Arizona prisoners may raise claims of ineffective assistance of trial counsel only in state collateral proceedings, not on direct review. In petitioner Martinez’s first state collateral proceeding, his counsel did not raise such a claim. On federal habeas review with new counsel, Martinez argued that he received ineffective assistance both at trial and in his first state collateral proceeding. He also claimed that he had a constitutional right to an effective attorney in the collateral proceeding because it was the first place to raise his claim of ineffective assistance at trial. The District Court denied the petition, finding that Arizona’s preclusion rule was an adequate and independent state-law ground barring federal review, and that under Coleman v. Thompson, 501 U. S. 722, the attorney’s errors in the postconviction proceeding did not qualify as cause to excuse the procedural default. The Court of Appeals for the Ninth Circuit affirmed.

Held:

1. Where, under state law, ineffective-assistance-of-trial-counsel claims must be raised in an initial-review collateral proceeding, a procedural default will not bar a federal habeas court from hearing those claims if, in the initial-review collateral proceeding, there was no counsel or counsel in that proceeding was ineffective. Pp. 5–14.

(a) Given that the precise question here is whether ineffective assistance in an initial-review collateral proceeding on an ineffective-assistance-at-trial claim may provide cause for a procedural default in a federal habeas proceeding, this is not the case to resolve the question left open in Coleman: whether a prisoner has a constitutional right to effective counsel in initial-review collateral proceedings. However, to protect prisoners with potentially legitimate ineffective-assistance claims, it is necessary to recognize a narrow exception to Coleman’s unqualified statement that an attorney’s ignorance or inadvertence in a postconviction proceeding does not qualify as cause to excuse a procedural default, namely, that inadequate assistance of counsel at initial-review collateral proceedings may establish cause. Pp. 5–6.

(b) A federal court can hear Martinez’s ineffective-assistance claim only if he can establish cause to excuse the procedural default and prejudice from a violation of federal law. Coleman held that a postconviction attorney’s negligence “does not qualify as ‘cause,’ ” because “the attorney is the prisoner’s agent,” and “the principal bears the risk of” his agent’s negligent conduct, Maples v. Thomas, ante, at 12. However, in Coleman, counsel’s alleged error was on appeal from an initial-review collateral proceeding. Thus, his claims had been addressed by the state habeas trial court. This marks a key difference between initial-review collateral proceedings and other collateral proceedings. Here, where the initial-review collateral proceeding is the first designated proceeding for a prisoner to raise the ineffective-assistance claim, the collateral proceeding is the equivalent of a prisoner’s direct appeal as to that claim because the state habeas court decides the claim’s merits, no other court has addressed the claim, and defendants “are generally ill equipped to represent themselves” where they have no brief from counsel and no court opinion addressing their claim. Halbert v. Michigan, 545 U. S. 605. An attorney’s errors during an appeal on direct review may provide cause to excuse a procedural default; for if the attorney appointed by the State is ineffective, the prisoner has been denied fair process and the opportunity to comply with the State’s procedures and obtain an adjudication on the merits of his claim. Without adequate representation in an initial-review collateral proceeding, a prisoner will have similar difficulties vindicating a substantial ineffective-assistance-at-trial claim. The same would be true if the State did not appoint an attorney for the initial-review collateral proceeding. A prisoner’s inability to present an ineffective-assistance claim is of particular concern because the right to effective trial counsel is a bedrock principle in this Nation’s justice system.

Allowing a federal habeas court to hear a claim of ineffective assistance at trial when an attorney’s errors (or an attorney’s absence) caused a procedural default in an initial-review collateral proceeding acknowledges, as an equitable matter, that a collateral proceeding, if undertaken with no counsel or ineffective counsel, may not have been sufficient to ensure that proper consideration was given to a substantial claim. It thus follows that, when a State requires a prisoner to raise a i claim of ineffective assistance at trial in a collateral proceeding, a prisoner may establish cause for a procedural default of such claim in two circumstances: where the state courts did not appoint counsel in the initial-review collateral proceeding for an ineffective-assistance-at-trial claim; and where appointed counsel in the initial-review collateral proceeding, where that claim should have been raised, was ineffective under Strickland v. Washington, 466 U. S. 668. To overcome the default, a prisoner must also demonstrate that the underlying ineffective-assistance-at-trial claim is substantial. Most jurisdictions have procedures to ensure counsel is appointed for substantial ineffective-assistance claims. It is likely that such attorneys are qualified to perform, and do perform, according to prevailing professional norms. And where that is so, States may enforce a procedural default in federal habeas proceedings. Pp. 6–12.

