Pitting, as the term is used in this paper, means a form of failure which occurs on gear teeth near the pitch circle, on rollers, in anti-friction bearings, and on other machine parts where cylindrical surfaces come in rolling contact under heavy loads. The purpose of the investigation described in this paper is to study the manner in which pits develop, to discover the reason for their development, and to find what conditions must be fulfilled to prevent their occurrence. A testing machine was used in which steel rollers approximately 1.5 in. in diameter, rolling together and loaded as shown in Fig. 2, could be tested. More than eighty tests have been made with rollers of different materials, with different lubricants, and with various loads. Pits are found to have a definitely characteristic shape and orientation with respect to the direction of rolling. They were observed to develop from minute cracks in the roller surfaces which appear after about 500,000 revolutions. Photographs were taken tracing the growth of these pitting cracks from their earliest observed stage. The cracks from the very beginning have a characteristic shape and orientation. The principal findings in the tests on the steel rollers were: (1) That a lubricant must be present if pitting is to take place; (2) that if the lubricant is of a viscosity above a certain critical value which depends on the load, pitting can be prevented; (3) that the nature of the surface finish on the rollers greatly influences the tendency to pit, pitting being prevented on a highly polished surface and accelerated on a rough-machined surface; (4) that nitrided rollers will not pit under conditions that would result in severe pitting of quenched and tempered mild carbon-steel rollers, also that pits are smaller on harder surfaces, in general. In the theoretical section of the paper, the load-carrying capacity for the oil film between two rollers having a common peripheral velocity is calculated, and it is shown that the critical viscosity necessary to prevent pitting is far below that necessary to prevent metal-to-metal contact, a fact also checked experimentally. The normal and tangential oil forces on the rollers after metal-to-metal contact are also calculated, and the tangential component is found to be small compared with the normal component. Calculation showed that no tensile stress would exist at the surface, where pitting cracks appear to start, due to contact pressure or oil forces. However, it is found by analysis that a small crack in the surface, if filled with oil, will tend to grow, provided it has a certain initial direction. This direction is the same as that of the cracks which precede pitting. This strongly suggests that oil penetration of very small surface cracks with a certain initial direction is the reason for the growth of these cracks until a particle is separated from the main body of material, leaving a pit. The experimental results harmonize, also, with this theory.