Kentucky permits "hazardous position" workers, e.g., policemen, to receive normal retirement benefits after working either 20 years or 5 years and attaining age 55 and pays "disability retirement" benefits to workers meeting specified requirements. Kentucky's "Plan" calculates normal retirement benefits based on actual years of service. The Plan calculates disability benefits by adding to an employee's actual years of service the number of years that the employee would have had to continue working in order to become eligible for normal retirement benefits, adding no more than the number of years the employee had previously worked. Charles Lickteig, who continued working after becoming eligible for retirement at age 55, became disabled and retired at age 61. He filed an age discrimination complaint with respondent (EEOC) after the Plan based his pension on his actual years of service without imputing any additional years. The EEOC filed suit against Kentucky and others (collectively Kentucky), arguing that the Plan failed to impute years solely because Lickteig became disabled after age 55. The District Court granted Kentucky summary judgment, holding that the EEOC could not establish age discrimination, but the Sixth Circuit ultimately reversed on the ground that the Plan violated the Age Discrimination in Employment Act of 1967 (ADEA).

Held: Kentucky's system does not discriminate against workers who become disabled after becoming eligible for retirement based on age. Pp. 4-14.

(a) The ADEA forbids an employer to "discriminate against any individual with respect to his compensation, terms, conditions, or privileges of employment, because of such individual's age." 29 U. S. C. sec.623(a)(1) (emphasis added). A plaintiff claiming age-related "disparate treatment" (i.e., intentional discrimination) must prove that age "actually motivated the employer's decision." Hazen Paper Co. v. Biggins, 507 U. S. 604 (emphasis added). In Hazen Paper, the Court found that, without evidence of intent, a dismissal based on pension status was not a dismissal "because ... of age," id., at 611-612, noting that, though pension status depended upon years of service, and years of service typically go hand in hand with age, the two concepts are "analytically distinct," id., at 611. And the dismissal at issue there, if based purely on pension status, would not embody the evils prompting the ADEA: It was not based on a "prohibited stereotype" of older workers, did not produce any "attendant stigma" to those workers, and was not "the result of an inaccurate and denigrating generalization about age." Id., at 612. However, the Court noted that discrimination based on pension status could violate the ADEA if pension status was a "proxy for age." Id., at 613. Pp. 4-6.

(b) Applying Hazen Paper, the circumstances here, taken together, show that the differences in treatment in this particular instance were not "actually motivated" by age. (1) Age and pension status remain "analytically distinct" concepts. (2) Here, several background circumstances eliminate the possibility that pension status serves as a "proxy for age." Rather than an individual employment decision, at issue here are complex systemwide rules involving not wages, but pensions--a benefit the ADEA treats somewhat more flexibly and leniently in respect to age. Further, Congress has otherwise approved programs, such as Social Security Disability Insurance, that calculate disability benefits using a formula that expressly takes account of age. (3) The disparity here has a clear non-age-related rationale. The Plan's disability rules track Kentucky's "normal retirement" rules by imputing only those additional years of service needed to bring the disabled worker's total to 20 or to the number of years that the individual would have worked had he worked to age 55. Thus, the disability rules' purpose is to treat a disabled worker as though he had become disabled after, rather than before, he had become eligible for "normal retirement" benefits. Age factors into the disability calculation only because the normal retirement rules themselves permissibly consider age. The Plan simply seeks to treat disabled employees as if they had worked until the point at which they would be eligible for a normal pension. Thus, the disparity turns upon pension eligibility and nothing more. (4) Although the Plan placed an older worker at a disadvantage here, in other cases, the rules can work to the advantage of older workers, who may get a bigger boost of imputed years than younger workers. (5) Kentucky's system does not rely on the sorts of stereotypical assumptions, e.g., the work capacity of "older" workers relative to "younger" workers, that the ADEA sought to eradicate. The Plan's "assumptions" that no disabled worker would have continued to work beyond the point at which he was both disabled and pension eligible do not involve age-related stereotypes, but apply equally to all workers regardless of age. (6) The nature of the Plan's eligibility requirements means that, unless Kentucky were severely to cut the benefits to disabled workers who are not yet pension eligible, it would have to increase the benefits available to disabled, pension-eligible workers, while lacking any clear criteria for determining how many extra years to impute for those already 55 or older. The difficulty of finding a remedy that can both correct the disparity and achieve the Plan's legitimate objective--providing each disabled worker with a sufficient retirement benefit--further suggests that this objective, not age, "actually motivated" the Plan.

