This article addresses the problem of eyewitness identification errors that can lead to false convictions of the innocent and false acquittals of the guilty. At the heart of our analysis based on signal detection theory is the separation of diagnostic accuracy—the ability to discriminate between those who are guilty versus those who are innocent—from the consideration of the relative costs associated with different kinds of errors. Application of this theory suggests that current recommendations for reforms have conflated diagnostic accuracy with the evaluation of costs in such a way as to reduce the accuracy of identification evidence and the accuracy of adjudicative outcomes. Our framework points to a revision in recommended procedures and a framework for policy analysis.