In 4 experiments it was found that contrast effects in person perception depend on the type and extremity of the primed information. Two previous models of priming effects, the standard-of-comparison and the set-reset models, make opposing predictions for the consequences of prime extremity on contrast effects. In Experiments 1 and 2 it was found that each model is descriptively accurate but in response to different priming stimuli. Exemplar primes (e.g., Dracula) produced greater contrast when extreme than when moderate, a pattern consistent with the standard-of-comparison model. Trait term primes (e.g., malevolent) produced greater contrast when moderate than when extreme, which is consistent with the set-reset model. In Experiments 3 and 4 it was demonstrated that the mechanisms through which contrast is produced are distinct for the 2 types of primes. Standard-of-comparison contrast is more perceptual and is not disrupted by cognitive load; set-reset contrast is effortful and requires sufficient cognitive capacity.
