Third grade subjects were classified as impulsive or reflective on the basis of the Matching Familiar Figures (MFF) test. In Experiment I, subjects were given either convergent thinking tasks, brainstorming tasks, or a control task. In Experiment II, half of the subjects received difficult tasks intended to induce failure while the other half received success tasks, with feedback or no feedback appropriate to each subject's task, provided. Response latencies on the MFF post tests were increased by convergent training to a lesser extent by the failure tasks. The data suggest that the convergent thought process may be potentially useful in modifying impulsive thinking. (Author)