When the Supreme Court says a precedent has been “abandoned,” the real work has already been done.
That was the story of Lemon v. Kurtzman, decided in 1971. For decades, the court invoked, revised, sidestepped, and criticized Lemon’s approach to the establishment clause of the First Amendment. The formal reports still contained the case. Lawyers still cited it. Lower courts still had to account for it. Yet the governing center of gravity had shifted. When the court later said in 2022’s Kennedy v. Bremerton School District that it had “long ago abandoned” Lemon, the statement confirmed a process already visible across years of doctrine.
That process is not unique to Lemon. Supreme Court precedents often lose authority gradually. But how this authority wanes varies dramatically. A case may be narrowed, distinguished, criticized, or confined to its facts. It may also be overruled sub silentio. A later line of decisions may begin doing the doctrinal work once assigned to the original case. The court may refuse to extend a precedent while leaving the original holding formally intact. And in some instances, it may condemn a decision as a constitutional failure without overruling it in the ordinary sense.
This article takes a look at the many ways in which precedents meet their demise.
Breaking down the data
Several major precedents now sit somewhere between full vitality and formal overruling. 1971’s Bivens v. Six Unknown Named Agents, allowing persons to bring certain suit against federal officers, remains on the books, but the court has made any extensions of it exceedingly difficult to recognize. Humphrey’s Executor v. United States, for nearly a century a central case on independent agencies and the president’s power to remove the heads of them, is under renewed pressure, especially with the soon-to-be-decided case of Trump v. Slaughter that directly questions this precedent. Employment Division v. Smith continues to govern free exercise claims under the First Amendment, yet the court’s more recent doctrines have narrowed the field of laws that receive such treatment.
Other precedents have already completed different versions of the same arc: Abood v. Detroit Board of Education, which upheld that public sector non-union employees could be required to pay union fees, was overruled in Janus v. AFSCME, while Korematsu v. United States, which upheld the internment of Japanese-American citizens during World War II, was repudiated as having “no place in law under the Constitution.”
Citation data offer one way to trace these developments. The numbers cannot answer by themselves whether a precedent remains healthy. They can, however, identify patterns: how much recognition a case receives, how often courts cite it negatively, how frequently the Supreme Court itself participates in that negative treatment, and whether citations are rising or falling in recent years.
The decisions have large disparities in their overall cite counts. Bivens has more than 38,000 citations in the data, far more than the other cases examined here. It also has the largest raw number of negative citations. That figure makes sense given the field. Bivens operates across criminal procedure, federal law enforcement, prison, detention, immigration, and constitutional-remedies litigation. A precedent that governs recurring claims against federal officers will naturally appear in far more opinions than a case about, say, public-sector union fees.
The more revealing figure is negative-treatment share, or negative citations as a percentage of total citations. On that measure, Employment Division v. Smith has the highest rate in this group, at 13.4%. Lemon follows at 9.2%. Humphrey’s Executor and Korematsu sit near 5%. Bivens, despite its large raw number of negative citations, has a negative-treatment share of only 1.7%.
The Supreme Court’s role adds another layer. Before Janus, Abood had 58 negative citations, seven of them from the Supreme Court. That is a notable share for a precedent that had not yet been overruled. It suggests an erosion pattern driven less by lower-court resistance than by signals from the court itself. In other words, lower courts continued to work within Abood while the Supreme Court increasingly treated the precedent as vulnerable.
Recent citation patterns also help separate dormant precedents from live ones. Humphrey’s Executor has only 527 total citations in the data, but 118 of them occurred from 2022 through 2026. That recent concentration reflects the renewed salience of presidential removal and independent-agency doctrine. Smith presents a different picture: a larger citation universe, the highest negative-treatment share in this set, and continued litigation over its meaning.
The cases fall into several recurring patterns. Abood shows erosion followed by express overruling. Lemon shows functional abandonment through replacement of the governing test. Bivens shows confinement without formal overruling. Korematsu shows historical repudiation. Humphrey’s Executor and Smith show live questioning. Together, they illustrate how much precedential change occurs before the court uses the word “overruled.”
Erosion followed by overruling: Abood
The clearest example in the group is Abood v. Detroit Board of Education. Decided in 1977, Abood permitted public-sector unions to collect agency fees from nonmembers for collective-bargaining expenses, while barring forced support for ideological or political activities. For decades, that compromise structured First Amendment challenges to public-sector union funding.
