1. Bidirectional paracellular fluxes using radioactive dextrans as inert molecular probes have been measured across Necturus gall-bladder epithelium during conditions of normal fluid absorption. There is a net flux at all radii analysed (0.4-2.2 nm) in the direction of fluid absorption. 2. The net flux is substantial at all radii within the range. The data extraplate to 2 x 10(-6) cm s-1 at zero probe radius, which is very close to the rate of epithelial fluid absorption. 3. The unstirred layers at the epithelial faces during transport have been determined; their contribution to the net fluxes is negligible. 4. Two possible mechanisms for the net flow of probes are considered: (i) that the probes diffuse across the junctions and are then entrained in a local osmotic flow along the interspaces and subepithelium; (ii) that the probes are entrained in volume flow across the junctions and the emergent solution subsequently passes through the interspaces and subepithelium. Model calculations clearly rule out mechanism (i) in which the maximum net flow obtainable is less than 10% of that observed. In addition the presence of leak paths shunting the junctions is not compatible with the observed fluxes. With mechanism (ii) the net flows are correctly predicted with all the fluid flow being transjunctional. The fluid absorption is therefore entirely paracellular. 5. The slope of the net flow curve shows no apparent change in magnitude over the range of the probe radii, indicating that effectively only one population of convective channels is present with parallel walls separated by about 7.7 nm. This agrees with the width previously determined by electron microscopy. 6. If the fluid absorption is junctional then the cellular route offers little if any relative contribution. The hydraulic conductivity of the junctions is not high enough, or the osmotic permeability of the membranes low enough, to accommodate this by osmosis and therefore the junctional fluid absorption must be non-osmotic.