Methods are described for a five-center study of flurazepam and midazolam, in which 107 patients with histories of benzodiazepine use for chronic insomnia were enrolled. Data were available for 99 of these patients. Staff and patient manuals and a behavior-based computer system were specially designed to measure and delineate clearly the study procedures and parameters. Patients were carefully followed and supported during a 20-day washout period and underwent extensive training on psychomotor performance tasks. They received placebo for 2 nights and were then randomly assigned to one of four study treatments--flurazepam 15 or 30 mg, midazolam 15 mg, or placebo--for 14 consecutive nights. All-night sleep electroencephalographic recordings were obtained on study nights--1 and 0 (placebo) and 1, 2, 7, 13, and 14 (treatment nights). Results of four computer-generated psychomotor performance tasks and three cognitive tasks, plus subjective evaluations of sleep, performance, and mood, were recorded in the morning after each night spent in the sleep laboratory. Blood and urine samples were analyzed for drug concentrations in plasma and for compliance with the protocol. A pilot study, using a high (nonclinical) dose of flurazepam (45 mg), preceded the multi-center study and was designed to evaluate tests in healthy volunteers and to familiarize staff with equipment and tests.