Eighty-six elementary and secondary undergraduate teacher education students exhibiting low and moderate levels of self-efficacy belief were randomly assigned to observe two types of instruction and two types of attributional feedback concerning teaching a child how to find the main idea of a paragraph. The research focused on the effects of two types of material presentation--direct instruction versus cognitive modeling, and the effects of watching a demonstration of that skill with two types of guiding comments--task-oriented versus self-efficacy statements. Analyses of variance were performed on each dependent measure. Results suggest that, for students exhibiting initial low self-efficacy beliefs, cognitive modeling is more effective for raising estimates of success and self-efficacy beliefs than is direct instruction. (Author/JD)