The Hague Convention on the Civil Aspects of International Child Abduction requires the judicial or administrative authority of a Contracting State to order a child returned to the child’s country of habitual residence if the authority finds that the child has been wrongfully removed to or retained in the Contracting State. The authority “is not bound to order the return of the child,” however, if the authority finds that return would expose the child to a “grave risk” of “physical or psychological harm or otherwise place the child in an intolerable situation.” The International Child Abduction Remedies Act (ICARA) implements the Convention in the United States, granting federal and state courts jurisdiction over Convention actions and directing those courts to decide cases in accordance with the Convention.

  Petitioner Narkis Golan, a United States citizen, married respondent Isacco Saada, an Italian citizen, in Italy, where they had a son, B. A. S., in 2016. In 2018, Golan flew with B. A. S. to the United States to attend a wedding and, instead of returning to Italy, moved into a domestic violence shelter with B. A. S. Saada thereafter timely filed a petition with the U. S. District Court for the Eastern District of New York, seeking an order returning B. A. S. to Italy pursuant to the Hague Convention. The District Court concluded that B. A. S. would face a grave risk of harm if returned to Italy, given evidence that Saada had abused Golan and that being exposed to this abuse harmfully affected B. A. S. The court, however, ordered B. A. S.’ return to Italy, applying Second Circuit precedent obligating it to “examine the full range of options that might make possible the safe return of a child” and concluding that ameliorative measures could reduce the risk to B. A. S. sufficiently to require his return. The Second Circuit vacated the return order, finding the District Court’s ameliorative measures insufficient. Because the record did not support concluding that no sufficient ameliorative measures existed, the Second Circuit remanded for the District Court to consider whether such measures, in fact, existed. After an examination over nine months, the District Court identified new ameliorative measures and again ordered B. A. S.’ return. The Second Circuit affirmed.

Held: A court is not categorically required to examine all possible ameliorative measures before denying a Hague Convention petition for return of a child to a foreign country once the court has found that return would expose the child to a grave risk of harm. Pp. 8–16.

 (a) “The interpretation of a treaty, like the interpretation of a statute, begins with its text.” Abbott v. Abbott, 560 U. S. 1, 10 (internal quotation marks omitted). When “a child has been wrongfully removed or retained” from his country of habitual residence, Article 12 of the Hague Convention generally requires the deciding authority (here, a district court) to “order the return of the child.” T. I. A. S. No. 11670, S. Treaty Doc. No. 99–11, p. 9. But Article 13(b) of the Convention leaves a court with the discretion to grant or deny return, providing that a court “is not bound to order the return of the child” if it finds that the party opposing return has established that return would expose the child to a “grave risk” of physical or psychological harm. Id., at 10. Nothing in the Convention’s text either forbids or requires consideration of ameliorative measures in exercising this discretion. Pp. 8–11.

  (1) Saada’s primary argument is that determining whether a grave risk of harm exists necessarily requires considering whether any ameliorative measures are available. The two questions, however, are separate. A court may find it appropriate to consider both questions at once, but this does not mean that the Convention imposes a categorical requirement on a court to consider any or all ameliorative measures before denying return based on a grave-risk determination. Pp. 9–10.

  (2) The discretion to courts under the Convention and ICARA includes the discretion to determine whether to consider ameliorative measures that could ensure the child’s safe return. The Second Circuit’s contrary rule—which imposes an atextual, categorical requirement that courts consider all possible ameliorative measures in exercising discretion under the Convention, regardless of whether such consideration is consistent with the Convention’s objectives—“in practice, rewrite[s] the treaty,” Lozano v. Montoya Alvarez, 572 U. S. 1, 17. Pp. 10–11.

 (b) A district court’s consideration of ameliorative measures must be guided by the legal principles and other requirements set forth in the Convention and ICARA. The Second Circuit’s rule improperly elevated return above the Convention’s other objectives. The Convention does not pursue return exclusively or at all costs. Courts must remain conscious of all the Convention’s objectives and requirements, which constrain courts’ discretion to consider ameliorative measures. First, the Convention explicitly recognizes that any consideration of ameliorative measures must prioritize the child’s physical and psychological safety. Second, consideration of ameliorative measures should abide by the Convention’s requirement that courts addressing return petitions do not usurp the role of the court that will adjudicate the underlying custody dispute. Third, any consideration of ameliorative measures must accord with the Convention’s requirement that courts “act expeditiously in proceedings for the return of children.” A court therefore reasonably may decline to consider ameliorative measures that have not been raised by the parties, are unworkable, draw the court into determinations properly resolved in custodial proceedings, or risk overly prolonging return proceedings. Pp. 11–15.

