In each actual test of a metal the stages are delimited by breaks, generally observed on kinetic curves plotted in the coordinates of log v versus log AK. However, there are also other physical grounds for such division into stages. In particular, it is thought that in stage II the kinetics of crack growth may be uniquely described by the Paris equation da/dN = C(AK) n, where C and n are constants of the material. At the same time, it is shown that in stage I the rate of crack growth depends substantially also on the initial loading conditions, including crack length and load amplitude [3]. Similar complications arise in attempting an analytical description of crack growth in stage ItI. There are grounds for stating that the kinetic features of propagation of fatigue cracks in alloys must be reflected in the microstructure of segments of the cracks corresponding to the individual stages of growth. In this study, fractographic analysis may be considered a means of investigating the micromechanism of the fatigue process. In recent years fractographic investigations have been widely carried out in connection with study of the kinetics of cracks in aluminum and titanium alloys [7, 8]. For structural steels such works are not very extensive. Investigations of this type deal chiefly with crack growth in steels with equilibrium structure [4]. The present work is devoted to a fractographic study of the growth mechanism (in all three stages) of fatigue cracks in high-strength, low-temper steels. For these investigations the steels 50KhN and ShKh15 were selected, differing primarily in a substantial difference in carbon content. They were subject'ed to heat treatment: quenched in oil from a temperature of 860~ with subsequent 2-h tempering at 150~ (50KhN) and 200~ (ShKh15). Kinetic fatigue curves were plotted on the basis of tests of beam specimens (6 !5 130 ram) having a crack, with cantilever bending in an asymmetrical cycle (R(~ =-0.33). The testing and processing of the resulting experimental data were carried out in accordance with the method discussed in [9]. The kinetic curves for each specimen were plotted on the basis of identical initial conditions (crack length and size of AK). The fractographic analysis was based on an emission electron-microsc opic examination of the fracture surface [10] with growth of the crack and, correspondingly, with increase in AK. The investigation also included optical analysis of the crack profile (first coated with nickel) and the structure of the crack tip (on polished side surfaces of the specimen). It was established first that with the chosen width of specimen (6 mm) and high strength of the investigated steels (HRC ~ 58-60) crack development is generally charPhysicomechanica l Institute, Academy of Sciences of the Ukrainian SSR, L'vov. Translated from Fiz