Twenty male and 20 female subjects, previously classified as high or low in test anxiety, participated in an experiment on visual vigilance under conditions designed either to enhance or minimize feelings of being tested. Detection rate declined across the 36 min of the task among subjects high in test anxiety who believed that they were being tested and in subjects low in test anxiety who did not. False detection rates revealed no significant differences among conditions. Analysis of sensitivity to signals according to statistical decision theory revealed low sensitivity in both the high-anxiety test and low-anxiety/no-test conditions than in the other two. Analysis of decisional criteria showed that subjects in the high-anxiety test condition were more conservative in setting a criterion than subjects in the other three conditions. No sex differences were found. The results are discussed in terms of an elaboration of Sarason's (1978) model of test anxiety.