The theoretical aspects of the problem of sieve-like membranes are developed. The method of preparing the dried collodion membrane is described, and the method of defining the property of a particular membrane is given. It consists of the measurement of the Co P, that is the P.D. between an 0.1 and an 0.01 M KCl solution separated by the membrane. Co P is in the best dried membranes 50 to 53 millvolts, the theoretically possible maximum value being 55 millivolts. Diffusion experiments have been carried out with several arrangements, one of which is, for example, the diffusion of 0.1 M KNO3 against 0.1 M NaCl across the membrane. The amount of K+ diffusing after a certain period was in membranes with a sufficiently high Co P (about 50 millivolts or more) on the average ten times as much as the amount of diffused Cl-. In membranes with a lower Co P the ratio was much smaller, down almost to the proportion of 1:1 which holds for the mobility of these two ions in a free aqueous solution. When higher concentrations were used, e.g. 0.5 M solution, the difference of the rate of diffusion for K+ and Cl- was much smaller even in the best membranes, corresponding to the fact that the P.D. of two KCl solutions whose concentrations are 10:1 is much smaller in higher ranges of concentration than in lower ones. These observations are confirmed by experiments arranged in other ways. It has been shown that, in general, the diffusion of an anion is much slower than the one of a cation across the dried collodion membrane. The ratio of the two diffusion coefficients would be expected to be calculable in connection with the potential difference of such a membrane when interposed between these solutions. The next problem is to show in how far this can be confirmed quantitatively.