Alternative training methods on self-efficacy and mastery of a computer software program were compared in the context of a field experiment involving 108 university managers. A behavioral modeling approach relative to a tutorial approach yielded higher self-efficacy scores and higher performance on an objective measure of computer software mastery. Participants scoring high in self-efficacy performed significantly better than participants with low computer self-efficacy scores. Participants low in self-efficacy reported greater confidence in their ability to master the software training in the modeling compared with the tutorial conditions. Participants in the modeling training reported more effective cognitive working styles, more ease with the task, more satisfaction with training, and less frustration compared with participants in tutorial training. Implications for training interventions are discussed.