The friction model and the pore model were studied comparatively, and the latter was modified in order to remove the discrepancy between these two models. Ultrafiltration data of six solutes of various molecular weights, obtained in our previous study, were analyzed by this model, but the results were not satisfactory.
Wall correction factors were then eliminated from the modified pore model. The data were reanalyzed by this new model, the "steric hindrance-pore model", and it became clear that this model adequately interpreted the experimental data. The values for membrane pore radius obtained were in good agreement. The A_k/ΔX value for each solute calculated was almost constant, and agreed well with that determined from pure water permeation.
The reason why this proposed model, from which the usual wall effect is eliminated, works well is not clear at this stage.