Petitioner John Montenegro Cruz was found guilty of capital murder by an Arizona jury and sentenced to death. Both at trial and on direct appeal, Cruz argued that under Simmons v. South Carolina, 512 U. S. 154, he should have been allowed to inform the jury that a life sentence in Arizona would be without parole. The trial court and Arizona Supreme Court held that Arizona’s capital sentencing scheme did not trigger application of Simmons. After Cruz’s conviction became final, this Court held in Lynch v. Arizona, 578 U. S. 613 (per curiam), that it was fundamental error to conclude that Simmons “did not apply” in Arizona. Id., at 615. Cruz then sought to raise the Simmons issue again in a state postconviction petition under Arizona Rule of Criminal Procedure 32.1(g), which permits a defendant to bring a successive petition if “there has been a significant change in the law that, if applicable to the defendant’s case, would probably overturn the defendant’s judgment or sentence.” The Arizona Supreme Court denied relief after concluding that Lynch was not “a significant change in the law.”

Held: The Arizona Supreme Court’s holding that Lynch was not a significant change in the law is an exceptional case where a state-court judgment rests on such a novel and unforeseeable interpretation of a state-court procedural rule that the decision is not adequate to foreclose review of the federal claim. Pp. 7–14.

  (a) This Court does not decide a question of federal law in a case if the state-court judgment “rests on a state law ground that is independent of the federal question and adequate to support the judgment.” Coleman v. Thompson, 501 U. S. 722, 729. In this case the Court focuses on the requirement of adequacy; whether Arizona’s “state procedural ruling is adequate is itself a question of federal law” Beard v. Kindler, 558 U. S. 53, 60. A state procedural ruling that is “ ‘firmly established and regularly followed’ ” will ordinarily “be adequate to foreclose review of a federal claim.” Lee v. Kemna, 534 U. S. 362, 376. This case is an exception, however, implicating this Court’s rule that “an unforeseeable and unsupported state-court decision on a question of state procedure does not constitute an adequate ground to preclude this Court’s review of a federal question.” Bouie v. City of Columbia, 378 U. S. 347, 354.

  At issue here is the Arizona Supreme Court’s decision that Cruz’s motion for postconviction relief failed to satisfy Arizona Rule of Criminal Procedure 32.1(g) because Lynch did not result in “a significant change in the law.” That court reasoned that Lynch was not a significant change in the law because it relied on Simmons, which was clearly established law at the time of Cruz’s trial. It so held even though Lynch overruled binding Arizona precedent foreclosing Simmons relief for Arizona capital defendants, and even though the Arizona Supreme Court had previously explained that the “archetype” of a “significant change in the law” is the overruling of “previously binding case law.” State v. Shrum, 220 Ariz. 115, 118, 203 P. 3d 1175, 1178. While the court reasoned that a significant change in the application of a law is not the same as a significant change in the law itself, Arizona can point to no other Rule 32.1(g) decision supporting that distinction. This interpretation of Rule 32.1(g) is entirely new and conflicts with prior Arizona case law. The novelty arises from the way in which the Arizona Supreme Court disregarded the effect of Lynch on Arizona law. Ordinarily, Arizona courts applying Rule 32.1(g) focus on how a decision changes the law that is operative in the State. Here, however, the Arizona Supreme Court disregarded the many state precedents overruled by Lynch, focusing instead on whether Lynch had wrought a significant change in federal law. Because the Arizona Supreme Court’s interpretation is so novel and unforeseeable, it cannot constitute an adequate state procedural ground for the challenged decision.

  Arizona’s interpretation generates a catch-22 for Cruz and other similarly situated capital defendants that only serves to compound its novelty. To obtain relief under Rule 32.1(g), a defendant must establish not just a significant change in the law but also that the law in question applies retroactively under Teague v. Lane, 489 U. S. 288. Prior to the Arizona Supreme Court’s decision below, it was possible to show that Lynch both was a “significant change in the law” and satisfied retroactivity because it merely applied Simmons. On the interpretation adopted below, however, the argument that Lynch applied “settled” federal law for retroactivity purposes also implies that Lynch does not represent a “significant change in the law.” Earlier Rule 32.1(g) decisions did not generate this catch-22. Given the Court’s conclusion that the Arizona Supreme Court’s application of Rule 32.1(g) to Lynch is so novel and unfounded that it does not constitute an adequate state procedural ground, it is unnecessary for the Court to determine whether the decision below is also independent of federal law. Pp. 7–11.

