Abstract In finite element analysis of soil-structure interaction problems involving firm to stiff overconsolidated clay, there have been difficulties in modelling the stress-strain response of the soil. Non-linearity and anisotropy of the soil depend on the inherent anisotropy of its particle structure and the induced anisotropy of its stress history and current stress path. In CRISP modelling of the centrifuge test of an abutment wall and its backfill of sand on the surface of a firm to stiff overconsolidated kaolin, the clay foundation was divided into 6 broad zones in accordance with the stress history and stress path. Undrained movements of the abutment and its subsoil were closely modelled in two analyses; one with a non-linear elastic model and the other with the Schofield model with shear modulus G assigned to the foundation zone in accordance with the estimated strain level as well as stress history and stress path. In the prediction of consolidation movement, there is a difficulty in the current critical state soil model in CRISP. The fe solution incorrectly predicted that substantial horizontal movement would accompany settlement due to consolidation, whereas the centrifuge test showed mainly vertical movement. This is attributable to the pronounced anisotropy separately observed in element tests.