Diffusion studies in vitro indicate that Ficoll behaves more like an ideal spherical molecule than does dextran, suggesting that Ficoll would be a better probe of glomerular pore size than the commonly used dextran. To examine the differences between these macromolecules in vivo, the fractional clearances of tritiated Ficoll and dextran were measured over a wide range of molecular sizes (Stokes-Einstein radius, rs, from 19 to 65 A) in normal euvolemic Munich-Wistar rats. Whole-kidney and single-nephron hemodynamic conditions were characterized through a combination of clearance and micropuncture measurements. The fractional clearance, or sieving coefficient (theta), for dextran significantly exceeded that of Ficoll at all molecular sizes examined, theta for dextran being approximately 10 times that for Ficoll for rs greater than 30 A. Thus, the results with Ficoll imply a more size-restrictive barrier than do the results with dextran. The values of theta for Ficoll approximated previously reported values for uncharged globular proteins. Although theta for Ficoll at rs = 35 A was much smaller than the corresponding value for dextran, it was still approximately 30 times greater than typical values of the filtrate-to-plasma concentration ratio reported for serum albumin (a polyanion) in the rat, in agreement with the concept that glomerular charge-selectivity normally plays an important role in the prevention of albuminuria. Three membrane-pore models were compared in their ability to represent the dextran and Ficoll sieving data. A lognormal pore-size distribution in parallel with a nonselective "shunt" pathway was found to be more effective than either an isoporous membrane with a shunt or a purely lognormal distribution. On the basis of these laboratory results and computations, Ficoll may be preferred over dextran in future studies of glomerular size-selectivity.