By the time the court decided Janus v. AFSCME in 2018, however, Abood had already been weakened. The data show 825 total citations before Janus, 58 negative citations, and seven Supreme Court negative citations. That produces a negative-treatment share of about 7%. The Supreme Court’s role was striking here: Roughly 12% of the negative citations came from the court itself.
That makes Abood a top-down erosion case. Lower courts continued to apply the precedent. The Supreme Court, meanwhile, increasingly framed the precedent as an anomaly in First Amendment law. Finally, Janus completed things. As Justice Samuel Alito, writing for the majority, declared in that case:
Abood was poorly reasoned. It has led to practical problems and abuse. It is inconsistent with other First Amendment cases and has been undermined by more recent decisions. Developments since Abood was handed down have shed new light on the issue of agency fees, and no reliance interests on the part of public-sector unions are sufficient to justify the perpetuation of the free speech violations that Abood has countenanced for the past 41 years. Abood is therefore overruled.
The Janus language matters because it shows the final stage of an erosion sequence. The court did not merely announce a new rule. It described the earlier precedent as poorly reasoned, inconsistent with later First Amendment doctrine, practically problematic, and insufficiently supported by reliance interests. That is the classic pathway from criticism to overruling: a precedent first becomes unstable in the court’s own opinions, then becomes untenable when the votes align to reconsider it directly.
Functional abandonment: Lemon
Lemon v. Kurtzman, which created a three-part test to determine if a government action violated the establishment clause, followed a different path. The court did not dispatch it with the same clean formal overruling language that Janus used for Abood. Yet its decline is hard to miss.
For years, the court treated the Lemon test as both familiar and unstable. It remained part of establishment clause vocabulary, especially in the lower courts. But the Supreme Court repeatedly declined to make it the exclusive test. In some settings, particularly cases involving public religious symbols, legislative prayer, monuments, and longstanding practices, the court turned instead to history, tradition, and context instead.
The progression can be seen in the court’s own language. First, two years after being decided, Lemon became a “helpful signpost.” Then, following decades of criticism inside and outside the court, it was declared “not useful” in certain cases. Finally, in Kennedy v. Bremerton School District, the court said it had “long ago abandoned” the Lemon test.
That sequence captures functional abandonment. Lemon did not disappear because courts stopped citing it. It disappeared because the court stopped treating the test as the organizing framework for establishment clause analysis before finally bringing it to an end.
Confinement without overruling: Bivens
Bivens v. Six Unknown Named Agents presents a different problem for citation-based analysis. By raw numbers, it appears to be the largest erosion story in the dataset. It has more than 37,650 citations, including 635 negative citations. No other case in this group comes close.
The scale is misleading unless its context is taken into account. Bivens sits at the center of a large and recurring litigation universe. It is cited in cases where individuals seek to sue federal officers, and thus touches on federal law enforcement, prison conditions, immigration detention, national security, criminal investigations, and constitutional remedies. That makes its citation profile very different from a narrower precedent such as Abood, which governed a specific First Amendment question involving public-sector union fees.
The modern court has repeatedly treated new Bivens claims as disfavored, especially when the claim arises outside of its original Fourth Amendment search-and-seizure context.
This pattern is confinement. The court leaves the original case in place while shrinking the circumstances in which it can do new work. Hesitation becomes the governing rule. Congress’ failure to create a damages action, the presence of alternative remedial schemes, institutional competence, and other concerns all become reasons for courts to stop short.
The court has not erased the original case. It has made the original case difficult to generalize.
Historical repudiation: Korematsu
Korematsu v. United States follows another path. For decades, Korematsu, in which the court allowed for the internment of people of Japanese descent during World War II, stood less as a working rule than as a warning. Courts, scholars, and lawyers invoked it as an example of judicial failure during wartime. Its authority eroded through constitutional memory as much as through ordinary doctrinal development. By the time the Supreme Court addressed the case in Trump v. Hawaii, its practical legitimacy had already collapsed. In Chief Justice John Roberts’ words:
The dissent’s reference to Korematsu … affords this Court the opportunity to make express what is already obvious: Korematsu was gravely wrong the day it was decided, has been overruled in the court of history, and—to be clear—“has no place in law under the Constitution.”
The court’s statement had the language of overruling but the posture of repudiation. Korematsu was not supplying the operative rule in Trump v. Hawaii in the way Abood supplied the rule later rejected in Janus. The court instead used the occasion to declare that Korematsu had long since lost any legitimate place in constitutional law.
