The purpose of the present study is to see if it is possible to estimate very short crack growth behavior using the data obtained from measurements on rather large scale fatigue crack growth. In this study, the material property was treated as a spatial random process along the path of the fatigue crack. The relation between the measured autocorrelation functions and averaging length was discussed for several correlation lengths. Applying the previously proposed method, the experimental data from stress-intensity-factor-controlled fatigue tests were analysed for different averaging lengths to determine the autocorrelation functions and probability distributions of parameters in Paris law. The main results obtained are : (1) The autocorrelation functions of Z (x) are almost same for different averaging lengths. (2) The variance decreases with the increasing averaging length δ. When the spatial correlation length is very small, the variance decreases significantly with the averaging length. (3) The autocorrelation functions of the material resistance Z (x) obtained from experiments are almost same for all δ, except for R (0). (4) The probability distribution functions of Z (x) show a slight dependence on the averaging length. As far as the present experiment is concerned, it can be concluded from these observations that a very small extension of a short crack can be predicted using the data obtained from measurements on rather large scale fatigue crack growth.