 (c) In this case, the District Court made a finding of grave risk, but never had the opportunity to inquire whether to order or deny return under the correct legal standard. Accordingly, it is appropriate to allow the District Court to apply the proper legal standard in the first instance, see Monasky v. Taglieri, 589 U. S. ___, ___. The District Court should determine whether the measures considered are adequate to order return in light of the District Court’s factual findings concerning the risk to B. A. S., bearing in mind that the Convention sets as a primary goal the safety of the child. Pp. 15–16.

833 Fed. Appx. 829, vacated and remanded.

 Sotomayor, J., delivered the opinion for a unanimous Court.

The Medicare statute lays out a formula that the Department of Health and Human Services must employ annually to set reimbursement rates for certain outpatient prescription drugs provided by hospitals to Medicare patients. 42 U. S. C. §1395l(t)(14)(A)(iii). That formula affords HHS two options. Option 1 applies if HHS has conducted a survey of hospitals’ acquisition costs for each covered outpatient drug. Under this option, the agency may set reimbursement rates based on the hospitals’ “average acquisition cost” for each drug, and may “vary” the reimbursement rates “by hospital group.” §1395l(t)(14)(A)(iii)(I). Absent a survey, option 2 applies, and HHS must set reimbursement rates based on “the average price” charged by manufacturers for the drug as “calculated and adjusted by the Secretary.” §1395l(t)(14)(A)(iii)(II). Option 2 does not authorize HHS to vary reimbursement rates for different hospital groups. From the time these provisions took effect in 2006 until 2018, HHS did not conduct surveys of hospitals’ acquisition costs, relied on option 2, set the reimbursement rates at about 106 percent, and did not vary those rates by hospital group. For 2018, HHS again did not conduct a survey. But this time it issued a final rule establishing separate reimbursement rates for hospitals that serve low-income or rural populations through the 340B program and all other hospitals. For 2019, HHS set reimbursement rates the same way.

 The American Hospital Association and other interested parties challenged the 2018 and 2019 reimbursement rates in federal court. In response, HHS first contended that various statutory provisions precluded judicial review of those rates. The agency also argued that it could vary the reimbursement rates by hospital group under its option 2 authority to “adjust” the price-based reimbursement rates. The District Court rejected HHS’s argument that the statute precluded judicial review, concluded that HHS had acted outside its statutory authority, and remanded the case to HHS to consider an appropriate remedy. The D. C. Circuit, however, reversed. The court ruled that the statute did not preclude judicial review, and upheld HHS’s reduced reimbursement rates for 340B hospitals.

Held:

 1. The statute does not preclude judicial review of HHS’s reimbursement rates. Judicial review of final agency action is traditionally available unless “a statute’s language or structure” precludes it, Mach Mining, LLC v. EEOC, 575 U. S. 480, 486, and this Court has long recognized a “strong presumption” in its favor, Weyerhaeuser Co. v. United States Fish and Wildlife Serv., 586 U. S. ___, ___. Here, no provision in the Medicare statute precludes judicial review of the 2018 and 2019 reimbursement rates. HHS cites two nearby provisions that preclude review of the general payment methodology that HHS employs to set rates for other Medicare outpatient services. See §§1395l(t)(12)(A), (C). But HHS sets rates for outpatient prescription drugs using a different payment methodology. HHS also argues that other statutory requirements would make allowing judicial review of the 2018 and 2019 reimbursement rates impractical. Regardless, such arguments cannot override the text of the statute and the traditional presumption in favor of judicial review of administrative action. Pp. 7–9.

 2. Absent a survey of hospitals’ acquisition costs, HHS may not vary the reimbursement rates only for 340B hospitals; HHS’s 2018 and 2019 reimbursement rates for 340B hospitals were therefore unlawful. The text and structure of the statute make this a straightforward case. Because HHS did not conduct a survey of hospitals’ acquisition costs, HHS acted unlawfully by reducing the reimbursement rates for 340B hospitals. HHS maintains that even when it does not conduct a survey, the agency still may “adjus[t]” the average price “as necessary.” §1395l(t)(14)(A)(iii)(II). But HHS’s power to increase or decrease the price is distinct from its power to set different rates for different groups of hospitals. Moreover, HHS’s interpretation would make little sense given the statute’s overall structure. Under HHS’s interpretation, the agency would never need to conduct a survey of acquisition costs if it could proceed under option 2 and then do everything under option 2 that it could do under option 1. That not only would render irrelevant the survey prerequisite for varying reimbursement rates by hospital group, but also would render largely irrelevant the provision of the statute that precisely details the requirements for surveys of hospitals’ acquisition costs. See §1395l(t)(14)(D). Finally, HHS’s argument that Congress could not have intended for the agency to “overpay” 340B hospitals for prescription drugs ignores the fact that Congress, when enacting the statute, was well aware that 340B hospitals paid less for covered prescription drugs. It may be that the reimbursement payments were intended to offset the considerable costs of providing healthcare to the uninsured and underinsured in low-income and rural communities. Regardless, this Court is not the forum to resolve that policy debate. Pp. 9–14.

