We experimentally investigate two questions that must be understood to effectively implement important normative prescriptions of optimal deterrence theory: i) does a non-monetary punishment and a fine of equivalent monetary value produce the same level of deterrence, and ii) should severe procedures, which maximize correct convictions of guilty defendants, be preferred to lenient procedures, which minimize errors in cases against innocent defendants? We examine these questions in an experiment where potential thieves face the possibility of punishment. As a non-monetary sanction, we require convicted individuals to perform a tedious real effort task. In the monetary treatments, sanctions are instead fines, which are based on individuals’ willingness to pay to avoid the real effort task to ensure comparability with the non-monetary treatment. The second manipulation in our experiment concerns the balance of errors in the adjudicative procedure (i.e. the conviction of innocents and acquittal of guilty individuals). We find that stealing is reduced most effectively by a sanction regime that combines non-monetary sanctions with a severe procedure. Our data are consistent with the notion that both monetary punishment and pro-defendant sanction regimes are less effective in communicating moral condemnation of an act.