Partition of sized FITC-dextrans in polyacrylamide gel showed a relationship between Kav and solute radius as predicted by the theory of Ogston, which is based solely on geometry of the spaces. Permeability data for the same dextrans were fit to several theories, including those based on geometry and those based on hydrodynamic interactions, and the gel structure predicted by the partition and permeability data were compared. The Brinkman effective-medium model (based on hydrodynamic interactions and requiring a measure of the hydraulic conductivity of the matrix) gave the best fit of permeability data with the values for fiber radius (rf) and void volume of the gel (epsilon) that were obtained from the partition data. The models based on geometry and the hydrodynamic screening model of Cukier, using the rf and epsilon from partition data, all predicted higher rates of permeation than observed experimentally, while the effective-medium model with added term for steric interaction predicted lower permeation than that observed. The size of cylindrical pores appropriate for the partition data predicted higher rates of permeation than observed. These relative results were unaffected by the method of estimating void volume of the gel. In sum, it appears that one can use data on partition of solute, combined with measurement of hydraulic conductivity, to predict solute permeation in polyacrylamide gel.