Critical defect size distributions are estimated for various metals and for hot pressed silicon nitride using the Monte Carlo method. With regard to the critical flaw size computation, results obtained for the numerous materials are compared to the sensitivity or probability of detection of particular flaw sizes based on existing nondestructive evaluation techniques. In several instances, defect detection is the limiting factor. Comparison of flaw estimates for the silicon nitride to fractographic observations demonstrates capability of order of magnitude accuracy. The technique is also used to explore the influence of variable parameters in a simple crack growth law. An important result was the nonnormality of life estimates, even though normal distributions were used for input variables. The consequence is a much lower probability of occurrence for the modal value of the life distribution than might be anticipated. With regard to the Monte Carlo method, the paper demonstrates that selection of the appropriate number of simulations must rely on consideration of third and higher order moments of the resulting statistical distributions. A large number of simulations were needed in order to obtain convergence of these moments. This, in turn, provides criteria for required number of simulations in the defect size and life computation. In summary, the paper illustrates the importance of statistically based fracture mechanics materials selection criteria. Furthermore, it demonstrates that the simple series approximation technique generally underestimates mean values and variance, particularly for large variances in operating stress and materials properties. Most importantly, the technique provides probability distributions for realistic fatigue life estimates.