An electrophysiological approach is used to analyze the possible routes of osmotically driven water flow across the isolated perfused Ambystoma proximal tubule. The minimum hydraulic conductivities (Lp) of the cell membranes were estimated from the initial rate of change of intracellular activities of Na+ and K+ in response to a step gradient of 50 or 100 mosmol/kg sucrose. The Lp of the apical membrane is 1.30 X 10(-4) cm.s-1.osM-1 referred to the luminal epithelial surface and 2.45 X 10(-6) cm.s-1.osM-1 when corrected for amplification of the brush border (n = 8). The Lp of the basolateral membrane is 1.42 X 10(-4) cm.s-1.osM-1 referred to the basement membrane surface and 6.39 X 10(-6) cm.s-1.osM-1 when corrected for the amplification of the basal and lateral membranes (n = 5). Transepithelial water flows were generated in either direction by a unilateral step increase of osmolality with 100 mosmol sucrose. Bath-to-lumen flow increased paracellular transepithelial resistance (R3) by 48%; lumen-to-bath flow decreased R3 by only 3%. A bilateral increase in the osmolality of both solutions by 50 mosM had no significant effect on R3. Streaming potentials were observed during trans-epithelial water flow induced by unilateral gradients of sucrose; their polarity, magnitude, site of generation, and insensitivity to change of paracellular resistance are all indicative of water flow through paracellular structures, especially the lateral intercellular spaces. Contrary to earlier suggestions (J. M. Diamond, J. Membr. Biol. 51: 195-216, 1979), these potentials are not primarily diffusion potentials across anion-selective tight junctions resulting from solute polarization in the unstirred layers. Instead, a true electrokinetic basis for these streaming potentials is indicated by their continued presence after deletion of all Cl-. Thus water moves through both cellular and paracellular pathways in this epithelium.