When cells are frozen slowly in aqueous suspensions, the solutes in the suspending solution concentrate as the amount of ice increases; the cells undergo osmotic dehydration and are sequestered in ever-narrowing liquid-filled channels. Cryoprotective solutes, such as glycerol, reduce the amount of ice that forms at any specified subzero temperature, thereby controlling the buildup in concentration of those other solutes present, as well as increasing the volume of the channels that remain to accommodate the cells. It has generally been thought that freezing injury is mediated by the increase in electrolyte concentration in the milieu surrounding the cells, rather than reduction of temperature or any direct action of ice. In this study we have frozen human erythrocytes in isotonic solutions of sodium chloride and glycerol and have demonstrated a correlation between the extent of damage at specific subzero temperatures, and that caused by the action at 0 degrees C of solutions having the same composition as those produced by freezing. The cell lysis observed increased directly with glycerol concentration, both in the freezing experiments and when the cells were exposed to corresponding solutions at 0 degrees C, showing that the concentration of sodium chloride alone is not sufficient to account quantitatively for the damage observed. We then studied the effect of freezing in anisotonic solutions to break the fixed relationship between solute concentration and the volume of the unfrozen fraction, as described by Mazur, P., W. F. Rall, and N. Rigopoulos (1981. Biophys. J. 653-675). We confirmed their experimental findings, but we explain them differently. We ascribe the apparently dominant effect of the unfrozen fraction to the fact that the cells were frozen in, and returned to, anisotonic solutions in which their volume was either less than, or greater than, their physiological volume. When similar cell suspensions were subjected to a similar cycle of increase and then decrease in solution strength, but in the absence of ice (at 20 degrees C), a similar pattern of hemolysis was observed. We conclude that freezing injury to human erythrocytes is due solely to changes that occur in the composition of their surrounding milieu, and is most probably mediated by a temporary leak in the plasma membrane that occurs during the thawing (reexpansion) phase.