In current theories of polyacrylamide gel electrophoresis, the idea prevails that molecular sieving relies on different accessibility of volume fractions and of cross-sectional area fractions (denoted pores) to different-sized ions due to the effect of geometric exclusion. This correlates with the assumption that all elements of a polyacrylamide network occupy fixed and unchangeable positions, thus forcing colliding macro-ions to diffuse laterally in order to find an accessible pore and to resume motion in direction of the electrical field. The alternative conception would be equally well-justified, i.e., the assumption that polyacrylamide chains represent smooth obstacles cleared aside under the electrokinetic pressure of a macro-ion. This explanation would be preferable with respect to the molecular sieving effects occurring in solutions of liquid polyacrylamide. Yet no theory exists as to describe such effects in quantitative terms. A parameter is defined and discussed, which can be estimated by experiment, and which seems to be apt to characterize local resistivity of polymer structures against dislocation and deformation: the fractional specific resistance. Definition of this parameter is based on the model of a viscosity-emulsion composed of 2 interpenetrating liquid compartments characterized by different levels of hydrodynamic friction and the spatial dimensions of which are inferred from Ogston''s theory. This concept of localized viscosity may also serve as a link between theories of molecular sieving and of macroscopic viscosity of flexible polymers. The data of Morris, formerly taken as verifications of the rigid-pore concept, are now interpreted in terms of 4 factors responsible for size discrimination: collision frequency, duration of single contacts, size-dependent frictional force and the extent of cooperation among fibers, due to crosslinking and to simultaneous contacts of several fibers to a single macro-ion. Some functions relevant for problems of MW determination by gel electrophoresis are discussed in relation to the suggested model.