We have analyzed glomerular sieving data from humans, rats in vivo, and from isolated perfused rat kidneys (IPK) and present a unifying hypothesis that seems to resolve most of the conflicting results that exist in the literature. Particularly important are the data obtained in the cooled IPK, because they allow a variety of experimental conditions for careful analysis of the glomerular barrier; conditions that never can be obtained in vivo. The data strongly support the classic concept of a negative charge barrier, but separate components seem to be responsible for charge and size selectivity. The new model is composed of a dynamic gel and a more static membrane layer. First, the charged gel structure close to the blood compartment has a charge density of 35-45 meq/l, reducing the concentration of albumin to 5-10% of that in plasma, due to ion-ion interactions. Second, the size-selective structure has numerous functional small pores (radius 45-50 A) and far less frequent large pores (radius 75-115 A), the latter accounting for 1% of the total hydraulic conductance. Both structures are required for the maintenance of an intact glomerular barrier.