Agents' beliefs in the imminent reform of a particular money supply process can have a powerful effect on their predictions of such variables as the rate of inflation. To find a criterion for monetary reform, we argue that any money supply process which does not provide a finite solution for price in a Cagan-type hyperinflationary money market will be rejected by the public. Such a process does not have the basic property of money which we call "process consistency," and we suggest that agents' subjective probabilities that their money is "process consistent" are identical to the probability that they attach to a currency reform. We compute agents' subjective probabilities that a particular money supply process is process consistent for the explosive part of the German hyperinflation, and we find that the probability of process consistency reached its lowest level in the very week in which the reform started.