The Court's opinion in no way unsettles the rule that a statute or policy that facially discriminates based on age suffices to show disparate treatment under the ADEA. The Court is dealing with the quite special case of differential treatment based on pension status, where pension status--with the explicit blessing of the ADEA--itself turns, in part, on age. Further, the rule for dealing with this sort of case is clear: Where an employer adopts a pension plan that includes age as a factor, and that employer then treats employees differently based on pension status, a plaintiff, to state a claim under the ADEA, must adduce sufficient evidence to show that the differential treatment was "actually motivated" by age, not pension status. Pp. 6-11.

(c) The Federal Government's additional arguments are rejected. Since Hazen Paper provides the relevant precedent here, an ADEA amendment made in light of Public Employees Retirement System of Ohio v. Betts, 492 U. S. 158 , is beside the point. And a contrary interpretation contained in an EEOC regulation and its compliance manual does not lead to a different conclusion. Pp. 11-13.

467 F. 3d 571, reversed.

Breyer, J., delivered the opinion of the Court, in which Roberts, C. J., and Stevens, Souter, and Thomas, JJ., joined. Kennedy, J., filed a dissenting opinion, in which Scalia, Ginsburg, and Alito, JJ., joined.

Petitioner Metropolitan Life Insurance Company (MetLife) is an administrator and the insurer of Sears, Roebuck & Company's long-term disability insurance plan, which is governed by the Employee Retirement Income Security Act of 1974 (ERISA). The plan gives MetLife (as administrator) discretionary authority to determine the validity of an employee's benefits claim and provides that MetLife (as insurer) will pay the claims. Respondent Wanda Glenn, a Sears employee, was granted an initial 24 months of benefits under the plan following a diagnosis of a heart disorder. MetLife encouraged her to apply for, and she began receiving, Social Security disability benefits based on an agency determination that she could do no work. But when MetLife itself had to determine whether she could work, in order to establish eligibility for extended plan benefits, it found her capable of doing sedentary work and denied her the benefits. Glenn sought federal-court review under ERISA, see 29 U. S. C. sec.1132(a)(1)(B), but the District Court denied relief. In reversing, the Sixth Circuit used a deferential standard of review and considered it a conflict of interest that MetLife both determined an employee's eligibility for benefits and paid the benefits out of its own pocket. Based on a combination of this conflict and other circumstances, it set aside MetLife's benefits denial.

Held:

1. Firestone Tire & Rubber Co. v. Bruch, 489 U. S. 101 , sets out four principles as to the appropriate standard of judicial review under sec.1132(a)(1)(B): (1) A court should be "guided by principles of trust law," analogizing a plan administrator to a trustee and considering a benefit determination a fiduciary act, id., at 111-113; (2) trust law principles require de novo review unless a benefits plan provides otherwise, id., at 115; (3) where the plan so provides, by granting "the administrator or fiduciary discretionary authority to determine eligibility," "a deferential standard of review [is] appropriate," id., at 111, 115; and (4) if the administrator or fiduciary having discretion "is operatingunder a conflict of interest, that conflict must be weighed as a 'facto[r] in determining whether there is an abuse of discretion,' " id., at 115. Pp. 3-5.

