Abstract Recent research suggests that the cognitive processes of low self-esteem people (low SEs) perpetuate low self-esteem and its undesirable emotional and behavioral correlates. For instance, it has been suggested that low SEs are more “self-conscious” than high self-esteem people (high SEs) and that self-consciousness can impair task performance. It was reasoned that if low SEs were led to focus their attention away from themselves and onto the task, performance would improve relative to high SEs. In a 2 × 3 between-subjects factorial design, subjects high and low in chronic self-esteem performed a concept formation task under three conditions: (1) in the presence of an audience, where self-focused attention is presumably high; (2) in a control group, in which attention was not manipulated; and (3) with instructions to concentrate on the task diligently. A significant interaction effect was obtained, indicating that low SEs performed worse than high SEs in the audience condition ( p < .025), no differently in the control, and better than high SEs ( p < .01) when instructed to concentrate on the task. Although the latter results are unusual in the context of self-esteem research, they are strikingly parallel to recent findings in the study of test anxiety.