The requirement of unanimous jury verdicts in criminal trials is widely believed to reduce the likelihood of convicting the innocent. This belief depends largely upon the assumption that jurors will vote nonstrategically based on their impression of the trial evidence. Recent literature, however, has questioned this assumption, and Feddersen and Pesendorfer propose a model in which it is never a Nash equilibrium for jurors to vote nonstrategically under unanimity rule, and equilibrium behavior produces higher probabilities of both convicting the innocent and acquitting the guilty under unanimity rule than under numerous alternatives. I extend this work by incorporating two additional features of actual jury procedure: the possibility of mistrial and communication among jurors. Under each circumstance, I demonstrate that nonstrategic voting is a Nash equilibrium under fairly general conditions and that unanimity performs better than any alternative rule in minimizing probability of trial error and maximizing expected utility.