  (b) Counterarguments presented in this case offer various reformulations of the argument that Lynch was not a “significant change in the law” for Rule 32.1(g) purposes, but they fail to grapple with the basic point that Lynch reversed previously binding Arizona Supreme Court precedent. The fact that Lynch was a summary reversal did not justify the Arizona Supreme Court in treating Lynch differently than other transformative decisions of this Court. Although Lynch did not change this Court’s interpretation of Simmons, it did change the operation of Simmons by Arizona courts in a way that matters for Rule 32.1(g). And it makes no difference that Lynch did not alter federal law. The analytic focus of Arizona courts applying Rule 32.1(g) has always been on the impact to Arizona law. Nor does this Court’s interpretation forestall Arizona’s ability to develop its Rule 32.1(g) jurisprudence in new contexts. That the Arizona Supreme Court had never before applied Rule 32.1(g) to a summary reversal did not present a new context in this case. Finally, no effective parallel can be drawn between Rule 32.1(g) and very different procedural rules governing federal prisoners, e.g., 28 U. S. C. §§2255(f),(h). Pp. 11–14.

251 Ariz. 203, 487 P. 3d 991, vacated and remanded.

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

Respondent Michael Hewitt filed an action against his employer, petitioner Helix Energy Solutions Group, seeking overtime pay under the Fair Labor Standards Act of 1938, which guarantees overtime pay to covered employees when they work more than 40 hours a week. From 2014 to 2017, Hewitt worked for Helix on an offshore oil rig, typically working 84 hours a week while on the vessel. Helix paid Hewitt on a daily-rate basis, with no overtime compensation. So Hewitt’s paycheck, issued every two weeks, amounted to his daily rate times the number of days he had worked in the pay period. Under that compensation scheme, Hewitt earned over $200,000 annually. Helix asserts that Hewitt was exempt from the FLSA because he qualified as “a bona fide executive.” 29 U. S. C. §213(a)(1). Under applicable regulations, an employee is considered a bona fide executive excluded from the FLSA’s protections if the employee meets three distinct tests: (1) the “salary basis” test, which requires that an employee receive a predetermined and fixed salary that does not vary with the amount of time worked; (2) the “salary level” test, which requires that preset salary to exceed a specified amount; and (3) the job “duties” test. See 84 Fed. Reg. 51230. The Secretary of Labor has implemented the bona fide executive standard through two separate and slightly different rules, one “general rule” applying to employees making less than $100,000 in annual compensation, and a different rule addressing “highly compensated employees” (HCEs) who make at least $100,000 per year. 29 CFR §§541.100, 541.601(a), (b)(1). The general rule considers employees to be executives when they are “[c]ompensated on a salary basis” (salary-basis test); “at a rate of not less than $455 per week” (salary-level test); and carry out three listed responsibilities— managing the enterprise, directing other employees, and exercising power to hire and fire (duties test). §541.100(a). The HCE rule relaxes only the duties test, while restating the other two. As litigated in this case, whether Hewitt was an executive exempt from the FLSA’s overtime pay guarantee turns solely on whether Hewitt was paid on a salary basis. The District Court agreed with Helix’s view that Hewitt was compensated on a salary basis and granted the company summary judgment. The Court of Appeals for the Fifth Circuit reversed, deciding that Hewitt was not paid on a salary basis and therefore could claim the FLSA’s protections. The court so held based on its examination of the two regulations that give content to the salary-basis test. The majority first concluded that a daily-rate employee (like Hewitt) does not fall within the main salary-basis provision of §541.602(a), which states:

  “An employee will be considered to be paid on a ‘salary basis’ . . . if the employee regularly receives each pay period on a weekly, or less frequent basis, a predetermined amount constituting all or part of the employee’s compensation, which amount is not subject to reduction because of variations in the quality or quantity of the work performed. Subject to [certain exceptions], an exempt employee must receive the full salary for any week in which the employee performs any work without regard to the number of days or hours worked.”

