In 1941 exposure tests were started on low-alloy steels in industrial and marine atmospheres. Analyses of selected rust samples are presented, along with an examination of weather data some weight losses. These results led to a new theory of the mechanism of rusting. It is postulated that corrosion rate depends on the quality and quantity of water reaching the steel surface. Relatively insoluble corrosion products decrease the amount of soluble constituents in the water and thereby decrease corrosion. Relatively insoluble corrosion products do not wash away, and thereby they decrease the porosity of the rust coating and the quantity of water reaching the steel. As time goes on pores become plugged and the position of low-alloy steels improves. In industrial atmospheres, copper and nickel in steel render sulfate corrosion products more insoluble by forming complex basic sulfates. On mild steel, sulfates in the rust are relatively soluble and promote corrosion but are washed away by rain. On alloy steels the sulfates are less soluble so that corrosion is slower, but less sulfate is washed away and more accumulates in the rust. The percentage of sulfates in the rust increases as with loss decreases. At the marine location, server corrosion of certain edges was due to soluble chlorides being swept to these edges by the wash of rain. Miscellaneous data and references are examined in the light of this mechanism. A factor which has been neglected in the literature is the composition of the film of electrolyte on the steel surface. It has been tacitly assumed that in a given atmosphere each test specimen of steel is exposed to the same environment. Actually, owing to the influence of corrosion products, each steel specimen is exposed to separate environment.