According to the attentional control theory of anxiety (Eysenck, Derakshan, Santos, & Calvo, 2007), anxiety impairs performance on cognitive tasks that involve the shifting function of working memory. This hypothesis was tested using a mixed antisaccade paradigm, in which participants performed single-task and mixed-task versions of the paradigm. The single task involved the completion of separate blocks of anti- and prosaccade trials, whereas in the mixed task, participants completed anti- and prosaccade trials in a random order within blocks. Analysis of switch costs showed that high-anxious individuals did not exhibit the commonly reported paradoxical improvement in saccade latency, whereas low-anxious individuals did. The findings are discussed within the framework of attentional control theory.