Retrograde cerebral perfusion (RCP) is a new method of cerebral protection that has been touted as an improvement over hypothermic circulatory arrest (HCA). However, RCP has been used clinically for durations and at temperatures that are "safe" for HCA alone. This study was designed to compare RCP to HCA and antegrade cerebral perfusion (ACP) deliberately exceeding "safe" limits, in order to determine unequivocally whether RCP provides better cerebral protection than HCA. Four groups of six Yorkshire pigs (20 to 30 kg) were randomly assigned to undergo 90 minutes of RCP, ACP, HCA, or HCA with heads packed in ice (HCA-HP) at an esophageal temperature of 20 degrees C. Arterial, mixed venous and cerebral venous oxygen, glucose and lactate contents; quantitative EEG; were monitored at baseline (37 degrees C); at the end of cooling cardiopulmonary bypass (20 degrees C); during rewarming (30 degrees C); and at two and four hours post intervention. Animals were recovered and were evaluated daily using a quantitative behavioral score (0 to 9). Mean behavioral score was lower in the HCA group than in the other three groups at seven days (HCA 5.8 +/- 1.1; RCP 8.5 +/- 0.2; ACP 9.0 +/- 0.0; HCA-HP 8.5 +/- 0.2, p < 0.05). Recovery of QEEG was better in the ACP group than in all others, but the RCP group had faster EEG recovery than HCA alone, although not better than HCA-HP (HCA 15 +/- 4; RCP 27 +/- 3; ACP 78 +/- 5; HCA-HP 19 +/- 3, p < 0.001). However, histopathological evidence of ischemic injury was present in 5 of 6 HCA animals and also in 4 of 6 of the HCP-HP group, but only in 1 of 6 RCP animals and in none of the ACP group. This study demonstrates that ACP affords the best cerebral protection by all outcome measures, but RCP provides clear improvement compared to HCA.