1. The interaction between Cl-, K+ and H2O fluxes were studied in the ventricular membrane of the choroid plexus epithelium from Necturus maculosus by means of ion-selective microelectrodes. The flux of H2O was measured by means of K+ electrodes as the dilution or concentration of intracellular choline ions, Ch+i. 2. In one series of experiments Cl- was readministered to the ventricular solution of tissues incubated in media with low Cl- concentrations. The resulting influx of Cl- was associated with an instantaneous influx of K+ and H2O. 3. Both the Cl- and the K+ influxes were reduced by the diuretic furosemide but were unaffected by inhibitors of Na+, K(+)-ATPase or changes in membrane potentials induced by Ba2+. Since the influx of K+ proceeds against its electrochemical gradient and is unaffected by changes in membrane potentials, the membrane exhibits secondary active, electroneutral transport of K+. 4. The influx of water, initiated simultaneously with the influx of K+ and Cl-, commenced before these ions had changed the osmolarity of the intracellular solution significantly. The influx of H2O could proceed against an osmotic gradient. The influx stopped when 100 mmol l-1 of mannitol was added to the ventricular solution at the same time as the Cl- ions. The influx of H2O was inhibited by K+ removal, furosemide or high external Ba2+ (10 mmol l-1), but not by strophanthidin, ouabain or low concentrations of Ba2+ (0.5 mmol l-1). The influx could not continue with other permeable anions, NO3-, acetate- or SCN-, replacing Cl-. 5. In another series of experiments Cl- was removed from the ventricular solution of tissues bathed in saline solutions with normal concentrations of Cl-. The resulting efflux of Cl- was associated with an instantaneous efflux of K+ and H2O. This efflux of H2O could proceed against an osmotic gradient of up to 70 mosmol l-1. This effect was inhibited by furosemide, in which case the water fluxes were entirely dependent on the osmotic gradients and the osmotic water permeability Lp of the ventricular membrane. 6. The data suggest that there is a coupling between the flux of KCl and of water in the ventricular membrane, which implies that the reflection coefficient sigma for KCl under the given circumstances is less than one. I suggest that the ability of leaky epithelia to transport against osmotic gradients depends on such a coupling, which derives from the properties of the proteins through which K+, Cl- and H2O leave the cell.