This paper describes a novel approach to the theoretical prediction of fatigue damage and failure in bone. A simulation was set up in which the growth of every crack in the material was described individually. A modified fracture-mechanics equation was used to describe crack growth, using stochastic variables to simulate scatter. This model was cumbersome to set up initially, having a large number of degrees of freedom, but it proved to be capable of predicting a wide variety of experimental data, including: crack density, length and other damage variables; stress/life data, and the effect of specimen size. It was also possible to simulate the increased density of cracks that occurs in older people. Explicit models of this kind, which include the underlying mechanisms of the process, are more useful in the long term because they can be extended to include other mechanisms, in this case the processes by which living bones can repair cracks, and adapt their shapes to altered stress levels. (C) 2003 Elsevier Science Ltd. All rights reserved.