Recent studies have suggested that oxygen-derived free radicals play an Important role in ischemia-reperfusion injury of the spinal cord. In other organ systems, reperfusion injury has been reduced by limiting the availability of oxygen in the reperfusion phase. The purpose of this study was to test the effect of normovolemic hemodilution and gradual reperfusion on spinal cord function after aortic cross-clamping in 84 New Zealand White rabbits. All animals underwent 21 min of infrarenal aortic cross-damping in the conscious state by means of a previously placed aortic occlusion device and were randomized to four groups. Group 1 animals were hemodiluted to a mean(s.e.m.) hematocrit of 28(2)% by extracting 25% of the effective blood volume and reinfusing the plasma component after centrifugation concurrently with a volume of normal saline three times that of the discarded red cells. Group 2 animals (controls) were bled similarly but both plasma and red cells were reinfused, resulting in a mean(s.e.m.) hematocrit of 38(2)%. In the next two groups, distal aortic flow was recorded via an implantable Doppler device. After cross-clamping, flow was returned gradually over 45 min in animals of group 3. and abruptly in group 4. Animals were observed for 5 days and neurologic function was graded by an independent observer. Paraplegia at 5 h after clamping occurred in 75% of animals in group 1 versus 32% in group 2 (P< 0.05), and in 33% of group 3 versus 28% in group 4 (not significant). Of those animals showing initial neurologic recovery, delayed-onset paraplegia was seen in 100% in group 1 versus 87% in group 4 (not significant), and in 50% of group 3 versus 92% of group 4 (P< 0.03). The findings that the incidence of delayed-onset paraplegia in rabbits is reduced after gradual reintroduction of blood flow supports the hypothesis that this phenomenon is caused by reperfusion injury. Ischemic spinal cord injury is worsened by normovolemic hemodilution in this model.