This study examines the role of affective assessment, a self-generated influence in learning acquisition, in Bandura's original modeling-of-aggression procedure. Thirty-two first- and second-grade children (both sexes) rated toys and televised acts of aggression against an inflated doll according to their personal affective preference. A matched control group of 32 children was yoked to the affective preferences of the experimental subjects. Control subjects were not shown the televised aggression. All children were then introduced to a free-play situation in which toys and an inflated doll were present. In line with Bandura's findings, we predicted that observational learning would be demonstrated across experimental and control conditions. However, in line with the tenets and previous research findings in support of logical learning theory, we predicted that children would model affectively preferred aggressive acts and toys more readily than affectively dispreferred acts and toys. The experimental predictions were supported in the data collection. Boys and girls differed significantly only in their imitation of positively rated aggressive actions and not in their imitation of negatively rated aggressive actions. The broader theoretical implications of these results, as well as the ramifications of these findings for the televised violence and aggression issue, are discussed.