One major source of error commonly encountered in determining the fatigue life of a given element arises from the loads involved, which are usually of random nature. As a rule, resistance analyses, carried out by simulation or tests, are done by using a loading history accounting for the real loads involved that is repeated indefinitely until failure is encountered. The result is one of the many different lives that can be obtained with the various histories representing the same loading process. The life obtained with one of these loading histories is a random variable even if the other parameters influencing it are assumed to be constant. In studying fatigue crack growth one should therefore take account of the additional randomness introduced in defining a given loading history to be representative for the actual loads involved. In this work we studied the influence of the loading histories used to analyse lives in relation to the number of cycles and spectral density of the loading process. For this purpose, we performed fatigue crack growth simulations by using a model taking account of the sequence effect. Analyses were carried out on loading histories obtained from different spectra1 densities of the nominal stress and various loading history lengths for each.