 Second, the court held that “daily-rate” workers can qualify as paid on a salary basis only through the “special rule” of §541.604(b), which focuses on workers whose compensation is “computed on an hourly, a daily or a shift basis.” Because Hewitt’s compensation concededly did not satisfy §604(b)’s conditions, the court concluded that Hewitt, although highly paid, was not exempt from the FLSA. Reaching the opposite conclusion, a dissenting opinion determined that Hewitt’s compensation satisfied the salary basis test of §602(a) and that §604(b) is not applicable to employees who fall within the HCE rule.

Held: Hewitt was not an executive exempt from the FLSA’s overtime pay guarantee; daily-rate workers, of whatever income level, qualify as paid on a salary basis only if the conditions set out in §541.604(b) are met. Pp. 7–20.

 (a) The critical question here is whether Hewitt was paid on a salary basis under §602(a). A worker may be paid on a salary basis under either §602(a) or §604(b). But Helix acknowledges that Hewitt’s compensation did not satisfy §604(b)’s conditions. And the Court concludes that Helix did not pay Hewitt on a salary basis as defined in §602(a), a conclusion that follows from the text and the structure of the regulations. Pp. 7–17.

  (1) The text of §602(a) excludes daily-rate workers. An employee, the regulation says, is paid on a salary basis only if he “receive[s] the full salary for any week in which [he] performs any work without regard to the number of days or hours worked.” Whenever an employee works at all in a week, he must get his “full salary for [that] week”—what §602(a)’s prior sentence calls the “predetermined amount.” That amount must be “without regard to the number of days or hours worked”—or as the prior sentence says, it is “not subject to reduction because” the employee worked less than the full week. Giving language its ordinary meaning, nothing in that description fits a daily-rate worker, who by definition is paid for each day he works and no others. Further, §602(a)’s demand that an employee receive a predetermined amount irrespective of days worked embodies the standard meaning of the word “salary.” The “concept of ‘salary’ ” is linked, “[a]s a matter of common parlance,” to “the stability and security of a regular weekly, monthly, or annual pay structure.” 15 F. 4th 289, 291. Helix responds by focusing on §602(a)’s use of the word “received,” contending that because Hewitt got his paycheck every two weeks, and that check contained pay exceeding $455 (the salary level) for any week in which he had worked, Hewitt was paid on a salary basis. But Helix offers no reason for hinging satisfaction of the salary-basis test on how often paychecks are distributed. And Helix’s interpretation of the “weekly basis” phrase is not the most natural one. A “basis” of payment typically refers to the unit or method for calculating pay, not the frequency of its distribution. And that is how neighboring regulations use the term. The “weekly basis” phrase thus works hand in hand with the rest of §602(a) to reflect the standard meaning of a “salary,” which connotes a steady and predictable stream of pay. Pp. 8–12.

  (2) The broader regulatory structure—in particular, the role of §604(b)—confirms the Court’s reading of §602(a). Section §604(b) lays out a second path for a compensation scheme to meet the salary-basis requirement. And that path is all about daily, hourly, or shift rates. An employee’s earnings, §604(b) provides, “may be computed on” those shorter bases without “violating the salary basis requirement” so long as an employer “also” provides a guarantee of weekly payment approximating what the employee usually earns. Section 604(b) thus speaks directly to when daily and hourly rates are “[ ]consistent with the salary basis concept.” 69 Fed. Reg. 22184. Reading §602(a) also to cover daily- and hourly-rate employees would subvert §604(b)’s strict conditions on when their pay counts as a “salary.” By contrast, when read as limited to weekly-rate employees, §602(a) works in tandem with §604(b), with §604(b) taking over where §602(a) leaves off.

