A method for predicting object movements in a driver assistance system of a motor vehicle, in which the movements of objects located periodically by a locating device (12) are precalculated by dynamic modeling of the objects, characterized in that a plurality of dynamic models (16) are held in readiness which are based on different hypotheses about the object, and that the models are selected and/or weighted as a function of the situation in accordance with the correctness probability (P1, P2) of the hypotheses for the prediction.