967 F. 3d 818, reversed and remanded.

 Kavanaugh, J., delivered the opinion for a unanimous Court.

The question for decision is whether the Federal Arbitration Act, 9 U. S. C. §1 et seq., preempts a rule of California law that invalidates contractual waivers of the right to assert representative claims under California’s Labor Code Private Attorneys General Act of 2004, Cal. Lab. Code §2698 et seq. PAGA enlists employees as private attorneys general to enforce California labor law. By its terms, PAGA authorizes any “aggrieved employee” to initiate an action against a former employer “on behalf of himself or herself and other current or former employees” to obtain civil penalties that previously could have been recovered only by the State in an enforcement action brought by California’s Labor and Workforce Development Agency (LWDA). California precedent holds that a PAGA suit is a “ ‘representative action’ ” in which the employee plaintiff sues as an “ ‘agent or proxy’ ” of the State. Iskanian v. CLS Transp. Los Angeles, LLC, 59 Cal. 4th 348, 380. California precedent also interprets the statute to contain what is effectively a rule of claim joinder—allowing a party to unite multiple claims against an opposing party in a single action. An employee with PAGA standing may “seek any civil penalties the state can, including penalties for violations involving employees other than the PAGA litigant herself.” ZB, N. A. v. Superior Court, 8 Cal. 5th 175, 185.

  Respondent Angie Moriana filed a PAGA action against her former employer Viking River Cruises, alleging a California Labor Code violation. She also asserted a wide array of other violations allegedly sustained by other Viking employees. Moriana’s employment contract with Viking contained a mandatory arbitration agreement. Important here, that agreement contained both a “Class Action Waiver”—providing that the parties could not bring any dispute as a class, collective, or representative action under PAGA—and a severability clause— specifying that if the waiver was found invalid, such a dispute would presumptively be litigated in court. Under the severability clause, any “portion” of the waiver that remained valid would be “enforced in arbitration.” Viking moved to compel arbitration of Moriana’s individual PAGA claim and to dismiss her other PAGA claims. Applying California’s Iskanian precedent, the California courts denied that motion, holding that categorical waivers of PAGA standing are contrary to California policy and that PAGA claims cannot be split into arbitrable “individual” claims and nonarbitrable “representative” claims. This Court granted certiorari to decide whether the FAA preempts the California rule.

Held: The FAA preempts the rule of Iskanian insofar as it precludes division of PAGA actions into individual and non-individual claims through an agreement to arbitrate. Pp. 7–21.

 (a) Based on the principle that “[a]rbitration is strictly ‘a matter of consent,’ ” Granite Rock Co. v. Teamsters, 561 U. S. 287, 299, this Court has held that “a party may not be compelled under the FAA to submit to class arbitration unless there is a contractual basis for concluding that the party agreed to do so,” Stolt-Nielsen S. A. v. AnimalFeeds Int’l Corp., 559 U. S. 662, 684. Because class-action arbitration mandates procedural changes that are inconsistent with the individualized and informal mode of bilateral arbitration contemplated by the FAA, see AT&T Mobility LLC v. Concepcion, 563 U. S. 333, 347, class procedures cannot be imposed by state law without presenting unwilling parties with an unacceptable choice between being compelled to arbitrate using such procedures and forgoing arbitration all together.

 Viking contends that the Court’s FAA precedents require enforcement of contractual provisions waiving the right to bring PAGA actions because PAGA creates a form of class or collective proceeding. If this is correct, Iskanian’s prohibition on PAGA waivers presents parties with an impermissible choice: Either arbitrate disputes using a form of class procedures, or do not arbitrate at all. Moriana maintains that any conflict between Iskanian and the FAA is illusory because PAGA creates nothing more than a substantive cause of action.

