The safety factor prescribed in codes and regulations is not always based on any reasonable background. Many researchers have endeavored to give to the safety factor a background which is based on probability of failure of structure and statistical nature of applied load. In order to obtain some information of probability of fracture in low cycle fatigue, experimental investigation was made into life distribution in strain controlled low cycle fatigue tests of quenched and tempered high tensile steels, HT 60 and HT 80, by testing 18 to 40 specimens at each of four strain amplitudes. Test results were examined by using normal, logarithmic normal, and Weibull distribution functions. It was found that distributions of low cycle fatigue lives agreed fairly well with logarithmic normal and Weibull distributions rather than normal distribution, and that scatter of crack initiation lives was, in general, larger than that of failure lives. Following relations were obtained between visible crack initiation life N_c and failure life N_f ; N_c =0.467 N ^1.043 _f for HT 60, and N_c =0.657 N ^0.953 _f for HT 80. The curve with 50% probability of fracture on N_c criterion was used, as a basis, for constructing _??_- N curves for 0.0001% probability of fracture, and modified _??_- N curves were also obtained by applying fatigue life reduction factor of 1/20 or strength reduction factors of 2/3 and 1/2 to the _??_- N curve of 50% probability of fracture. The concept of fatigue design curves were discussed by examining the strength and life reduction ratios.