In this study, an attempt has been made to implement back propagation network (BPN) for modeling the compression behavior of a saturated sand. In the implementation of BPN, data are categorized as input patterns and target patterns. The input patterns are fed to the network, which then performs feed-forward computations to calculate target patterns. A mapping between input patterns and target patterns can be achieved through internal learning algorithms of BPN, resulting in a network capable of simulating the target patterns for a given input patterns. The simulating data are acquired from the repeated Ko compression tests of saturated Ottawa sand performed by automatic triaxial test system, which was developed by author. The work presented in this paper demonstrates that the simulating results agree with measured data.