Updating behavior in cascade experiments is usually investigated on the basis of urn prediction. But urn predictions alone can only provide a very rough information on individual updating behavior. Therefore, we implement a BDM mechanism. Subjects have to submit maximum prices that they are willing to pay to participate in the prediction game. This enables us to study subjects' probabilty formation less crudely. The results show that in many situations herding occurs in accordance with the standard BHW model but cannot be explained neither by rational Bayesian updating nor by heuristics identi...ed in former cascade experiments.