 This Court disagrees with both characterizations of the statute. Moriana’s premise that PAGA creates a unitary private cause of action is irreconcilable with the structure of the statute and the ordinary legal meaning of the word “claim.” A PAGA action asserting multiple violations under California’s Labor Code affecting a range of different employees does not constitute “a single claim” in even the broadest possible sense. Viking’s position, on the other hand, elides important structural differences between PAGA actions and class actions. A class-action plaintiff can raise a multitude of claims because he or she represents a multitude of absent individuals; a PAGA plaintiff, by contrast, represents a single principal, the LWDA, that has a multitude of claims. As a result, PAGA suits exhibit virtually none of the procedural characteristics of class actions.

 This Court’s FAA precedents treat bilateral arbitration as the prototype of the individualized and informal form of arbitration protected from undue state interference by the FAA. See, e.g., Epic Systems Corp. v. Lewis, 584 U. S. ___, ___. Viking posits that a proceeding is “bilateral” only if it involves two and only two parties and “is conducted by and on behalf of the individual named parties only.” Wal-Mart Stores, Inc. v. Dukes, 564 U. S. 338, 348. Thus, Iskanian’s prohibition on PAGA waivers is inconsistent with the FAA because PAGA creates an intrinsically representational form of action and Iskanian requires parties either to arbitrate in that format or forgo arbitration altogether.

 This Court disagrees. Nothing in the FAA establishes a categorical rule mandating enforcement of waivers of standing to assert claims on behalf of absent principals. Non-class representative actions in which a single agent litigates on behalf of a single principal necessarily deviate from the strict ideal of bilateral dispute resolution posited by Viking, but this Court has never held that the FAA imposes a duty on States to render all forms of representative standing waivable by contract or that such suits deviate from the norm of bilateral arbitration. Unlike procedures distinctive to multiparty litigation, single-principal, single-agent representative actions are “bilateral” in two registers: They involve the rights of only the absent real party in interest and the defendant, and litigation need only be conducted by the agent-plaintiff and the defendant. Nothing in this Court’s precedent suggests that in enacting the FAA, Congress intended to require States to reshape their agency law governing who can assert claims on behalf of whom to ensure that parties will never have to arbitrate disputes in a proceeding that deviates from bilateral arbitration in the strictest sense. Pp. 7–17.

 (b) PAGA’s built-in mechanism of claim joinder is in conflict with the FAA. Iskanian’s prohibition on contractual division of PAGA actions into constituent claims unduly circumscribes the freedom of parties to determine “the issues subject to arbitration” and “the rules by which they will arbitrate,” Lamps Plus, Inc. v. Varela, 587 U. S. ____, ____, and does so in a way that violates the fundamental principle that “arbitration is a matter of consent,” Stolt-Nielsen, 559 U. S., at 684. For that reason, state law cannot condition the enforceability of an agreement to arbitrate on the availability of a procedural mechanism that would permit a party to expand the scope of the anticipated arbitration by introducing claims that the parties did not jointly agree to arbitrate. A state rule imposing an expansive rule of joinder in the arbitral context would defeat the ability of parties to control which claims are subject to arbitration by permitting parties to superadd new claims to the proceeding, regardless of whether the agreement committed those claims to arbitration. When made compulsory by way of Iskanian, PAGA’s joinder rule functions in exactly this way. The effect is to coerce parties into withholding PAGA claims from arbitration. Iskanian’s indivisibility rule effectively coerces parties to opt for a judicial forum rather than “forgo[ing] the procedural rigor and appellate review of the courts to realize the benefits of private dispute resolution.” Stolt-Nielsen, 559 U. S., at 685. Pp. 17–19.

 (c) Under this Courts holding, Iskanian’s prohibition on wholesale waivers of PAGA claims is not preempted by the FAA. But Iskanian’s rule that PAGA actions cannot be divided into individual and non-individual claims is preempted, so Viking was entitled to compel arbitration of Moriana’s individual claim. PAGA provides no mechanism to enable a court to adjudicate non-individual PAGA claims once an individual claim has been committed to a separate proceeding. And under PAGA’s standing requirement, a plaintiff has standing to maintain non-individual PAGA claims in an action only by virtue of also maintaining an individual claim in that action. As a result, Moriana would lack statutory standing to maintain her non-individual claims in court, and the correct course was to dismiss her remaining claims. Pp. 20–21.

Reversed and remanded.

