The anxiolytic activity of methylclonazepam was compared to lorazepam and placebo in a double-blind, randomized cross-over study, using a latin square design, in 18 inpatients meeting Research Diagnostic Criteria for Generalized Anxiety Disorders. Patients presented at least 1 year of symptomatology and had a minimum score of 20 on the Hamilton Anxiety Scale, despite chronic anxiolytic pharmacotherapy. Daily dosage was flexible, from three to six tablets of methylclonazepam 1 mg, lorazepam 2.5 mg, or placebo. Clinical evaluation included Hamilton Anxiety Scale, Clinical Global Impression (CGI), a side-effects checklist, completed every 2 days, and the global preference of the patient for one of the treatment periods. Results showed a highly significant superiority of both benzodiazepines over placebo on the Hamilton Scale (P less than 0.000001) and CGI (P less than 0.001), and also a significant superiority of methylclonazepam over lorazepam on the Hamilton Scale (P less than 0.01), CGI-1 (P less than 0.01), and in the number of patient preferences (14 versus 1; P less than 0.001), with no significant differences in side-effects or related to position in the trial. These results support the value of the cross-over design in chronic and severe anxious inpatients for the demonstration of differences in efficacy between anxiolytic pharmacotherapies.