OBJECTIVE--To assess the sedative effects in overdose of temazepam and oxazepam compared with other benzodiazepines to determine if this explains reported differences in fatal toxicity. DESIGN--Cohort study of patients admitted with benzodiazepine poisoning. SETTING--Newcastle, Australia. SUBJECTS--303 patients who had ingested benzodiazepine alone or in combination with alcohol and presented to a general hospital which served a well defined geographical area. MAIN OUTCOME MEASURES--Degree of sedation: Glasgow coma score, McCarron Score, and whether patients were stuporose or comatose. RESULTS--Oxazepam produced less and temazepam more sedation than other benzodiazepines. Unadjusted odds ratios for coma with oxazepam and temazepam compared with other benzodiazepines were 0.0 (95% confidence interval 0.0 to 0.85) and 1.86 (0.68 to 4.77) respectively, chi 2 = 7.08, 2df, P = 0.03. After adjustment for potentially confounding effects of age, dose ingested, and coingestion of alcohol, the odds ratios were 0.22 (0.0 to 1.43) for oxazepam and 1.94 (0.57 to 6.23) for temazepam. Similar results were obtained for other measures of sedation. CONCLUSIONS--These results were in accordance with fatal toxicity indices derived from coroners' data on mortality and rates of prescription. The relative safety of benzodiazepines in overdose should be a consideration when they are prescribed.