduates were asked to classify old, new, and prototypical stimuli both immediately and after 1 wk. In addition, Ss were asked to indicate their preference between a feature prototype (containing the modal feature values) and an average prototype (containing the average feature values) for each of these conditions. Results are consonant with previous studies showing the facilitative influence of category size on abstraction. Analyses revealed that Ss were more likely to count features than to average them, except when variability was low and discriminability was difficult. Both classification performance and reaction time were best predicted by the city-block distance of an exemplar to the feature prototype. Evidence suggests that more dimensions were utilized during abstraction when category discriminability was made more difficult. (PsycINFO Database Record (c) 2006 APA, all rights reserved).