We have tested the proposal (M. Potschka (1987) Anal. Biochem. 162, 47-64) that the elution position of macromolecules by gel chromatography is better correlated with the viscosity-based Stokes radius (R eta) than with the Stokes radius (RS) calculated from the frictional coefficient. By the use of different gel matrices (agarose, Sephadex, TSK silica gel columns) we found that the elution of dextran fractions and reduced proteins, denatured with guanidinium hydrochloride, agreed with their R eta, using water-soluble, globular proteins for gel calibration. However, the elution of large sodium dodecyl sulfate-protein complexes and other elongated macromolecules (fibrinogen, tropomyosin) was systematically retarded while polyethylene glycol and polyethylene glycol detergents eluted earlier than water-soluble, globular protein as a function of R eta. The same was the case for bacteriorhodopsin, solubilized by C12E8 or Triton X-100. It is concluded that for steric reasons size exclusion chromatography is more sensitive than hydrodynamic measurements to the detailed conformation of macromolecules (rods and random coils) and that for this reason gels with inert pores cannot be universally calibrated for all kinds of macromolecules.