Brooklyn Law Review


The fate of defendants facing lengthy federal sentence enhancements often turns on what the U.S. Supreme Court calls the categorical approach. The approach controls whether a federal defendant might face an additional decade or longer in prison based solely on having prior convictions of a certain type. At a time when many question the wisdom of mass incarceration, the Court has taken great care to delimit the circumstances in which a federal sentencing judge can lengthen sentences based on recidivism. The categorical approach also governs most immigration cases involving deportation for a crime. As Congress has cut back deportation defenses for lawful permanent residents with criminal records, the Court has demanded that convictions used for deportation strictly correspond to a federal removal ground. In both Descamps v. United States and Mathis v. United States, recent Armed Career Criminal Act opinions authored by Justice Kagan, the Supreme Court interpreted the categorical approach as requiring a true elements test, declaring that at least two decades of the Court’s precedent required this result. Yet, Justice Breyer and Justice Ginsburg, two justices who had been in the majority in Descamps, dissented in Mathis. This article analyzes the trajectory of the Court’s categorical approach decisions, using the Mathis dissent authored by Justice Breyer and joined by Justice Ginsburg, to explain an ambiguity in the Court’s jurisprudence that Descamps and Mathis have now settled. Justice Kagan’s penchant for oversimplification has led her not only to overstate the relationship between the Court’s early and late categorical approach decisions but to include a discussion in Mathis that leaves room for confusion about how to apply the categorical approach in practice. This dictum suggests that adjudicators can take a “peek” at the record of conviction to help determine whether state law has defined a fact as a means or element. This article argues that Descamps and Mathis require the categorical approach to operate like a true elements test and that the take a “peek” suggestion in Mathis is inconsistent with the case’s holding.