Benzodiazepines (BZs) produce transient anterograde amnesia when given to normal subjects. The present longitudinal study assessed whether BZs impair memory functions in a clinically anxious group. Eighty-two agoraphobics with panic disorder were randomly allocated to one of four treatment groups resulting from a combination of two drug treatments (alprazolam or placebo) and two psychological treatments (exposure or relaxation). Of these, 38 subjects were assessed on a range of objective and subjective indices of memory and mood at three time points: before treatment, after 8 weeks of treatment and again at 24 weeks when patients had been free of medication from 5-8 weeks. Alprazolam produced pronounced impairments on a word recall task. At the 24-week medication-free follow-up, alprazolam patients were still impaired on the task compared with placebo patients. Alprazolam did not impair performance on an implicit memory task and did not affect digit span. Differences between psychological treatments emerged mainly in subjective memory effects. Findings are discussed in terms of the specificity of BZ-induced amnesia and differential tolerance to the varying effects of BZs. Implications are drawn out for the patient's ability to function optimally in daily life while taking alprazolam.