Two simulations are involved when a centrifuge is used to test models. First, the behaviour of the model in a uniform ng field is assumed to be similar to that of the prototype. Then the centrifuge is assumed to produce an equivalent ng gravitational field. For most static problems, the centrifuge does produce an equivalent ng gravitational field, but for some dynamic problems involving saturated soil these assumptions can break down. When the soil particles and fluid are moving relative to one another, the behaviour in the ng field is not similar to that in the 1g field unless the Reynolds number in both conditions is less than unity. Since this is a special circumstance, the centrifugal behaviour is not similar to that of the prototype in most cases. To illustrate this, the similarity requirements are examined for a single particle moving in a fluid. If different fluids are used in the model and prototype, then the difference in densities must also be accounted for.