An experimental model which duplicates bleeding and immediate replacement with blood substitutes was designed and used in 100 experiments on dogs. In hemo-dilution to asystole with dextrans 40 and 75 and with hydroxyethyl starch, four phases of response were seen: complete compensation, partial compensation, reversible decompensation (above 10% hematocrit), and irreversible decompensation (below 10% hematocrit). After hemodilution to 10% hematocrit with infusion volumes equal to blood withdrawn, long-term survival was achieved in 14 of 18 dogs (78%) with dextran 75, in 10 of 12 (83%) with dextran 40, and in 11 of 13 (84%) with hydroxyethyl starch; none of the dogs hemodiluted with buffered plasma or Ringer's lactate solution survived. Ringer's lactate solution, even in volumes of 2.5 times the blood volume withdrawn, did not maintain blood volume and cardiac output (after hemodilution to 20% and 10% hematocrit). Dextrans and starch resulted in bleeding diathesis below 20% hematocrit. In severe hemodilution, dextran 40 and hydroxyethyl starch proved as effective as dextran 75 in maintaining homeostasis and supporting survival. In treatment of moderate to severe blood loss, therefore, colloid solutions proved superior to crystalloid solution.