To observe and quantify cerebrovascular microembolic events in the central nervous system during cardiopulmonary bypass, 40 patients having elective uncomplicated coronary surgery had retinal fluorescein angiograms 5 min before bypass was discontinued. Each patient also had 10 neuropsychological tests before and after surgery. A Harvey H1700 bubble oxygenator was used for 23 patients and a Cobe CML sheet membrane oxygenator was used for 17 patients. All 23 (100%) of patients in the bubble oxygenator group had retinal microvascular occlusions consistent with microembolism compared to 8/17 (47%) in the membrane oxygenator group (P less than 0.001). In those retinas with occlusions, the mean resultant area of non-perfusion was less in the membrane oxygenator group (0.11 mm2; n = 8) than in the bubble oxygenator group (0.29 mm2; P less than 0.01). Arterial PO2 levels during bypass were similar in both groups at moderate hypothermia, but the mean PaO2 during rewarming was higher in the bubble oxygenator group (27 kPa) than in the membrane group (13 kPa; P less than 0.001). Neuropsychological deficits were more common and more severe after bubble oxygenation than after membrane oxygenation, but in this small patient group, the difference was not statistically significant. We conclude that flat sheet membrane oxygenation during cardiopulmonary bypass may confer significant protection against cerebrovascular microembolism.