Under the formal rules of criminal procedure, fact finders are required to apply a uniform standard of proof in all criminal cases. Experimental studies as well as real world examples indicate, however, that fact finders often adjust the evidentiary threshold for conviction in accordance with the severity of the applicable sanction. All things being equal, the higher the sanction, the higher the standard of proof that fact finders will apply in order to convict. Building on this insight, this Article introduces a new paradigm for criminal punishments-a paradigm that focuses on designing penalties that will reduce the risk of unsubstantiated convictions. By setting mandatory penalties of sufficient size, the legal system can induce fact finders to convict only if sufficient admissible evidence proves a defendant's guilt. This Article applies this theoretical framework to three concrete contexts that involve a high risk of erroneous convictions: inchoate crimes, the right to silence, and the punishment of recidivists. It shows that a sanctioning regime that is attuned to the probative function of punishment can protect innocent defendants from unsubstantiated convictions while obeying the dictates of both deterrence and retribution.