 Alito, J., delivered the opinion of the Court, in which Breyer, Sotomayor, Kagan, and Gorsuch, JJ., joined, in which Roberts, C. J., joined as to Parts I and III, and in which Kavanaugh and Barrett, JJ., joined as to Part III. Sotomayor, J., filed a concurring opinion. Barrett, J., filed an opinion concurring in part and concurring in the judgment, in which Kavanaugh, J., joined, and in which Roberts, C. J, joined as to all but the footnote. Thomas, J., filed a dissenting opinion.

Chief Justice Roberts, with whom Justice Thomas, Justice Alito, and Justice Gorsuch join, concurring.

 This case involves a regulation known as the Public Charge Rule, promulgated by the Department of Homeland Security in 2019. See 84 Fed. Reg. 41292 (2019). The Rule set out the test the Department planned to use to determine whether an applicant for admission into the country or adjustment to lawful permanent resident status is “likely at any time to become a public charge,” which would make him ineligible. 8 U. S. C. §1182(a)(4)(A). Several parties filed lawsuits arguing that the Rule was unlawful because it defined “public charge” too broadly.

 We granted certiorari in this case not to address the merits of that argument, but to decide whether the petitioners—13 States which support the Rule—should have been permitted to intervene in this litigation to defend the Rule’s legality in the Court of Appeals. Petitioners argue that the answer is yes, in light of the Government’s actions.

 When this and other suits challenging the Rule were first brought in 2019, the Government defended it. And when multiple lower courts, including the District Court here, found the Rule unlawful, the Government appealed those decisions. After a change in administrations, though, the Government reversed course and opted to voluntarily dis miss those appeals, leaving in place the relief already entered.

 A new administration is of course as a general matter entitled to do that. But the Government then took a further step. It seized upon one of the now-consent judgments against it—a final judgment vacating the Rule nationwide, issued in a different litigation—and leveraged it as a basis to immediately repeal the Rule, without using notice-and-comment procedures. 86 Fed. Reg. 14221 (2021) (“Because this rule simply implements the district court’s vacatur of the August 2019 rule . . . DHS is not required to provide notice and comment.”). This allowed the Government to circumvent the usual and important requirement, under the Administrative Procedure Act, that a regulation originally promulgated using notice and comment (as the Public Charge Rule was) may only be repealed through notice and comment, 5 U. S. C. §551(5); see Perez v. Mortgage Bankers Assn., 575 U. S. 92, 101 (2015). As part of this tactic of “rulemaking-by-collective-acquiescence,” City and County of San Francisco v. United States Citizenship and Immigration Servs., 992 F. 3d 742, 744 (CA9 2021) (VanDyke, J., dissenting), the Government successfully opposed efforts by other interested parties—including petitioners here—to intervene in order to carry on the defense of the Rule, including possibly before this Court.

 These maneuvers raise a host of important questions. The most fundamental is whether the Government’s actions, all told, comport with the principles of administrative law. But bound up in that inquiry are a great many issues beyond the question of appellate intervention on which we granted certiorari, among them standing; mootness; vacatur under United States v. Munsingwear, Inc., 340 U. S. 36 (1950); the scope of injunctive relief in an APA action; whether, contrary to what “[t]he government has long argued,” the APA “authorize[s] district courts to vacate regulations or other agency actions on a nationwide basis,” Brief for Federal Respondents 5, n. 3; how the APA’s procedural requirements apply in this unusual circumstance, cf. §551(5); FCC v. Fox Television Stations, Inc., 556 U. S. 502, 515 (2009); and more.

 It has become clear that this mare’s nest could stand in the way of our reaching the question presented on which we granted certiorari, or at the very least, complicate our resolution of that question. I therefore concur in the Court’s dismissal of the writ of certiorari as improvidently granted. But that resolution should not be taken as reflective of a view on any of the foregoing issues, or on the appropriate resolution of other litigation, pending or future, related to the 2019 Public Charge Rule, its repeal, or its replacement by a new rule. See Cook County v. Mayorkas, 340 F. R. D. 35 (ND Ill. 2021), appeal pending, No. 21–2561 (CA7); 87 Fed. Reg. 10571 (2022) (new proposed rule that would “implement a different policy than the 2019 Final Rule”).

When petitioner Kevin George joined the Marine Corps in 1975, he did not disclose his history of schizophrenic episodes, and a medical examination noted no mental disorders. After George suffered an episode during training, the Marines medically discharged him. George then applied to the Department of Veterans Affairs under 38 U. S. C. §1110 for veterans’ disability benefits based on his schizophrenia. A regional office of the VA denied George’s claim, and the VA’s Board of Veterans’ Appeals denied his appeal in 1977.

