Earlier experiments have shown that the friction between metals is due to the shearing of junctions formed by adhesion or welding at the points of intimate contact. This suggests that when the load is removed the junctions should remain and an appreciable normal force should be needed to separate the surfaces. Experiments show that with clean hard surfaces in dry air the adhesion is negligibly small. In moist air appreciable adhesion may be observed, and it is shown that this is due to the surface tension of a thin film of adsorbed water. The surface-tension forces due to thin films of liquid trapped between solid surfaces may be very large. Under certain conditions the viscosity of the liquid may also be important. The absence of adhesion between clean hard surfaces is not due to the non-formation of metallic junctions. Experiments show that it is due to the released elastic stresses which break the junctions one by one as the load is removed. With very soft metals, such as lead or indium, where the effect of released elastic stresses is very much less important, marked adhesion is observed in air, if the surfaces are freed of grosser contaminants. This adhesion provides direct evidence for the formation of metallic junctions by a process of cold welding or pressure welding at the points of contact. If the surfaces are covered with oxide films of appreciable thickness, the amount of metallic interaction is diminished with a corresponding reduction in the adhesion. Lubricant films have a similar effect, and in general those materials which are most effective in reducing the adhesion are also most effective, as boundary lubricants, in reducing the friction.