2. A plan administrator's dual role of both evaluating and paying benefits claims creates the kind of conflict of interest referred to in Firestone. That conclusion is clear where it is the employer itself that both funds the plan and evaluates the claim, but a conflict also exists where, as here, the plan administrator is an insurance company. For one thing, the employer's own conflict may extend to its selection of an insurance company to administer its plan. For another, ERISA imposes higher-than-marketplace quality standards on insurers, requiring a plan administrator to "discharge [its] duties" in respect to discretionary claims processing "solely in the interests of the [plan's] participants and beneficiaries," 29 U. S. C. sec.1104(a)(1); underscoring the particular importance of accurate claims processing by insisting that administrators "provide a 'full and fair review' of claim denials," Firestone, supra, at 113; and supplementing marketplace and regulatory controls with judicial review of individual claim denials, see sec.1132(a)(1)(B). Finally, a legal rule that treats insurers and employers alike in respect to the existence of a conflict can nonetheless take account of different circumstances by treating the circumstances as diminishing the conflict's significance or severity in individual cases. Pp. 5-8.

3. The significance of the conflict of interest factor will depend upon the circumstances of the particular case. Firestone's "weighed as a 'factor' " language, 489 U. S., at 115, does not imply a change in the standard of review, say, from deferential to de novo. Nor should this Court overturn Firestone by adopting a rule that could bring about near universal de novo review of most ERISA plan claims denials. And it is not necessary or desirable for courts to create special burden-of-proof rules, or other special procedural or evidentiary rules, focused narrowly upon the evaluator/payor conflict. Firestone means what the word "factor" implies, namely, that judges reviewing a benefit denial's lawfulness may take account of several different considerations, conflict of interest being one. This kind of review is no stranger to the judicial system. Both trust law and administrative law ask judges to determine lawfulness by taking account of several different, often case-specific, factors, reaching a result by weighing all together. Any one factor will act as a tiebreaker when the others are closely balanced. Here, the Sixth Circuit gave the conflict some weight, but focused more heavily on other factors: that MetLife had encouraged Glenn to argue to the Social Security Administration that she could do no work, received the bulk of the benefits of her success in doing so (being entitled to receive an offset from her retroactive Social Security award), and then ignored the agency's finding in concluding that she could do sedentary work; and that MetLife had emphasized one medical report favoring denial of benefits, had deemphasized other reports suggesting a contrary conclusion, and had failed to provide its independent vocational and medical experts with all of the relevant evidence. These serious concerns, taken together with some degree of conflicting interests on MetLife's part, led the court to set aside MetLife's discretionary decision. There is nothing improper in the way this review was conducted. Finally, the Firestone standard's elucidation does not consist of detailed instructions, because there "are no talismanic words that can avoid the process of judgment." Universal Camera Corp. v. NLRB, 340 U. S. 474 . Pp. 8-13.

461 F. 3d 660, affirmed.

Breyer, J., delivered the opinion of the Court, in which Stevens, Souter, Ginsburg, and Alito, JJ., joined, and in which Roberts, C. J., joined as to all but Part IV. Roberts, C. J., filed an opinion concurring in part and concurring in the judgment. Kennedy, J., filed an opinion concurring in part and dissenting in part. Scalia, J., filed a dissenting opinion, in which Thomas, J., joined.

When the National Government ordered its contractor, respondent Knolls, to reduce its work force, Knolls had its managers score their subordinates on "performance," "flexibility," and "critical skills"; these scores, along with points for years of service, were used to determine who was laid off. Of the 31 employees let go, 30 were at least 40 years old. Petitioners (Meacham, for short) were among those laid off, and they filed this suit asserting, inter alia, a disparate-impact claim under the Age Discrimination in Employment Act of 1967 (ADEA), 29 U. S. C. sec.621 et seq. To show such an impact, Meacham relied on a statistical expert's testimony that results so skewed according to age could rarely occur by chance; and that the scores for "flexibility" and "criticality," over which managers had the most discretionary judgment, had the firmest statistical ties to the outcomes. The jury found for Meacham on the disparate-impact claim, and the Second Circuit initially affirmed. This Court vacated the judgment and remanded in light of its intervening decision in Smith v. City of Jackson, 544 U. S. 228 . The Second Circuit then held for Knolls, finding its prior ruling untenable because it had applied a "business necessity" standard rather than a "reasonableness" test in assessing the employer's reliance on factors other than age in the layoff decisions, and because Meacham had not carried the burden of persuasion as to the reasonableness of Knolls's non-age factors.

