The optimal hematocrit (Hctopt) hypothesis was tested by altering Hct (and arterial blood oxygen content, CaO2) between extreme states of anemia and polycythemia (Hct = 8-55%) in the rainbow trout. Since blood viscosity (eta) effects on cardiac output (Q) and O2 transport (TO2) are likely to be greatest when O2 demand and Q are maximal, we challenged fish to swim to their critical swimming velocity (Ucrit) in a swim-tunnel respirometer at 13 degrees C and measured maximal oxygen uptake (VO2max), maximum Q(Qmax), and other cardiovascular variables. In addition, experimental temperature was lowered to 5 degrees C to increase eta. Consistent with the Hctopt hypothesis, the decreased CaO2 in anemic (Hct < 22%) fish caused significant reductions in Ucrit and VO2max. In contradiction to the Hctopt hypothesis, and despite an exponential relationship between eta and Hct, maximal TO2 (TO2max) and Ucrit increased with polycythemia up to Hct 55%. Although there was a peak for VO2max, it occurred at an Hct (42%) well above the normocythemic range (23-33%). These results clearly demonstrate that eta is not significant in setting normocythemia in rainbow trout. The novel finding of an Hct-dependent relationship for exercise-induced arterial hypoxemia may be indicative of a diffusion limitation to normocythemia. We suggest that factors involved in setting normocythemia in vertebrates should include diffusion limitations to oxygen transfer in addition to blood viscosity and oxygen transport constraints.