It has generally been assumed that the therapeutic action of benzodiazepines results from the effect of these drugs on mood. We suggest, however, that in reducing anxiety, benzodiazepines may have a direct effect on anxiety-related cognitions. The investigation was designed to examine the question of whether anxiety-related cognitive bias is reduced by diazepam in subjects selected according to DSM-III(R) criteria for generalized anxiety disorder (American Psychiatric Association, 1987). A modification of the Stroop color-naming task was used to measure bias toward the processing of threatening material. The results demonstrate that the reduction in anxiety shown by anxious patients after diazepam is not accompanied by a reduction in cognitive bias toward the processing of threatening material. This suggests that diazepam fails to reduce anxiety-related cognitive bias in clinically anxious subjects. It would seem, therefore, that diazepam alleviates anxious mood rather than cognitive manifestations of anxiety.