Three predictions of effects of cognitive dissonance on task evaluation and task performance were examined in two experiments, which were derived respectively from Weick's “dissonance reduction process” model, Hull-Spence's drive theory, and Takada-Hashimoto's “one way alternative” hypothesis. The design of Exp. I was dissonance (2)×task interest (2)×time for assessing task evaluation (3). Four hundred and eighty-one 7th-graders served as subjects. The results seemed to support the prediction derived from the drive theory. But there remained some possibility for Weick's model to be valid. In Exp. II, with similar design and subjects, the result tended to be supported by Weick's model better than by the drive theory. These findings suggested a new model, which is not merely a modification of Weick's model, but also implies conception from the drive theory.