During hyperfiltration of a salt solution through a charged clay membrane two types of electrical potential gradients can develop, i.e. a streaming potential and a membrane potential. In the absence of electrical shorting, hydraulic flow induces a streaming potential gradient by the drag of excess double layer cations in downstream direction. In turn this potential difference hampers the further downstream movement of ions, and therefore of water. Thus the streaming potential causes an electroosmotic counterflow of solution. A membrane potential gradient develops when the membrane interfaces are in contact with solutions of different salt concentrations. It arises in part from the difference in ionic mobility between cations and anions and is enhanced by the perm-selectivity of the membrane, e.g. the exclusion of anions in clay membranes. Also thís potential difference hinders the downstream movement of ions, be it by convection or diffusion, and works as an electroosmotic counterflow of solution. For perfectly semi-permeable membranes the membrane potential difference is given by the Nernst-equation, with the positive potential at the low concentration side of the membrane. Since clay membranes are not ideal the membrane potential difference is smaller than this maximum value. For bentonite clay we measured membrane potential gradients between 3.9 and 5.8 V/m, positively correlated with NaCl-concentration gradients between 1.2x104 and 3.1x104 mol/m4. In a laboratory hyperfiltration experiment with bentonite reported in the literature, a salt concentration increase is observed at the inflow side of the membrane, which is ascribed to semi-permeability of the clay causing reverse osmosis. The clay specimen was not electrically shorted; therefore, although not considered, a streaming potential gradient and an electroosmotic counterflow must have been present. For estimation of the streaming potential gradient in this experiment we considered several literature sources, amongst them our own work. Estimated gradients of streaming potential in this experiment vary between –41 and –1227 V/m. Using coefficients of electroosmotic conductivity and streaming potential gradients reported in literature, an electroosmotic counterflow of water of the same order of magnitude as the applied downstream flow is predicted. A maximum value of the membrane potential gradient in the experiment is calculated with the Nernst-equation to be 12 V/m. Since for non-ideal membranes the membrane potential gradient is substantially smaller than the Nernst value, it will have been negligible in the experiment compared to the streaming potential. The counterflow of water will have an effect on the required pressure on the input solution. This pressure must have been higher than in the absence of the streaming potential, i.e. in shorted condition. Taking electroosmotic counterflow into account leads to the conclusion that the semi-permeability of the clay (quantified by a reflection coefficient) is much smaller than predicted when an induced streaming potential is neglected. The observed salt accumulation is influenced by the electroosmotic counterflow.