It has already been shown that in fatigue tests in vitro human cortical bone accumulates damage in the form of microcracks and that the total number of microcracks generated prior to the creation of the fatal macrocrack, and their effect (softening) on the material properties, depends on the level of applied stress. At each stress level the amount of accumulated damage has also been shown to be a non-linear function of the cycle number (Zioupos et al., 1996a, Zioupos et al., 1996b;​ Pattin et al., 1996). The theoretical implications of the previous findings and two possible models for cumulative damage were put to the test here by performing tensile fatigue tests in two-step level (high/low or low/high) loading on human cortical bone specimens. The results indicate that the accumulation of damage in-vitro is highly dependent on the level of stress and the stress history. Usual linear expressions for fatigue lifetime predictions, like the Palmgren–Miner rule, substantially over or underestimate the outcome depending on whether the stress was applied in a high/low or a low/high sequence, respectively. In view of these discrepancies we conclude that predicting the fatigue lifetime of any bone in vivo under variable loading and complex history regimes is an extremely difficult task to which the study of accumulation of damage can offer a significant but, perhaps, still limited contribution.