Two experiments were conducted to test three models of person memory concer
ning multitrait expectancies: the trait-specific, complete association, and
dual-coding models. Participants in Experiment 1 were bd to expect a targe
t person to display two traits, then they were presented with congruent and
incongruent behaviors ostensibly performed by the target person, and final
ly they were asked to recall the behaviors. In contradiction. to the comple
te association model which predicts that participants form congruent-congru
ent associations as often as congruent-incongruent associations, incongruen
t items were better recalled than were congruent items. Furthermore, condit
ional probability analyses contradicted the implications of the trait-speci
fic and dual-coding models. Findings from Experiment 2 reinforced the lack
of support for all three models.