Fasting serum glucose, glycosylated hemoglobin, and fructosamine concentrations were determined in 304 consecutive subjects admitted with acute stroke, within 48 hours of ictus. Based on the medical history and these results, subjects were divided into known diabetics, newly diagnosed diabetics, subjects with stress hyperglycemia, and nondiabetics. The type of stroke was classified as lacunar infarct, cerebral infarct, or intracerebral hemorrhage, based on clinical examination by a neurologist and computed tomographic brain scan and/or autopsy. Immediate and 3-month outcomes were examined in relation to (1) fasting glucose, glycosylated hemoglobin, and fructosamine levels by stroke subtypes, and (2) glucose tolerance categories by stroke subtypes. A high fasting glucose level was associated with an increased mortality, but this was observed only among patients with intracerebral hemorrhage. Patients with stress hyperglycemia, but not diabetics, had increased mortality. In spite of having similar glucose concentrations to those patients with stress hyperglycemia, diabetics did not have a worse outcome compared with nondiabetics. It is concluded that the association between glucose concentration and outcome is a reflection of stress relating to stroke severity, rather than a direct harmful effect of glucose on damaged neurons.