Summary: We previously found mild hypothermia (34-36°C), induced before cardiac arrest, to improve neu rologic outcome. In this study we used a reproducible dog model to evaluate mild hypothermia by head cooling dur ing arrest, continued with systemic cooling (34°C) during recirculation and for 1 h after arrest. In four groups of dogs, ventricular fibrillation (no flow) of 12.5 min at 37 SC was reversed with cardiopulmonary bypass and defibrillation in �5 min, and followed by controlled ven tilation to 20 h and intensive care to 96 h. In Study A we resuscitated with normotension and normal hematocrit; Control Group A-I (n = 12) was maintained normother mic, while Treatment Group A-II (n = 10) was treated with hypothermia. In Study B we resuscitated with hy pertension and hemodilution. Control Group B-1 (n = 12) was maintained normothermic (6 of 12 were not hemodi luted), while Treatment Group B-II (n = 10) was treated with hypothermia. Best overall performance categories This study concerns two relatively unexplored concepts: (a) hypothermia initiated during ischemia (preservation) and continued after ischemia (resus citation) in contrast to hypothermia initiated before ischemia (protection), which is known to mitigate postischemic brain damage (Bigelow et aI. , 1950;Rupp and Severinghaus, 1986); (b) mild hypother mia (temperature 34-36°C) in contrast to moderate hypothermia (28-32°C), which is known to protect the brain (Bigelow et aI. , 1950) and may be harmful. This study involves cardiac arrest (i. e. , total body