A stochastic model of age-related bone loss capable of predicting age-specific incidence of fractures has been implemented by Monte Carlo simulation. Each simulation involves aging a large cohort of individuals from 20 to 100 years. Every individual is randomly allocated a particular amount of bone as a young adult, an age of onset of bone loss, and parameter values that determine the subsequent bone loss. Fracture risk is assumed to be zero when the amount of bone is above a global threshold level, increasing progressively as the amount of bone falls below the threshold. From the individual fracture risks, a fracture subpopulation is identified and age-specific incidence evaluated numerically. By adjusting the model parameters, predicted and observed incidences of femoral neck fracture can be closely matched in both sexes by use of a linear function to describe age-related bone loss. For fracture of the distal radius, a close match can be achieved in women by use of an exponential function to describe the bone loss phase. In men, the incidence is independent of age; trauma, rather than the amount of bone in the forearm, appears to be the main determinant of fracture risk.