Held: An employer defending a disparate-impact claim under the ADEA bears both the burden of production and the burden of persuasion for the "reasonable factors other than age" (RFOA) affirmative defense under sec.623(f)(1). Pp. 5-17.

(a) The ADEA's text and structure indicate that the RFOA exemption creates an affirmative defense, for which the burden of persuasion falls on the employer. The RFOA exemption is listed alongside one for bona fide occupational qualifications (BFOQ), which the Court has recognized to be an affirmative defense: "It shall not be unlawful for an employer ... to take any action otherwise prohibited under subsections (a), (b), (c), or (e) ... where age is a [BFOQ] reasonably necessary to the normal operation of the particular business, or where the differentiation is based on [RFOA] ... ." sec.623(f)(1). Given that the statute lays out its exemptions in a provision separate from the general prohibitions in sec.sec.623(a)-(c), (e), and expressly refers to the prohibited conduct as such, it is no surprise that this Court has spoken of both the BFOQ and RFOA as being among the ADEA's "five affirmative defenses," Trans World Airlines, Inc. v. Thurston, 469 U. S. 111 . This reading follows the familiar principle that "[w]hen a proviso ... carves an exception out of the body of a statute or contract those who set up such exception must prove it," Javierre v. Central Altagracia, 217 U. S. 502 . As this longstanding convention is part of the backdrop against which the Congress writes laws, the Court respects it unless there is compelling reason to think that Congress put the burden of persuasion on the other side. See Schaffer v. Weast, 546 U. S. 49 . The Court has given this principle particular weight in enforcing the Fair Labor Standards Act of 1968, Corning Glass Works v. Brennan, 417 U. S. 188 ; and it has also recognized that "the ADEA [is] enforced in accordance with the 'powers, remedies, and procedures' of the FLSA," Lorillard v. Pons, 434 U. S. 575 . Nothing in sec.623(f)(1) suggests that Congress meant it to march out of step with either the general or specifically FLSAdefault rules placing the burden of proving an exemption on the party claiming it. Any further doubt would be dispelled by the natural implication of the "otherwise prohibited" language prefacing the BFOQ and RFOA defenses. Pp. 5-9.

(b) Knolls argues that because the RFOA clause bars liability where action is taken for reasons "other than age," it should be read as mere elaboration on an element of liability. But City of Jackson confirmed that sec.623(a)(2)'s prohibition extends to practices with a disparate impact, inferring this result in part from the presence of the RFOA provision. 544 U. S., at 239, 243. And City of Jackson made it clear that action based on a "factor other than age" is the very premise for disparate-impact liability, not a negation of it or a defense to it. Thus, it is assumed that a non-age factor was at work in such a case, and the focus of the RFOA defense is on whether the factor relied on was "reasonable." Pp. 10-11.

(c) The business necessity test has no place in ADEA disparate-impact cases; applying both that test and the RFOA defense would entail a wasteful and confusing structure of proof. The absence of a business necessity enquiry does not diminish, however, the reasons already given for reading the RFOA as an affirmative defense. City of Jackson cannot be read as implying that the burden of proving any business-related defense falls on the plaintiff, for it confirmed that the BFOQ is an affirmative defense, see 544 U. S., at 233, n. 3. Moreover, in referring to "Wards Cove's interpretation of identical language [in Title VII]," City of Jackson could not have had the RFOA clause in mind, for Title VII has no like-worded defense. And as Wards Cove did not purport to construe any Title VII defenses, only an over-reading of City of Jackson would find in it an assumption that Wards Cove has anything to say about statutory defenses in the ADEA. Pp. 12-15.

(d) City of Jackson confirmed that an ADEA disparate-impact plaintiff must " ' "isolat[e] and identif[y] the specific employment practices that are allegedly responsible for any observed statistical disparities." ' " 544 U. S., at 241. This is not a trivial burden, and it ought to allay some of the concern that recognizing an employer's burden of persuasion on an RFOA defense will encourage strike suits or nudge plaintiffs with marginal cases into court; but in the end, such concerns have to be directed at Congress, which set the balance by both creating the RFOA exemption and writing it in the orthodox format of an affirmative defense. Pp. 15-17.

