term basis to medical inpatients. Since there is no pharmacological evidence to support the current practice of using two different benzodiazepines together, and a potential reduction in cost to the Health Service of changing the practice, we undertook a randomised double-blind patient-preference study to test the hypothesis that one of the three commonly used benzodiazepines (flurazepam, nitrazepam, and diazepam) was better than the others as a short-term hypnotic in medical inpatients.This study design was chosen because of simplicity, economy in staff time, minimum patient interference, use of each patient as his own control, and the benefits of a sequential analysis design. Such trial methods have been shown to be reproducible and reliable in assessing symptomatic drug treatment qualitatively provided only that the basic symptoms under treatment do not vary on a daily basis.' Variability of symptoms in patientpreference studies are of considerable significance and may invalidate the use of this technique. Nevertheless, this did not constitute a significant problem in our study, and Zelvelder,8 in an extensive review of the techniques of evaluating hypnotics, recommended this approach as the most efficient one available for evaluating short-term hypnotic efficacy. Our studies confirmed that the benzodiazepines are effective hypnotics9 and failed to show a clinically significant difference in efficacy between them when used in this setting. Despite the relatively small numbers of patients, the incidence of undesired effects (usually excessive sedation) seemed to be higher with nitrazepam than with diazepam or flurazepam. Our data also show that older patients admitted to medical wards have a significant preference for a benzodiazepine as a hypnotic rather than an antihistamine.