Tensile stress-induced crazing in polystyrene, poly(methyl methacrylate), and polycarbonate has been carefully examined by optical and electron microscopy. Examination of the surface of crazed specimens and the cross sections of individual crazes leads to the conclusion that the crazes are not void cracks, but are filled with a craze matter. The craze matter is readily distinguishable from the surrounding resin and is seen to exist in continuity with it. Further experiments confirm the existence of the craze matter and tend to indicate its structural and mechanical properties. These experiments include: microscopic examination of the walls of fractured crazes, micro x-ray diffraction of craze matter, studies of the strength of crazed specimens under static loads and under increasing tensions, studies of the effect of heat and solvent on crazed specimens, and observations on the ability of crazes to form special networks. The results indicate that the craze matter may be formed by localized resin deformation leading to a load bearing oriented structure. A hypothesis of the mechanism of craze formation is proposed in light of the new and varied information reported in the paper. Study of the kinetics of craze growth suggests the division of the mechanism into three parts: initiation, propagation, and termination. The initiation step describes the concentration of strain energy and the precursory molecular arrangements in the immediate vicinity of inhomogeneities. These are the changes which occur during the time lapse between stress application and first appearance of crazes. The propagation step comprises the sudden and relatively rapid localized resin deformation which creates the craze matter. The shapes of the growing crazes and their diminishing growth rates are attributable to known resin properties. The termination step represents the apparent cessation of craze growth with continued long times of stress application. At this stage in the existence of crazes the proposed hypothesis must blend into the theories of time delayed rupture. It is concluded that stress crazing is basically a molecular slippage rather than a molecular cleavage phenomenon. Considerations, therefore, are to be directed more towards intermolecular forces than intramolecular forces.