1. Individual capillaries of the transilluminated frog mesentery have been perfused with suspensions of human red cells in frog Ringer solution containing 1-0 g albumin 100 ml.-1. The outer surface of the mesentery has been washed with normal frog Ringer solution and with frog Ringer solutions made hypertonic by addition of one of the following solutes: sodium chloride (100 m-mole. 1.-1); urea (100 m-mole.1.-1); sucrose (20-50 m-mole. 1.-1); cyanocobalamin (8-5 m-mole. 1.-1). The temperature of the mesentery was between 14 and 16 degrees C in all experiments. 2. Wtih the mesentery superfused with normal Ringer, the filtration coefficient was determined from measurements of the rate of fluid filtration across the capillary wall, at a series of known capillary pressures (Michel, Mason, Curry & Tooke, 1974). Filtration coefficient varied from 0-69 X 10(-3) to 4-45 X 10(-3) mum. sec-1 .cm H2O-1 with an average value of 1-87 X 10(-3) mum. sec-1. cm H2O-1. 3. When the superfusate was made hypertonic by the addition of a test solute, the osmotic reflextion coefficient (sigma) of the capillary wall to test solute was calculated from the additional rate of filtration, the concentration of test solute in the superfusate and the filtration coefficient. Average values for sigma were: sodium chloride, 0-068 +/- 0-03 (three capillaries); urea, 0-071 +/- 0.015 (four capillaries); sucrose, 0-115 +/- 0-023 (seven capillaries); cyanocobalamin, 0-100 +/- 0-03 (three capillaries). 4. In further experiments, the osmotic reflextion coefficients to sodium chloride, urea and sucrose were determined in the same capillary. Five technically acceptable experiments were carried out. Although there were differences in the value of sigma between different capillaries, in any one capillary values of sigma were of the same magnitude and there appeared to be no significant trend with the molecular size of the test solute. 5. Our findings are inconsistent with the hypothesis that there is a single pathway for water and small hydrophilic molecules across the capillary wall. 6. Our results may be interpreted in terms of an exclusive channel for water in parallel with a channel shared by both water and small hydrophilic molecules. It is suggested that the exclusive water channel may be the membranes and cytoplasm of the endothelial cells and the shared channel may be located in the intercellular junctions. 7. Our data suggest the exclusive water channel represents about 10% of the total filtration coefficient in frog mesenteric capillaries. The shared channel shows relatively little restriction to the molecules investigated. Estimates of the volume flow throught the two channels are made for conditions where hydrostatic pressure differences and osmotic pressure differences are the driving forces.