  In 2014, George asked the Board to revise its final decision. When the VA denies a benefits claim, that decision generally becomes “final and conclusive and may not be reviewed by any other official or by any court” after the veteran exhausts the opportunity for direct appeal. §511(a); see §7104(a). But George sought collateral review under a statutory exception allowing a veteran to seek revision of a final benefits decision at any time on grounds of “clear and unmistakable error.” §§5109A, 7111; see 38 CFR §§3.105, 20.1400–20.1411. In particular, he claimed that the Board clearly and unmistakably erred by applying a later invalidated regulation to deny his claim for benefits without holding the VA to its burden of proof to rebut the statutory presumption that he was in sound condition when he entered service.

  The Board denied George’s claim for collateral relief, and the Veterans Court affirmed. The Federal Circuit also affirmed, concluding that the application of a later invalidated regulation does not fall into the narrow category of “clear and unmistakable error” permitting revision of a final decision under 38 U. S. C. §§5109A and 7111.

Held: The invalidation of a VA regulation after a veteran’s benefits decision becomes final cannot support a claim for collateral relief based on clear and unmistakable error. Pp. 5–12.

 (a) This case turns on the meaning of the 1997 statute subjecting a final veterans’ benefits decision to collateral review on grounds of “clear and unmistakable error.” 111 Stat. 2271 (38 U. S. C. §§5109A, 7111). No statute defines the term “clear and unmistakable error,” but the modifiers “clear” and “unmistakable” as well as the statutory structure suggest a narrow category. A robust regulatory backdrop fills in the details. Where Congress employs a term of art “ ‘ “obviously transplanted from another legal source,” ’ it ‘ “brings the old soil with it.” ’ ” Taggart v. Lorenzen, 587 U. S. ___, ___. That principle applies here. The Court agrees with the Federal Circuit that Congress “codif[ied] and adopt[ed] the [clear-and-unmistakable-error] doctrine as it had developed under” decades of prior agency practice. Cook v. Principi, 318 F. 3d 1334, 1344 (en banc). That history reveals that this category of error does not encompass a subsequent “change in law . . . or a change in interpretation of law.” 38 CFR §3.105 (Cum. Supp. 1963). And the invalidation of a prior regulation constitutes a “change in interpretation of law” under historical agency practice. Defined by this regulatory history, the statutory term “clear and unmistakable error” does not encompass a claim like George’s. Pp. 5–8.

 (b) In response, George argues that the VA has distorted the history of agency practice that the 1997 statute codified. But across a century of review for clear and unmistakable error, George can muster only one uncertain outlier case sustaining a claim that arguably resembles his, which does not move the mountain of contrary regulatory authority. He alternatively argues that the VA is wrong to call a later decision invalidating a regulation a “change in interpretation of law.” But that is a perfectly natural use of language. George tries to bolster his position by invoking cases explaining that a judicial decision states what the statute “always meant,” Rivers v. Roadway Express, Inc., 511 U. S. 298, 313, n. 12, and an unauthorized regulation is a “ ‘nullity,’ ” Dixon v. United States, 381 U. S. 68, 74. But those general principles do not disturb the conclusion that the Board’s application of a then-binding regulation is not the kind of “clear and unmistakable error” for which collateral relief is available under §§5109A and 7111. And that longstanding VA approach is consistent with the general rule that the new interpretation of a statute can only retroactively affect decisions still open on direct review.

 George also leans on what he describes as the plain meaning of the words “clear and unmistakable error.” But as he concedes elsewhere, the real question is not what might be called clear and unmistakable error in the abstract, but what the prevailing understanding of this term of art was when Congress codified it. The fact that Congress did not expressly enact the specific regulatory principle barring collateral relief for subsequent changes in interpretation does not mean that the principle did not carry over to the statute. Statutory “silence” on the details of prior regulatory practice indicates that Congress “left the matter where it was pre-[codification].” Kucana v. Holder, 558 U. S. 233, 250. Pp. 8–12.

991 F. 3d 1227, affirmed.

 Barrett, J., delivered the opinion of the Court, in which Roberts, C. J., and Thomas, Alito, Kagan, and Kavanaugh, JJ., joined. Sotomayor, J., filed a dissenting opinion. Gorsuch, J., filed a dissenting opinion, in which Breyer, J., joined, and in which Sotomayor, J., joined as to all but Part II–C.

