Dependence to benzodiazepines is difficult to induce in animals but has been induced by high doses in man. Case reports of benzodiazepine dependence are rare compared with the usage of these drugs, but provide no proper epidemiological framework for the estimation of risk. Patients taking these drugs for four months or more may develop symptoms on withdrawal, characterized by anxiety, dysphoria, malaise, depersonalization, and by perceptual changes such as hyperacusis and unsteadiness. In our first study we compared four patients withdrawing from high doses of benzodiazepines with six patients withdrawing from therapeutic doses. In all patients the typical withdrawal syndrome was noted and was equal in intensity in both groups. In the second study, long-term, normal-dose benzodiazepine treatment was discontinued in 24 patients believed to be dependent on their medication. The withdrawal was gradual, placebo-controlled and double-blind. All experienced some form of withdrawal reaction, which ranged from anxiety and dysphoria to moderate affective and perceptual changes. Symptom ratings rose as the drugs were discontinued, but usually subsided to pre-withdrawal levels over the next two to four weeks. Electroencephalograhic (EEG) changes comprised marked reduction in fast-wave activity as the drugs were withdrawn, and an improvement in psychological performance was noted. It is concluded that a risk of dependencies present in all patients taking benzodiazepines even in therapeutic doses for more than a few months. Caution is urged in the prescribing of these drugs.