To observe microembolic events in the central nervous system during cardiopulmonary bypass, we subjected 21 patients undergoing elective coronary operations to retinal fluorescein angiograms 5 minutes before cardiopulmonary bypass was discontinued. Patients with diabetes or clinically evident cerebrovascular disease were excluded. Bubble oxygenation and nonpulsatile perfusion were used for bypass. After 31 to 167 minutes of cardiopulmonary bypass, all 21 (100%) patients had retinal microvascular occlusions indicative of microembolism. Control fluorescein angiograms obtained preoperatively and in five patients immediately before bypass but after aortic cannulation showed retinal perfusion. During bypass there was a mean of 3.5 (range 1 to 7) blocked arterioles of less than 50 micron caliber and a mean of 6.3 (range 1 to 10) focal areas of capillary nonperfusion per 30-degree field of retina centered on the macula per patient. Repeat studies 30 minutes after discontinuation of bypass showed partial reperfusion with occlusions in four of the five (80%) patients and a decrease in the mean total microembolic count from 12.6 to 4.8 (38%). In later studies at a median of 8 days postoperatively, only two of 16 (12.5%) patients had persistently occluded retinal vessels. The total microembolic count did not correlate with bypass time (r = 0.14) and was not reduced by arterial line filtration with a Pall 40 micron filter (Ultipore) in a subgroup of 11 patients. Twenty patients completed four standardized psychometric tests. Retinal microvascular occlusions were more numerous in those patients with a psychometric deficit (n = 7) than in those with no deficit (n = 13; p = 0.075). In a dog model of cardiopulmonary bypass, eight of nine (88.9%) had retinal microvascular occlusions after 10 to 90 minutes of bypass. Retinal histologic studies revealed intravascular platelet-fibrin microaggregates 20 to 70 micron in diameter and focal ischemic changes in seven of the nine (77.8%) dogs. Six dogs undergoing sham bypass had normal retinal perfusion and histologic findings. This study demonstrates a very high incidence of microvascular occlusions in the territory of the internal carotid artery during bypass consistent with a microembolic origin. The clinical significance of these findings is uncertain.