 Helix’s argument to the contrary relies on the premise that the HCE rule operates independently of §604(b). Even if so, a daily-rate worker like Hewitt is not paid on a salary basis under the plain text of §602(a). And supposing that the HCE rule incorporates only §602(a), and not §604(b), those two provisions still must be read to complement each other because §602(a) cannot change meanings depending on whether it applies to the general rule or the HCE rule. Regardless, Helix is wrong that the HCE rule operates independently of §604(b). The HCE rule refers to the salary-basis (and salary-level) requirement in the same way that the general rule does. Compare §541.601(b)(1) (requiring “at least $455 per week paid on a salary or fee basis”) with §541.100(a)(1) (requiring payment “on a salary basis at a rate of not less than $455 per week”). And the two provisions giving content to that requirement—explaining when a person is indeed paid on a salary basis—are §602(a) and §604(b). So both those provisions apply to both the general and the HCE rule. There is a difference between the HCE and general rule; it just has nothing to do with the salary-basis requirement. That difference instead involves the duties standard, which is more flexible in the HCE rule. Pp. 13–17.

 (b) The Court’s reading of the relevant regulations properly concludes this case. Helix urges the Court to consider supposed policy consequences of that reading, but even the most formidable policy arguments cannot overcome a clear textual directive. See BP p.l.c. v. Mayor and City Council of Baltimore, 593 U. S. ___, ___. And anyway, Helix’s appeal to consequences appears less than formidable in the context of the FLSA’s regulatory scheme. Helix’s complaint about “windfalls” for high earners fails, as the HCE rule itself reflects Congress’s choice not to set a simple income level as the test for exemption. As to Helix’s cost-based objections, the whole point of the salary-basis test is to preclude employers from paying workers neither a true salary nor overtime. So too, Helix’s complaints about retroactive liability lack force because the salary-basis test is not novel, but rather traces back to the FLSA’s beginnings. Pp. 17–20.

15 F. 4th 289, affirmed.

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

Kate and David Bartenwerfer decided to remodel the house they jointly owned in San Francisco and to sell it for a profit. David took charge of the project, while Kate remained largely uninvolved. They eventually sold the house to respondent Kieran Buckley. In conjunction with the sale, Kate and David attested that they had disclosed all material facts related to the property. After the purchase, Buckley discovered several defects that the Bartenwerfers had failed to disclose. Buckley sued in California state court and won, leaving the Bartenwerfers jointly responsible for more than $200,000 in damages. Unable to pay that judgment or their other creditors, the Bartenwerfers filed for Chapter 7 bankruptcy. Buckley then filed an adversary complaint in the bankruptcy proceeding, alleging that the debt owed him on the state-court judgment was nondischargeable under the Bankruptcy Code’s exception to discharge of “any debt . . . for money . . . to the extent obtained by . . . false pretenses, a false representation, or actual fraud.” 11 U. S. C. §523(a)(2)(A). The Bankruptcy Court found that David had committed fraud and imputed his fraudulent intent to Kate because the two had formed a legal partnership to renovate and sell the property. The Bankruptcy Appellate Panel disagreed as to Kate’s culpability, holding that §523(a)(2)(A) barred her from discharging the debt only if she knew or had reason to know of David’s fraud. On remand, the Bankruptcy Court determined that Kate lacked such knowledge and could therefore discharge her debt to Buckley. The Bankruptcy Appellate Panel affirmed. The Ninth Circuit reversed in relevant part. Invoking Strang v. Bradner, 114 U. S. 555, the court held that a debtor who is liable for her partner’s fraud cannot discharge that debt in bankruptcy, regardless of her own culpability.

Held: Section 523(a)(2)(A) precludes Kate Bartenwerfer from discharging in bankruptcy a debt obtained by fraud, regardless of her own culpability. Pp. 3–12.

