As soil behaviour is stress-dependent, a centrifuge model with tilting sample box was developed to generate liquefaction flow slides at higher confining stresses. The design and experimental set-up were based on the large liquefaction tank (de Jager et al., 2017). Fluidisation was used as sample preparation technique to produce a saturated, loose and uniform sand bed. The performance of the fluidisation system was evaluated by experimental investigation of the sample by considering the relative density, uniformity, degree of saturation and the influence of viscous pore fluid. The reproducibility of the initial sample was considered acceptable. A series of centrifuge experiments was conducted where the fluidised sand bed was accelerated to varying gravity levels and inclined to a slope with constant tilting rate. In most tests the soil response was characterised by a rapid liquefaction flow slide and a sudden increase in pore pressures. The moment of failure was consistently influenced by a variation in fluid viscosity and tilting rate, regardless of the gravity level; these effects indicated that instability was caused by the restricted seepage rate during loading. It is believed that the liquefaction potential is governed by the extremely loose and highly contractive top layer, which yields a sudden loss of strength under limited drainage conditions. The pore pressure measurements, which showed no excess pore pressures building up prior to failure, can lead to a misunderstanding of the failure mechanism and false assumption of fully drained conditions. Monitoring the pore pressures is therefore not suitable to predict liquefaction flow slides in submarine slopes. Mitigation of liquefaction should be focussed on densification of the looser part of the sandy slope, which is usually the top layer.