The ability of extracranial-intracranial bypass surgery to alter favorably the natural history of ischemic cerebrovascular disease remains in question. A recently completed prospective randomized multicenter cooperative trial failed to confirm the hypothesis that the procedure prevents further cerebral ischemia in patients with atherosclerotic internal carotid artery or middle cerebral artery disease. We analyze findings of the study in detail, including possible effects on the natural history of the disease beyond the immediate perioperative period. Potential sources of bias that may have unpredictably affected the study are discussed. These include observational bias (patient and therapist not blinded), "randomization-to-treatment" bias (high morbidity after randomization but before operation), and "prerandomization" or allocation bias (patients in the study representing a selected sample of the population with cerebrovascular disease). The extensive analysis of secondary subgroups with small numbers of patients is discussed in light of the statistical methods used. Two particular classes of patients not addressed in the study who might benefit from the procedure are defined. They are patients failing the best available medical therapy and patients with clearly documented hemodynamic compromise. Possible indications for bypass surgery are suggested for the various lesions in light of the trial and of recent reports on the natural history and pathophysiology of ischemic cerebrovascular disease. Selected cases illustrating these indications are presented.