Discretion allows judges to tailor appropriate sentences to offenders with complex circumstances, but discretion also facilitates unwarranted sentencing disparity, that is, different treatment for similarly situated offenders. Guidelines aim to control unwarranted disparity by restricting sentencing variation and, in doing so, may inhibit proportionality or a punishment that fits the crime. Scholars have argued recently, based on extensive nonstatistical evidence, that the federal sentencing guidelines promote excessive uniformity. This paper proposes a statistical method for evaluating guidelines. Under illustrative assumptions about both guidelines and about how actual sentences deviate from appropriate sentences, we find that reductions in unwarranted disparity effected by the hypothetical guidelines studied do not offset the associated loss in proportionality. Because of the maintained assumptions of the analysis, we cannot condone or condemn guidelines on this basis alone. Rather, the results of this study restrict the admissible justifications of guidelines.