Abstract: T HE PENETRATION of capillary walls by water and dissolved substances appears to take place solely by processes which require no energy transformations on the part of the capillary endothelial cells. The rate of net fluid movement across the capillary wall has been shown to be simply proportional to the difference between hydrostatic and osmotic forces acting across the capillary membranes (I, 2). The chemical composition of ascitic fluid (3), edema fluid (4) or glomerular fluid (5) closely resembles that obtainable by filtration of plasma through inert artificial membranes of suitable porosities. Filtration through peripheral capillaries, like that through artificial membranes, varies inversely with the viscosity of the filtrate as this is altered by temperature (6). These striking similarities between the permeability characteristics of living capillaries, on the one hand, and artificial porous membranes on the other, have given rise to the ‘Pore Theory’ of capillary exchange. In its simplest form, the pore theory supposes that the capillary walls are pierced with numerous ultramicroscopic openings which are in general too small to allow the passage of plasma protein molecules, but are of sufficient size and number to account for the observed rates of passage of water and nonprotein constituents of the plasma. Many important questions arise when the pore theory of capillary exchange is examined in detail. The glomerular membranes allow molecules as large as inulin (effective diffusion diameter, d, = 30 A) to pass with no detectable hindrance. Egg albumin (d, = 56 A) passes rapidly through the glomerular membranes (7-9) and even hemoglobin (d, = 62 A) is believed to pass into the glomerular filtrate in appreciable concentration (7, IO). Theoretical and experimental studies on the sieving of molecules by ultrafiltration through porous membranes (I I, 12) indicate that an effective pore diameter of at least IOO A would be required to explain permeability of this magnitude. If this is the case, how are we to imagine the physical structure of the openings in the capillary membranes? Do pores of diameter IOO A penetrate through the endothelial cells and their plasma membranes on both surfaces? Or are the openings confined to the narrow intercellular regions as postulated by Chambers and Zweifach (13)? In the latter case, is the small area available between cells sufficient to explain the observed filtration rates in such rapidly filtering systems as the kidney? If there is a distribution of pore sizes, or if water is free to