The general mechanism of transepithelial water transport remains unexplained. Uphill transport of water across epithelia, for example, cannot result from osmotic mechanisms; the driving forces are directed the wrong way. Such problems are circumvented in epithelial models which combine osmotic transport i.e. in aquaporins with direct coupling of water and substrate transport in cotransporters. The relations between substrate and water transport were studied in Na⁺-coupled cotransporters of glucose (SGLT1) and of iodide (NIS) expressed in Xenopus oocytes and compared to that of the uniport GLUT2. The water transport was monitored from changes in oocyte volume at a resolution of 20 pl, more than one order of magnitude better than previous investigations (Zeuthen et al. 2005). The rate of cotransport was monitored as the clamp current obtained from two-electrode voltage clamp. The data demonstrated a fixed ratio between the turn-over of the cotransporter and the rate of water transport under a variety of experimental procedures: isosmotic application of substrate, rapid changes in clamp voltage, or poisoning. Transport of larger substrates gave rise to less water transport: For the rabbit SGLT1 about 380 water molecules were cotransported along with the 2 Na⁺ ions and the glucose analogue α-MDG (MW 194), using the larger sugar arbutin (MW 272) the water coupling ratio was reduced by a factor of about 0.85. For the human SGLT1 the respective numbers were 235 and 0.85. For NIS about 250 water molecules were cotransported for each 2 Na⁺ and 1 thiocyanate (SCN^–, MW 58), with I^– as anion (MW 127) only 160 water molecules. The effect of substrate size supports a molecular mechanism for water cotransport and is opposite to what would be expected from unstirred layer effects. Fluxes into the oocyte via GLUT2 showed that diffusion of sugar in the cytoplasm did not give rise to significant unstirred layer effects; a similar conclusion has been reached for transport of Na⁺ ions maintained by the ionophore gramicidin. We conclude that coupling between water and substrate transport takes place by a molecular mechanism in the cotransporter itself. An epithelial model based upon the Na⁺/glucose and K⁺/Cl^– cotransporters and the GLUT2 uniport is presented; the K⁺/Cl^– cotransporter has previously been shown to cotransport about 500 water molecules per turnover (Zeuthen, 1994; Zeuthen & MacAulay, 2002). The ability of the epithelial model to transport water uphill is discussed.