The present study investigated the relationship between individual differences in anxiety and eyewitness performance. Instructions designed to produce evaluative threat were given to anxious and nonanxious subjects at encoding and/or at retrieval. The performance of those subjects who were given anxiety-arousing instructions at encoding and retrieval and who scored high on the Test Anxiety Scale (TAS; Sarason, 1972) was less accurate on an eyewitness task than was that of the subjects who scored low on the scale. This difference in performance was attributable to improvement in the performance of nonanxious subjects, rather than debilitation in the performance of anxious subjects. Faced with the threat of failure, low-anxious subjects appeared to have the potential for increased effort. High-anxious subjects seemed to perform at or near capacity under all the experimental conditions because of their predisposition for task-irrelevant worry, which limited the possibility of an increase in working memory capacity.