 (a) Kate (hereinafter, Bartenwerfer) disputes a straightforward reading of §523(a)(2)(A)’s text. Bartenwerfer argues that an ordinary English speaker would understand that “money obtained by fraud” means money obtained by the individual debtor’s fraud. This Court disagrees. The passive voice in §523(a)(2)(A) does not hide the relevant actor in plain sight, as Bartenwerfer suggests—it removes the actor altogether. Congress framed §523(a)(2)(A) to “focu[s] on an event that occurs without respect to a specific actor, and therefore without respect to any actor’s intent or culpability.” Dean v. United States, 556 U. S. 568, 572. It is true that context can confine a passive-voice sentence to a likely set of actors. See, e.g., E. I. du Pont de Nemours & Co. v. Train, 430 U. S. 112, 128–129. But the legal context relevant to §523(a)(2)(A)—the common law of fraud—has long maintained that fraud liability is not limited to the wrongdoer. Understanding §523(a)(2)(A) to reflect “agnosticism” as to the identity of the wrongdoer is consistent with the age-old rule of fraud liability.

 Bartenwerfer points out that “ ‘exceptions to discharge should be confined to those plainly expressed.’ ” Bullock v. BankChampaign, N. A., 569 U. S. 267, 275. The Court, however, has never used this principle to artificially narrow ordinary meaning, invoking it instead to stress that exceptions should not extend beyond their stated terms. See, e.g., Gleason v. Thaw, 236 U. S. 558, 559–562.

 Bartenwerfer also seeks support from §523(a)(2)(A)’s neighboring provisions in subparagraphs (B) and (C), both of which require some culpable action by the debtor herself. Bartenwerfer claims that these neighboring provisions make explicit what is unstated in (A). This argument turns on its head the rule that “ ‘[w]hen Congress includes particular language in one section . . . but omits it in another section of the same Act,’ ” the Court generally takes “the choice to be deliberate.” Badgerow v. Walters, 596 U. S. ___, ___. If there is an inference to be drawn here, the more likely one is that (A) excludes debtor culpability from consideration given that (B) and (C) expressly hinge on it. Bartenwerfer suggests it would defy credulity to think that Congress would bar debtors from discharging liability for fraud they did not personally commit under (A) while allowing debtors to discharge debt for (potentially more serious) fraudulent statements they did not personally make under (B). But the Court offered a possible answer for this disparity in Field v. Mans, 516 U. S. 59, 76–77. Whatever the rationale, it does not defy credulity to think that Congress established differing rules for (A) and (B). Pp. 3–8.

 (b) Any remaining doubt about the textual analysis is eliminated by this Court’s precedent and Congress’s response to it. In Strang v. Bradner, 114 U. S. 555, the Court held that the fraud of one partner should be imputed to the other partners, who “received and appropriated the fruits of the fraudulent conduct.” Id., at 561. The Court so held despite the fact that the relevant 19th-century discharge exception for fraud disallowed the discharge of debts “created by the fraud or embezzlement of the bankrupt.” 14 Stat. 533 (emphasis added). And when Congress next overhauled bankruptcy law, it deleted the phrase “of the bankrupt” from the discharge exception for fraud. The unmistakable implication is that Congress embraced Strang’s holding. See Ysleta Del Sur Pueblo v. Texas, 596 U. S. ___, ___. Pp. 8–10.

 (c) Finally, Bartenwerfer insists that the preclusion of faultless debtors from discharging liabilities run up by their associates is inconsistent with bankruptcy law’s “fresh start” policy. But the Bankruptcy Code is not focused on the unadulterated pursuit of the debtor’s interest, and instead seeks to balance multiple, often competing interests. Bartenwerfer’s fairness-based critiques also miss the fact that §523(a)(2)(A) does not define the scope of one’s liability for another’s fraud. Section 523(a)(2)(A) takes the debt as it finds it, so if California did not extend liability to honest partners, §523(a)(2)(A) would have no role here. And while Bartenwerfer paints a picture of liability being imposed on hapless bystanders, fraud liability generally requires a special relationship to the wrongdoer and, even then, defenses to liability are available. Pp. 10–12.

860 Fed. Appx. 544, affirmed.

 Barrett, J., filed an opinion for a unanimous Court. Sotomayor, J., filed a concurring opinion, in which Jackson, J., joined.