A method and a device for course prediction in motor vehicles, which have a position-finding system for objects situated in front of the vehicle, where a function describing the path of the roadway is calculated on the basis of measured distance and angular data, in that several fixed targets are identified and tracked and supplied (subjected) to a statistical evaluation, a plausibility criterion being that at least one parameter of the functions, which describe these roadway paths, has a significant frequency maximum at the value which corresponds to the real roadway path. At the beginning of the evaluation, the frequency distribution for all parameters is set to a predefined frequency value, and the frequency values of the parameters are reduced or increased by a predefined numerical value as a function of the position of fixed targets or vehicles.