A method based on multivariate statistical analysis is developed which shows great promise of providing a definitive process for bed form prediction. This method uses statistics developed from flume, canal and river data in a discriminant equation which provides the basis for bed form prediction. Using four parameters, it is shown that approximately ninety percent success is achieved in correctly predicting bed form type. This method is limited at present by the accuracy and variety of the available data. It does provide, however, a needed method to apply more than two parameters to the problem of bed form prediction.