Plurality Opinions and the Marks Rule
Freeman v. United States asked a fairly narrow sentencing question in 2011: could a defendant who pleaded guilty under an agreement fixing a specific prison term later ask for a reduced sentence once the Sentencing Commission lowered the guideline range that agreement had referenced? Four justices, in a plurality opinion, said the guideline range always counts as the basis for a sentence, even one fixed by agreement, so the defendant qualified. Four dissenting justices said a defendant sentenced under that kind of fixed-term agreement never qualifies at all. Justice Sotomayor, writing only for herself, landed in between: eligible, but only where the agreement itself expressly used a guideline range to arrive at the number. Nobody commanded five votes for any reasoning, only for the result. For the next seven years, defendants in different circuits got different answers to the identical question depending on which of three opinions their court of appeals decided actually spoke for the Court.
That last part is the failure worth dwelling on, because it is a failure of Marks itself, not merely a disagreement about sentencing law. Some circuits read Sotomayor's solo opinion as the narrowest grounds and treated it as controlling. Others concluded her plea-agreement-based reasoning was not a logical subset of the plurality's guideline-always-counts reasoning at all, just a different theory that happened to produce overlapping results in some cases, and held that Freeman therefore controlled nothing beyond its own four corners. Both readings are defensible applications of the same 1977 rule to the same 2011 opinion, which is itself the clearest evidence that the rule does not resolve what it claims to resolve. Two defendants who pleaded guilty under identical Type-C agreements, sentenced to identical terms, could ask for the identical guideline-driven reduction and get opposite answers depending only on which circuit had custody of their case, not on anything either defendant did differently or any distinction Congress ever drew.
The logical-subset test King v. Palmer built sounds precise until it meets an opinion like Sotomayor's. A narrower opinion is supposed to control when anyone who agreed with the broader opinion would necessarily have agreed with the narrower one too, the way a rule limited to red cars is a logical subset of a rule covering all cars. Kennedy's plurality would have made every Type-C defendant whose plea referenced a guideline range eligible; Sotomayor's opinion made only some of that same group eligible, the ones whose agreements used the range in a particular textual way. Every defendant who satisfies Sotomayor's narrower test also satisfies Kennedy's broader one, which looks at first glance like exactly the nested relationship King v. Palmer requires. But Kennedy's opinion never mentions the textual distinction Sotomayor's turns on, and a judge applying only the plurality's own stated reasoning would have no way to know that distinction mattered at all. Whether one opinion counts as a subset of another, it turns out, depends on how generously a court is willing to read the broader opinion's silence, and that judgment call is exactly the kind of thing Marks was supposed to make unnecessary.
Hughes v. United States gave the Court a direct chance to settle which reading was correct, since the question presented asked exactly that. The six-member majority did not take it. It held, over a three-justice dissent, that a guideline range matters even to a sentence fixed by a Type-C plea agreement, arriving at essentially the plurality's old destination by a fresh route the majority argued out on its own terms, and simply had no occasion left to say whether Sotomayor's narrower opinion had been binding on lower courts for the seven years in between. The sentencing question got answered. The interpretive question that had actually divided the circuits did not.
The same pattern shows up far outside criminal sentencing, on a doctrinal problem the criminal courts never touch at all. Rapanos v. United States produced its own fractured opinion in 2006 over how far the Clean Water Act's "waters of the United States" reaches: a four-justice plurality by Justice Scalia limited the term to relatively permanent, continuously flowing bodies of water, while Justice Kennedy, concurring only in the judgment, proposed a broader significant-nexus test asking whether a wetland significantly affects the integrity of a navigable water nearby. Kennedy's test is not a narrower version of Scalia's. It is a different test, built on different premises, that happens to reach the same result on Rapanos's own facts. Circuits that took up the question split several ways rather than converging on one, and at least one ducked the choice while still deciding the case in front of it: in United States v. Cundiff, the Sixth Circuit found jurisdiction proper under both Scalia's test and Kennedy's test at once, working through Kennedy's significant-nexus test first, and said outright that it would "leave ultimate resolution of the Marks-meets-Rapanos debate to a future case that turns on which test in fact controls." This one did not. The Eleventh Circuit went furthest, holding expressly that Kennedy's significant-nexus opinion is the sole controlling rule, and refusing even to ask whether the plurality's narrower test would have made a jury instruction error harmless. The Seventh and Ninth Circuits reached a similar result without committing nearly as far: the Seventh held Kennedy's test governs the case in front of it while flagging, without resolving, the awkward possibility of a case where Kennedy's own vote would deny jurisdiction only to be outvoted eight to one; the Ninth held Kennedy's test controls "in almost all cases," a hedge neither circuit read as foreclosing the plurality's test as an alternative basis where the facts called for it. The First Circuit reasoned differently again, in United States v. Johnson, borrowing language Justice Stevens had used in his own Rapanos dissent: since all four dissenters said they would find jurisdiction whenever either the plurality's test or Kennedy's test was satisfied, a court could count their votes either way, so, as Johnson put it, quoting Stevens, the government "may elect to prove jurisdiction under either test." The Eighth Circuit reached the same disjunctive result as its own holding, and the Third later joined both of them expressly. A wetland that fails Kennedy's broader test but happens to meet Scalia's narrower, more literal one, a comparatively rare fact pattern but not an empty one, is covered by federal law in the circuits running the disjunctive rule and is not covered in the circuits that treat Kennedy's opinion as controlling on its own. For the better part of two decades, whether a given wetland fell under the Clean Water Act at all depended on which of at least three different circuit-level answers to the same unresolved Marks question happened to govern, assuming the circuit in question had settled on an answer at all.
Sackett v. Environmental Protection Agency, 598 U.S. 651 (2023), finally resolved which of the two Rapanos tests survives, nearly seventeen years later, and the resolution came by two different margins depending on which half of the question is asked. All nine justices agreed that Kennedy's significant-nexus test is finished; Justice Kavanaugh's concurrence, joined by Justices Sotomayor, Kagan and Jackson, said so in as many words, endorsing the majority's "decision not to adopt the significant nexus test" even while rejecting what the majority put in its place. That replacement, a continuous-surface-connection test drawn from Scalia's old plurality, commanded only five votes, Justice Alito's opinion joined by the Chief Justice and Justices Thomas, Gorsuch and Barrett; the same four justices who agreed Kennedy's test should die argued the majority's own substitute departs from the statute's text, from decades of agency practice and from the Court's own precedents. Sackett is an opinion about what the Clean Water Act means. It is not an opinion about what Marks required lower courts to do with Rapanos in the meantime, and it does not say those courts had been applying the wrong controlling opinion all along. It simply made the argument about which Rapanos opinion controlled irrelevant going forward, the same way Hughes made the argument about which Freeman opinion controlled irrelevant to future defendants.
Notice what that means for nearly seventeen years in between. A property owner sued under the significant-nexus test during that window was judged by a standard that, on Sackett's own account of what the statute actually means, was never the correct one, applied by courts that had every reason to think Marks required them to use it anyway. Sackett does not revisit any of those judgments. It only makes sure nobody gets judged that way again, which is a different kind of fix than saying the old rule was wrong to apply when it was applied.
Twice now, in two fields with nothing else in common, a fragmented opinion split lower courts for the better part of a decade or two, and the eventual fix arrived as a fresh majority answering the underlying substantive question rather than as a ruling on what Marks itself demands. A rule built to tell courts how to read a fragmented decision has instead trained the Court, whenever a case squarely tests that rule, to make the fragmented decision irrelevant rather than to say what reading it required. The King v. Palmer question Hughes left open is, as far as the Court's own opinions show, still open.