To test interphysician agreement on the diagnosis of subtype of ischemic stroke, we sent subtype definitions and 18 case summaries (clinical features and pertinent laboratory data) to 24 neurologists who have a special interest in stroke, and asked them to determine the most likely subtype diagnosis. The overall agreement was 0.64 (Kappa [K] = 0.54). Interphysician agreement was highest for the diagnoses of stroke secondary to cardioembolism (K = 0.75) or to large-artery atherosclerosis (K = 0.69). Individual physicians varied widely; four agreed with the consensus diagnosis in all 18 cases, while six others disagreed with the consensus diagnosis in three to five cases. Our level of interphysician agreement is greater than that reported in other studies and was substantial. However, despite using subtype definitions and being given extensive information often not available in the acute setting, physicians still disagree about the etiology of stroke, particularly in regard to stroke due to small-artery occlusion or of undetermined etiology. Physicians seem reluctant not to attribute stroke to a specific etiology. The uncertainty about subtype diagnosis will affect interpretation of the results of clinical trials in patients selected by the subtype of ischemic stroke and also suggests that results of treatment as affected by subtype should be cautiously interpreted unless efforts to assure uniformity are included in the trial's operations. Refinement of algorithms for determining subtype of ischemic stroke do improve interphysician agreement. Such criteria should be applied strictly, and trials should include measures to assure the most uniform diagnosis of stroke subtype possible.