This case represents the latest conflict between Texas gaming officials and the Ysleta del Sur Pueblo Indian Tribe. In 1968, Congress recognized the Ysleta del Sur Pueblo as an Indian tribe and assigned its trust responsibilities for the Tribe to Texas. 82 Stat. 93. In 1983, Texas renounced its trust responsibilities as inconsistent with the State’s Constitution. The State also expressed opposition to any new federal trust legislation that did not permit the State to apply its own gaming laws on tribal lands. Congress restored the Tribe’s federal trust status in 1987 when it adopted the Ysleta del Sur and Alabama and Coushatta Indian Tribes of Texas Restoration Act. 101 Stat. 666. The Restoration Act also “prohibited” as a matter of federal law “[a]ll gaming activities which are prohibited by the laws of the State of Texas.” Id., at 668. Shortly thereafter, Congress adopted its own comprehensive Indian gaming legislation: the Indian Gaming Regulatory Act (IGRA). IGRA established rules for separate classes of games. As relevant here, IGRA permitted Tribes to offer so-called class II games—like bingo—in States that “permi[t] such gaming for any purpose by any person, organization or entity.” 25 U. S. C. §2710(b)(1)(A). IGRA allowed Tribes to offer class III games—like blackjack and baccarat—but only pursuant to negotiated tribal/state compacts. §2703(8).

  Pursuant to IGRA, the Tribe sought to negotiate a compact with Texas to offer class III games. Texas refused, arguing that the Restoration Act displaced IGRA and required the Tribe to follow all of the State’s gaming laws on tribal lands. In subsequent federal litigation, the District Court held that Texas violated IGRA by failing to negotiate in good faith. The Fifth Circuit reversed, holding that the Resto ration Act’s directions superseded IGRA’s and guaranteed that the entirety of “Texas’ gaming laws and regulations” would “operate as surrogate federal law on the Tribe’s reservation.” 36 F. 3d 1325, 1326, 1334 (Ysleta I). In 2016, the Tribe began to offer bingo, including “electronic bingo” machines, on the view that IGRA treats bingo as a class II game for which no state permission is required so long as the State permits the game to be played on some terms by some persons. The State then sought to shut down all of the Tribe’s bingo operations. Bound by Ysleta I, the District Court sided with Texas and enjoined the Tribe’s bingo operations, but the court stayed the injunction pending appeal. The Fifth Circuit reaffirmed Ysleta I and held that the Tribe’s bingo operations were impermissible because they did not conform to Texas’s bingo regulations.

Held: The Restoration Act bans as a matter of federal law on tribal lands only those gaming activities also banned in Texas. Pp. 8–20.

 (a) Section 107 of the Restoration Act directly addresses gaming on the lands of the Ysleta del Sur Pueblo. It provides in subsection (a) that “gaming activities which are prohibited by [Texas law] are hereby prohibited on the reservation and on lands of the tribe.” Subsection (b) insists that the statute does not grant Texas “civil or criminal regulatory jurisdiction” with respect to matters covered by §107. The State reads the Act as effectively subjecting the Tribe to the entire body of Texas gaming laws and regulations. The Tribe, however, understands the Act to bar it from offering only those gaming activities the State fully prohibits, and that if Texas merely regulates bingo, the Tribe may also offer that game subject only to federal-law, not state-law, limitations.

 The language of §107—particularly its dichotomy between prohibition and regulation—presents Texas with a problem. Texas concedes that its laws do not “forbid,” “prevent,” “effectively stop,” or “make impossible” bingo operations in the State. Webster’s Third International Dictionary 1813 (defining “prohibit”). Instead, the State admits that it allows the game “according to rule[s]” that “fix the time,” place, and manner in which it may be conducted. Id., at 1913 (defining “regulate”). From this alone, Texas’s bingo laws appear to fall on the regulatory rather than prohibitory side of the line. In response, Texas describes its laws as “prohibiting” bingo unless the State’s regulations are followed and insists that it is merely seeking to do what subsection (a) allows.

