24 
exemplars in a recall list, communalities should be sought. 
The goal should be to determine what characteristics of 
exemplars predict that an item will be listed as an 
exemplar. Perhaps the simplest way to do this is to have 
subjects recall negative exemplars of the target semantic 
category as well as positive exemplars. Then multivariate 
procedures can be used to determine which exemplar 
characteristics discriminate the positive exemplars from the 
negative. Als, on recall tasks, factors other than 
category criteria might influence the listing of exemplars. 
By having subjects list both positive and negative exemplars 
of the target category, the effects of such extraneous 
variables could be moderated in the analysis, since they 
should be relatively constant across both recall tasks. 
Method 
Pilot Studies 
Although semantic studies using free recall can be 
designed such that each subject generates criteria by which 
the exemplars can be differentiated (Coltheart & Evans, 
1981), there is such heterogeneity among subjects that the 
results cannot be pooled or easily compared. Since the 
present study sought to understand a consensus On the 
meaning of a social concept rather than individual 
differences, all subjects were to make judgements on the