We compared the effect of abrupt discontinuation of therapeutic doses of short half-life and long half-life benzodiazepines in 57 benzodiazepine-dependent patients (daily use, greater than 1 year). Despite the use of a mean daily dose of 14.1 mg of diazepam equivalents, there were notable residual symptoms of anxiety and depression present at intake (Hamilton Rating Scale for Anxiety score, 17.0; Hamilton Rating Scale for Depression score, 14.0). Benzodiazepine intake was stabilized for 3 weeks before double-blind assignment to placebo (n = 47), or continued benzodiazepine use (n = 10). Clinical assessments were performed daily, including benzodiazepine plasma levels. Depending on the outcome criteria used, anywhere from 58% to 100% of patients were judged to have experienced a withdrawal reaction, with a peak severity at 2 days for short half-life and 4 to 7 days for long half-life benzodiazepines. Relapse onto benzodiazepines occurred in 27% of patients who were receiving long half-life benzodiazepines and in 57% of patients who were receiving short half-life benzodiazepines. Baseline predictors of relapse were nonpanic diagnoses, a higher benzodiazepine dose, and a higher Eysenck neuroticism score. A short half-life and higher daily doses were associated with greater withdrawal severity, as were personality traits, such as dependency and neuroticism, less education and higher baseline levels of anxious and depressive symptoms. Patients who were able to remain free of benzodiazepines for at least 5 weeks obtained lower levels of anxiety than before benzodiazepine discontinuation. These results provide a detailed picture of the symptoms, time course, and multidimensional determinants of the benzodiazepine withdrawal syndrome.