The standard model of optimal deterrence predicts that the probability of wrongful conviction of innocents is at the margin as detrimental to deterrence as the wrongful acquittal of guilty individuals. We extend the model in several directions: using expected utility as well as non expected utility to consider the role of risk-aversion, non-linear probability weighting and loss-aversion. We also consider how relevant emotions such as guilt, shame and indignation play out. Several of these factors support the intuition that wrongful convictions of innocents do have a larger detrimental impact on deterrence and thus the policy implications is reconciled with the widely shared maxim that in Dubio Pro Reo. We then draw some theoretical implications such as a novel justification for the different standard of proof in criminal vs civil law as well as other policy implications.