 Texas’s understanding of the word “prohibit” would risk turning the Restoration Act’s terms into an indeterminate mess. In Texas’s view, laws regulating gaming activities become laws prohibiting gaming activities—an interpretation that violates the rule against “ascribing to one word a meaning so broad” that it assumes the same meaning as another statutory term. Gustafson v. Alloyd Co., 513 U. S. 561, 575. Indeterminacy aside, the State’s interpretation would leave subsection (b)—denying the State regulatory jurisdiction—with no work to perform. As a result, Texas’s interpretation also defies another canon of statutory construction—the rule that courts must normally seek to construe Congress’s work “so that effect is given to all provisions.” Corley v. United States, 556 U. S. 303, 314 (internal quotation marks omitted). Seeking to give subsection (b) real work to perform, Texas submits that the provision serves to deny its state courts and gaming commission “jurisdiction” to punish violations of subsection (a) by sending such disputes to federal court instead. But that interpretation only serves to render subsection (c), which grants federal courts “exclusive” jurisdiction over subsection (a) violations, a nullity. A full look at the statute’s structure suggests a set of simple and coherent commands; Texas’s competing interpretation renders individual statutory terms duplicative and leaves whole provisions without work to perform. Pp. 8–12.

 (b) Important contextual clues resolve any remaining questions. Congress passed the Restoration Act six months after this Court handed down its decision in California v. Cabazon Band of Mission Indians, 480 U. S. 202. There, the Court interpreted Public Law 280—a statute Congress had adopted in 1953 to allow a handful of States to enforce some of their criminal laws on certain tribal lands—to mean that only “prohibitory” state gaming laws could be applied on the Indian lands in question, not state “regulatory” gaming laws. The Cabazon Court held that California’s bingo laws—materially identical to Texas’s laws here—fell on the regulatory side of the ledger. This Court generally assumes that, when Congress enacts statutes, it is aware of this Court’s relevant precedents. Ryan v. Valencia Gonzales, 568 U. S. 57, 66. At the time Congress adopted the Restoration Act, Cabazon was not only a relevant precedent; it was the precedent. In Cabazon’s immediate aftermath, Congress also adopted other laws governing tribal gaming that appeared to reference and employ in different ways Cabazon’s distinction between prohibition and regulation. See, e.g., Wampanoag Tribal Council of Gay Head, Inc., Indian Claims Settlement Act of 1987, §9, 101 Stat. 709–710.

 None of this is to say that the Tribe may offer gaming on whatever terms it wishes. The Restoration Act provides that a gaming activity prohibited by Texas law is also prohibited on tribal land as a matter of federal law. Other gaming activities are subject to tribal regulation and must conform to the terms and conditions set forth in federal law, including IGRA to the extent applicable. Pp. 12–15.

 (c) The State’s remaining arguments are unavailing. Pp. 15–19.

  (1) Texas asks the Court to focus on subsection (a) of the Restoration Act, which ends with the statement that “[t]he provisions of this subsection are enacted in accordance with the tribe’s request in Tribal Resolution No. T. C.–02–86.” 101 Stat. 668–669. In that referenced resolution, the Tribe announced its opposition to Texas’s legislative efforts to have its gaming laws apply on tribal lands. At the same time, the Tribe also announced its own intention to prohibit gaming on its reservation and authorized the acceptance of federal legislation prohibiting gaming on tribal lands. Texas claims that the reference to the tribal resolution suggests the Restoration Act should be read “broadly” to allow Texas to apply its gaming regulations on tribal lands. As an initial matter, subsection (a) does not purport to incorporate that resolution into federal law—something Congress knows how to do when it wishes, see e.g., 25 U. S. C. §5396(b). In addition, Texas’s “broad” reading suffers from the same interpretative challenges already mentioned and defies Congress’s apparent adoption of Cabazon’s prohibitory/regulatory distinction. Finally, on this Court’s interpretation of the Restoration Act, Congress did legislate “in accordance with” the Tribe’s resolution by expressly granting the Tribe federal recognition and choosing not to apply Texas gaming regulations as surrogate federal law on tribal land. Pp. 15–18.

  (2) Texas appeals to public policy and argues that attempts to distinguish between prohibition and regulation are sure to prove “unworkable.” It is not, however, this Court’s place to question whether Congress adopted the wisest or most workable policy. That the Restoration Act’s prohibitory/regulatory distinction can and will generate borderline cases hardly makes it unique among federal statutes. And courts have applied the same prohibitory/regulatory framework for decades under Public Law 280. Moreover, Texas’s alternative interpretation poses its own “workability” challenges, as federal courts would be charged with enforcing the minutiae of state gaming regulations governing the conduct of permissible games. Pp. 18–19.

955 F. 3d 408, vacated and remanded.

 Gorsuch, J., delivered the opinion of the Court, in which Breyer, Sotomayor, Kagan, and Barrett, JJ., joined. Roberts, C. J., filed a dissenting opinion, in which Thomas, Alito, and Kavanaugh, JJ., joined.