The thickening of oxide (or sulfide) films can occur in different ways. Outward migration of cations and electrons, normally by lattice defects, gives parabolic film growth, but mechanical breakdowns due to compressional stresses (of which good evidence exists) may lead to other equations. It is shown that blistering (likely if adhesion is poor and cohesion good) should lead to rectilinear or logarithmic thickening according as the rifts in the blister walls do or do not admit oxygen. Flaking (which should be rare but catastrophic) should lead to periodical recommencement of rapid oxidation. Shear‐cracking (likely if cohesion is poor and adhesion good) may produce parabolic, rectilinear or intermediate forms. The theory is applied to interpret the results of Vernon and others, and particularly to explain why conditions prevailing at the time when a specimen is first exposed decide the corrosion‐rate, which often continues unchanged even if subsequently conditions fluctuate.