461 F. 3d 134, vacated and remanded.

Souter, J., delivered the opinion of the Court, in which Roberts, C. J., and Stevens, Kennedy, Ginsburg, and Alito, JJ., joined, and in which Thomas, J., joined as to Parts I and II-A. Scalia, J., filed an opinion concurring in the judgment. Thomas, J., filed an opinion concurring in part and dissenting in part. Breyer, J., took no part in the consideration or decision of the case.

After Indiana charged respondent Edwards with attempted murder and other crimes for a shooting during his attempt to steal a pair of shoes, his mental condition became the subject of three competency proceedings and two self-representation requests, mostly before the same trial judge. Referring to the lengthy record of psychiatric reports, the trial court noted that Edwards suffered from schizophrenia and concluded that, although it appeared he was competent to stand trial, he was not competent to defend himself at trial. The court therefore denied Edwards' self-representation request. He was represented by appointed counsel at trial and convicted on two counts. Indiana's intermediate appellate court ordered a new trial, agreeing with Edwards that the trial court's refusal to permit him to represent himself deprived him of his constitutional right of self-representation under the Sixth Amendment and Faretta v. California, 422 U. S. 806 . Although finding that the record provided substantial support for the trial court's ruling, the Indiana Supreme Court nonetheless affirmed the intermediate appellate court on the ground that Faretta and Godinez v. Moran, 509 U. S. 389 , required the State to allow Edwards to represent himself.

Held: The Constitution does not forbid States from insisting upon representation by counsel for those competent enough to stand trial but who suffer from severe mental illness to the point where they are not competent to conduct trial proceedings by themselves. Pp. 4-13.

(a) This Court's precedents frame the question presented, but they do not answer it. Dusky v. United States, 362 U. S. 402 , and Drope v. Missouri, 420 U. S. 162 , set forth the Constitution's "mental competence" standard forbidding the trial of an individual lacking a rational and factual understanding of the proceedings and sufficient ability to consult with his lawyerwith a reasonable degree of rational understanding. But those cases did not consider the issue presented here, namely, the relation of that "mental competence" standard to the self-representation right. Similarly the Court's foundational "self-representation" case, Faretta, supra--which held that the Sixth and Fourteenth Amendments include a "constitutional right to proceed without counsel when" a criminal defendant "voluntarily and intelligently elects to do so," 422 U. S., at 807--does not answer the question as to the scope of the self-representation right. Finally, although Godinez, supra, presents a question closer to the one at issue in that it focused upon a borderline-competent defendant who had asked a state trial court to permit him to represent himself and to change his pleas from not guilty to guilty, Godinez provides no answer here because that defendant's ability to conduct a defense at trial was expressly not at issue in that case, see 509 U. S., at 399-400, and because the case's constitutional holding that a State may permit a gray-area defendant to represent himself does not tell a State whether it may deny such a defendant the right to represent himself at his trial. Pp. 4-8.

(b) Several considerations taken together lead the Court to conclude that the Constitution permits a State to limit a defendant's self-representation right by insisting upon trial counsel when the defendant lacks the mental competency to conduct his trial defense unless represented. First, the Court's precedent, while not answering the question, points slightly in that direction. By setting forth a standard that focuses directly upon a defendant's ability to consult with his lawyer, Dusky and Drope assume representation by counsel and emphasize counsel's importance, thus suggesting (though not holding) that choosing to forgo trial counsel presents a very different set of circumstances than the mental competency determination for a defendant to stand trial. Also, Faretta rested its self-representation conclusion in part on pre-existing state cases that are consistent with, and at least two of which expressly adopt, a competency limitation on the self-representation right. See 422 U. S., at 813, and n. 9. Second, the nature of mental illness--which is not a unitary concept, but varies in degree, can vary over time, and interferes with an individual's functioning at different times in different ways--cautions against using a single competency standard to decide both whether a defendant who is represented can proceed to trial and whether a defendant who goes to trial must be permitted to represent himself. Third, a self-representation right at trial will not "affirm the dignity" of a defendant who lacks the mental capacity to conduct his defense without the assistance of counsel, see McKaskle v. Wiggins, 465 U. S. 168 , and may undercut the most basic of the Constitution's criminal law objectives, providing a fair trial. The trial judge--particularly one such as the judge in this case, who presided over one of Edwards' competency hearings and his two trials--will often prove best able to make more fine-tuned mental capacity decisions, tailored to the particular defendant's individualized circumstances. Pp. 8-12.

