It is well known that metals will break down under repeated application, and especially under repeated reversal, of stresses greatly less than those that have to he applied when the “ultimate strength” of the material is tested in the ordinary way. The researches of Wöhler have shown, for example, that iron capable of bearing about 20 tons per sq. inch of steady load will break when it is exposed to some millions of reversals of a stress of 8 or 9 tons per sq. inch, alternately in compression and extension. When the alternating stress is increased a smaller number of reversals suffices to produce rupture. On the other hand, examples such as are furnished in the balance-spring of a watch, or in a railway axle, show that very many million repetitions may be applied with impunity, provided the limit of greatest stress be kept sufficiently low. The mild steel axle of a railway carriage is exposed to many million reversals of a stress which, in some cases, approaches as high a value as 5 tons on the sq. inch, apparently with perfect impunity, for it seems probable that in the rare instances where fracture of such axles has occurred an explanation is to be found in the gradual spreading of a crack from an origin supplied by an air-bubble or other primitive defect in the material. But Wöhler’s researches, which have been confirmed by other observers, give evidence that a stress not very much greater than this, and far below not only the ultimate strength but even the “yield-point” of the metal, will produce what is called “fatigue” and bring about fracture when reversal of the stress is repeated many times. The purpose of this paper is to describe experiments in which the microscope has been applied to study the nature of the process of fatigue by which breakdown occurs under repeated reversals of stress. The experiments have been made during the past year in the Engineering Laboratory at Cambridge. The metal chosen for experiment was Swedish iron, of high and very uniform quality. It had the further advantage for our purpose of possessing a clearly defined and fairly large crystalline structure, well adapted when polished and etched to exhibit the characteristic lines known as “sliplines” or “slip-bands,” which appear in ordinary testing when any portion of the material has passed its limit of elasticity under strain. We used the metal in the form of rods with a rectangular section, the dimensions being approximately 0·3 inch by 0·1 inch, and to make the structure as uniform as possible these were in all cases annealed by being kept for about two hours at a dull red heat, while enclosed in a tube filled with lime, in a muffle furnace. One of the surfaces of each rod was polished and etched, and the rod was subjected to reversals of stress by bending, so that the polished surface was alternately extended and compressed. This was done, as in Wöhler's original experiments, by making the rod project from a revolving shaft with a load on the projecting end. As the process went on the rod was from time to time examined under the microscope, and in several cases photographs of the same crystals weie taken at each stage to record the progressive effect of repeated reversals of stress.