According to several studies, the amount of subarachnoid blood on the initial computed tomogram of patients with aneurysmal subarachnoid hemorrhage has predictive value with respect to infarction and outcome. Of several methods for assessing the amount of subarachnoid blood, none has been subjected to a study of interobserver agreement. We describe our own method, applied in previous studies, in which the amounts of blood in 10 basal cisterns and fissures and in four ventricles are graded separately. In grading single computed tomograms of 182 consecutive patients with subarachnoid hemorrhage, the agreement between pairs of three observers, studied with kappa statistics, was relatively good for individual cisterns or fissures (kappa between 0.35 and 0.65) and ventricles (kappa between 0.47 and 0.74). The Spearman rank correlation coefficients for the sum of the scores for subarachnoid and intraventricular blood were very high. Summed scores for extravasated blood are suitable as a baseline variable in follow-up studies of patients with subarachnoid hemorrhage.