(c) Indiana's proposed standard, which would deny a criminal defendant the right to represent himself at trial if he cannot communicate coherently with the court or a jury, is rejected because this Court is uncertain as to how that standard would work in practice. The Court also declines Indiana's request to overrule Faretta because today's opinion may well remedy the unfair trial concerns previously leveled against the case. Pp. 12-13.

866 N. E. 2d 252, vacated and remanded.

Breyer, J., delivered the opinion of the Court, in which Roberts, C. J., and Stevens, Kennedy, Souter, Ginsburg, and Alito, JJ., joined. Scalia, J., filed a dissenting opinion, in which Thomas, J., joined.

Organizations whose members do business with California sued to enjoin enforcement of "Assembly Bill 1889" (AB 1889), which, among other things, prohibits employers that receive state grants or more than $10,000 in state program funds per year from using the funds "to assist, promote, or deter union organizing." Cal. Govt. Code Ann. sec.sec.16645.2(a), 16645.7(a). The District Court granted the plaintiffs partial summary judgment, holding that the National Labor Relations Act (NLRA) pre-empts sec.sec.16645.2 and 16645.7 because they regulate employer speech about union organizing under circumstances in which Congress intended free debate. The Ninth Circuit reversed, concluding that Congress did not intend to preclude States from imposing such restrictions on the use of their own funds.

Held: Sections 16645.2 and 16645.7 are pre-empted by the NLRA. Pp. 4-16.

(a) The NLRA contains no express pre-emption provision, but this Court has held pre-emption necessary to implement federal labor policy where, inter alia, Congress intended particular conduct to "be unregulated because left 'to be controlled by the free play of economic forces.' " Machinists v. Wisconsin Employment Relations Comm'n, 427 U. S. 132 . Pp. 4-5.

(b) Sections 16645.2 and 16645.7 are pre-empted under Machinists because they regulate within "a zone protected and reserved for market freedom." Building & Constr. Trades Council v. Associated Builders & Contractors of Mass./R. I., Inc., 507 U. S. 218 . In 1947, the Taft-Hartley Act amended the NLRA by, among other things, adding sec.8(c), which protects from National Labor Relations Board (NLRB) regulation noncoercive speech by both unions and employers about labor organizing. The section both responded to prior NLRB rulings that employers' attempts to persuade employees not to organize amounted to coercion prohibited as an unfair labor practice by the previous version of sec.8 and manifested a "congressional intent to encourage free debate on issues dividing labor and management." Linn v. Plant Guard Workers, 383 U. S. 53 . Congress' express protection of free debate forcefully buttresses the pre-emption analysis in this case. California's policy judgment that partisan employer speech necessarily interferes with an employee's choice about union representation is the same policy judgment that Congress renounced when it amended the NLRA to preclude regulation of noncoercive speech as an unfair labor practice. To the extent sec.sec.16645.2 and 16645.7 actually further AB 1889's express goal, they are unequivocally pre-empted. Pp. 5-8.

(c) The Ninth Circuit's reasons for concluding that Machinists did not pre-empt sec.sec.16645.2 and 16645.7--(1) that AB 1889's spending restrictions apply only to the use of state funds, not to their receipt; (2) that Congress did not leave the zone of activity free from all regulation, in that the NLRB still regulates employer speech on the eve of union elections; and (3) that California modeled AB 1889 on federal statutes, e.g., the Workforce Investment Act--are not persuasive. Pp. 8-16.

463 F. 3d 1076, reversed and remanded.

Stevens, J., delivered the opinion of the Court, in which Roberts, C. J., and Scalia, Kennedy, Souter, Thomas, and Alito, JJ., joined. Breyer, J., filed a dissenting opinion, in which Ginsburg, J., joined.