To assess steady-state transepithelial osmotic water permeability (Pf), rabbit proximal convoluted tubules were perfused in vitro with the impermeant salt, sodium isethionate at 26 degrees C. Osmotic gradients (delta pi) were established by varying the bath concentration of the impermeant solute, raffinose. When lumen osmolality was 300 mOsm and bath osmolality was 320, 360 and 400 mOsm, apparent Pf decreased from 0.5 to 0.10 to 0.08 cm/sec, respectively. Similar data were obtained when lumen osmolality was 400 mOsm. Five possible causes of the delta pi dependence of apparent Pf were considered experimentally and/or theoretically: (1) external unstirred layer (USL); (2) cytoplasmic USL; (3) change in surface area; (4) saturation of water transport; (5) down-regulation of Pf. Apparent Pf was inhibited 83% by p-chloromercuribenzene sulfonate (pCMBS) at 20 mOsm, but not at 60 mOsm delta pi, suggesting presence of a serial barrier resistance to water transport. Increases in perfusate or bath solution flow rate and viscosity did not alter apparent Pf, ruling out an external USL. A simple cytoplasmic USL, described by a constant USL thickness and solute diffusion coefficient, could not account for the delta pi dependence of apparent Pf according to a mathematical model. The activation energy (Ea) for apparent Pf increased from 7.0 to 12.5 kcal/mol when delta pi was increased from 20 to 60 mOsm, not consistent with a simple USL or a change in membrane surface area with transepithelial water flow. These findings are most consistent with a complex cytoplasmic USL, where the average solute diffusion coefficient and/or the area available for osmosis decrease with increasing delta pi. These results (1) indicate that true Pf (at physiologically low delta pi) is very high (greater than 0.5 cm/sec) in the rabbit proximal tubule; (2) provide an explanation for the wide variation in Pf values reported in the literature using different delta pi, and (3) suggest the presence of a flow-dependent cytoplasmic barrier to water flow.