That makes Korematsu a category of its own. Some precedents fade because later decisions narrow their holdings. Others are displaced by new tests. Korematsu was condemned as a constitutional symbol. Its death was not mainly a product of doctrinal mechanics, but of historical judgment.
Live testing: Humphrey’s Executor and Employment Division v. Smith
The last category involves precedents that have not yet reached a final endpoint – but may soon.
Humphrey’s Executor v. United States, which could be formally overturned this term in Trump v. Slaughter, is the lower-volume example. It has 527 total citations in the data, 28 negative citations, and two Supreme Court negative citations, for a negative-treatment share of about 5.3%. Standing alone, those numbers would not place it among the most visibly eroded precedents in the group. The recent trend tells a different story, however. From 2022 through 2026, Humphrey’s Executor received 118 citations, about 22.4% of its total citation universe.
That concentration reflects renewed attention to presidential removal. For decades, Humphrey’s Executor stood for Congress’ ability to provide removal protection for members of certain independent agencies. More recent cases have recentered the analysis around Article II and presidential control. 2020’s Seila Law LLC v. CFPB, for example, did not overrule Humphrey’s Executor, but it narrowed the precedent’s orbit by describing removal protection as the exception rather than the rule:
But text, first principles, the First Congress’s decision in 1789, [and prior precedent] all establish that the President’s removal power is the rule, not the exception. While we do not revisit Humphrey’s Executor or any other precedent today, we decline to elevate it into a freestanding invitation for Congress to impose additional restrictions on the President’s removal authority.
The importance of that language lies in its supposed restraint. The court said it was not revisiting Humphrey’s Executor, but it also refused to let the case become a general permission slip for removal restrictions. That is how erosion can operate before overruling. A precedent remains in place, but later doctrine reduces its domain and signals that further extension is unlikely.
Employment Division v. Smith is different. It has a much larger citation universe and the highest negative-treatment share in this set. The data show 2,901 total citations, 388 negative citations, and 11 Supreme Court negative citations. Its negative-treatment share is 13.4%.
Smith also sits in a field where the court has repeatedly adjusted the surrounding doctrine. In that case, the court held that neutral and generally applicable laws ordinarily do not violate the free exercise clause merely because they burden religious exercise. The modern pressure on Smith comes through the meaning of “neutral” and “generally applicable.” If more laws affecting religious groups are characterized as underinclusive, discretionary, targeted, or non-neutral, fewer cases fall within Smith’s deferential rule. From Fulton v. Philadelphia, decided in 2021:
Not only is it difficult to square Smith’s interpretation with the terms of the Free Exercise Clause, the absence of any language referring to equal treatment is striking. If equal treatment was the objective, why didn’t Congress say that? And since it would have been simple to cast the Free Exercise Clause in equal-treatment terms, why would the state legislators who voted for ratification have read the Clause that way?
That makes Smith a live erosion case. The court can alter its practical reach without immediately overruling it. Each decision that narrows neutrality or general applicability leaves the formal rule standing while reducing the set of laws that can benefit from it. The current posture of free exercise litigation therefore resembles the pre-overruling stage in some respects, even though the court has not taken the full overruling question in the same direct way.
Five ways precedent loses force
Viewed together, the cases show several recurring paths of precedential decline.
To be clear: The categories are not necessarily rigid. A precedent may move from one to another. A case that begins as a confinement story may eventually be overruled. A precedent that appears merely weakened may become the subject of direct reconsideration once a new case presents the right vehicle. The categories instead help clarify what the data are showing.
Conclusion
Formal overruling remains the clearest sign that a precedent has died. But it is rarely the only relevant event. Often, by the time the court overrules a case, abandons a test, or declares a precedent illegitimate, the surrounding doctrine has been signaling the change for years.
The six precedents here capture different versions of that process. Abood was criticized and narrowed before it was overruled. Lemon was displaced before it was declared abandoned. Bivens survives, but with little appetite for extension. Korematsu was repudiated as a constitutional failure. Humphrey’s Executor and Smith remain live, but each is being tested through the logic of newer cases.
Citation data can make those patterns more visible. They show when a case remains active, when negative treatment accumulates, when the Supreme Court is participating in that erosion, and when a precedent suddenly receives renewed attention. For litigants and lower courts, it is key to understand whether the Supreme Court still treats the precedent as a rule to be applied, or as a problem to be managed.
