When capillary condensation takes place, it is accompanied by multilayer adsorption in empty capillaries. To apply the Kelvin equation to pore-size distributions, it is necessary to make allowance for multilayers, and a method of carrying this out is presented. Any such calculation requires either an estimated or experimental adsorption isotherm for the same surface when capillary condensation is absent. In this paper, it was possible to apply it to the data of part I, in which adsorption isotherms were compared for loose powders and for the same powders compressed into porous plugs. Modification of the calculation when capillaries fill by blockage instead of condensation is discussed. Blockage can prevail over capillary condensation only in pores with radii < 8 to 10 Å, and even then only if heats of adsorption are high, so that a large fraction of the surface is covered at low relative pressures. Owing to presence of adsorbed layers, the Kelvin radius is less than the true pore radius, but it is difficult to decide upon the proper correction, and much uncertainty results for radii < 30 Å. Pore-size distribution curves were derived and compared for plugs of varying porosity. They provided strong evidence of tightly packed aggregates of particles, and of an increasing uniformity of pore size as porosity is decreased. In Carbolac 1 plugs of low porosity, a few pores with radii only a little larger than the radius of a CF
2
CI
2
molecule were encountered, but no pores were smaller. This agrees with the observation that the total particle surface was initially accessible to CF
2
CI
2
molecules, but that a rapid decrease of accessible surface took place before even a complete monolayer had been adsorbed.
