Are the original Donnan assumptions applicable to serum?

Analyses of the diffusible fraction of serum for sodium, potassium, and chloride have indicated that the ratios of the levels of the protein solution to those of the filtrate do not follow the predicted Donnan values. Specifically, measurements by sodium glass electrode showed that sodium activities are the same for both protein solution and filtrate in contradiction to the Donnan model which predicts greater activity for the sodium in the protein solution because of its greater analyzed content. An alternative model that fits the experimental findings is discussed, and calculations made possible by this model are presented that give reasonable values for bound sodium, bound water, and a radius for an assumed spherical albumin molecule. A review of Donnan's original data discloses discrepancies in his assumptions.