The objective, sensitive, and reliable quantitation of withdrawal symptoms is essential to assess physical dependence on drugs. Data collected from 23 patients abusing high doses of benzodiazepines (mean diazepam dose equivalents, 150 mg/day; range, 40-500) and from 40 long-term therapeutic users randomized to receive either placebo (N = 19) or diazepam (N = 21; mean diazepam dose equivalents, 15 mg/day; range, 5-40; mean duration of use, 72 months; range, 6-240) were analyzed. Information on the type and severity of symptoms was obtained from several assessment instruments. In the high-dose abuse group, plasma benzodiazepine concentrations were measured daily and the 3 consecutive days of greatest relative daily fall were considered the critical withdrawal period. Selection of the 22 items of the Clinical Institute Withdrawal Assessment-Benzodiazepines (CIWA-B) was based on statistically significant differences between baseline and critical withdrawal periods in high-dose subjects and between symptoms associated with placebo and diazepam in low-dose subjects, using contingency tables and logistic regression analysis. Of the 104 symptoms measured by the assessment instruments, 22 symptoms were found to distinguish withdrawal from prewithdrawal.(ABSTRACT TRUNCATED AT 250 WORDS)