(c) This limited qualification to Coleman does not implicate stare decisis concerns. Coleman’s holding remains true except as to initial-review collateral proceedings for claims of ineffective assistance at trial. The holding in this case should not put a significant strain on state resources. A State facing the question of cause for an apparent default may answer that the ineffective-assistance-of-trial-counsel claim is insubstantial. The limited circumstances recognized here also reflect the importance of the right to effective assistance at trial. Other claims may not implicate the same fundamentals of the adversary system. The Antiterrorism and Effective Death Penalty Act of 1996 does not speak to the question presented here, and thus does not bar Martinez from asserting attorney error as cause for a procedural default. Pp. 12–14.

2. Whether Martinez’s attorney in his first collateral proceeding was ineffective and whether his ineffective-assistance-at-trial claim is substantial, as well as the question of prejudice, are questions that remain open for a decision on remand. P. 15.

623 F. 3d 731, reversed and remanded.

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

The Family and Medical Leave Act of 1993 (FMLA) entitles an employee to take up to 12 work weeks of unpaid leave per year for (A) the care of a newborn son or daughter; (B) the adoption or foster-care placement of a child; (C) the care of a spouse, son, daughter, or parent with a serious medical condition; and (D) the employee’s own serious health condition when the condition interferes with the employee’s ability to perform at work. 29 U. S. C. §2612(a)(1). The FMLA also creates a private right of action for equitable relief and damages “against any employer (including a public agency) in any Federal or State court.” §2617(a)(2). For present purposes, subparagraphs (A), (B), and (C) are referred to as the family-care provisions, and subparagraph (D) as the self-care provision. In Nevada Dept. of Human Resources v. Hibbs, 538 U. S. 721−732, this Court held that Congress could subject States to suit for violations of subparagraph (C) based on evidence of family-leave policies that discriminated on the basis of sex.

Petitioner filed suit, alleging that his employer, the Maryland Court of Appeals, an instrumentality of the State, violated the FMLA by denying him self-care leave. The Federal District Court dismissed the suit on sovereign immunity grounds. The Fourth Circuit affirmed, holding that unlike the family-care provision in Hibbs, the self-care provision was not directed at an identified pattern of gender-based discrimination and was not congruent and proportional to any pattern of sex-based discrimination on the part of States.

Held: The judgment is affirmed.

626 F. 3d 187, affirmed.

Justice Kennedy, joined by The Chief Justice, Justice Thomas, and Justice Alito, concluded that suits against States under the self-care provision are barred by sovereign immunity. Pp. 3−12.

(a) Under the federal system, States, as sovereigns, are immune from damages suits, unless they waive that defense. See, e.g., Kimel v. Florida Bd. of Regents, 528 U. S. 62−73. Congress may also abrogate the States’ immunity pursuant to its powers under §5 of the Fourteenth Amendment, but it must make that intention “unmistakably clear in the language of the statute,” Hibbs, supra, at 726. It did so in the FMLA. Congress also “must tailor” legislation enacted under §5 “to remedy or prevent” “conduct transgressing the Fourteenth Amendment’s substantive provisions.” Florida Prepaid Postsecondary Ed. Expense Bd. v. College Savings Bank, 527 U. S. 627. “There must be a congruence and proportionality between the injury to be prevented or remedied and the means adopted to that end.” City of Boerne v. Flores, 521 U. S. 507. Pp. 3−5.

(b) The sex-based discrimination that supported allowing subparagraph (C) suits against States is absent with respect to the self-care provision. Petitioner’s three arguments to the contrary are unpersuasive. Pp. 5–12.

(1) Petitioner maintains that the self-care provision addresses sex discrimination and sex stereotyping. But the provision, standing alone, is not a valid abrogation of the States’ immunity from suit. At the time the FMLA was enacted, there was no evidence of such discrimination or stereotyping in sick-leave policies. Congress was concerned about the economic burdens imposed by illness-related job loss on employees and their families and about discrimination based on illness, not sex. Although the self-care provision offers some women a benefit by allowing them to take leave for pregnancy-related illnesses, the provision, as a remedy, is not congruent and proportional to any identified constitutional violations. When the FMLA was enacted, Congress had no evidence that States were excluding pregnancy-related illnesses from their leave policies. Pp. 6–7.

(2) Petitioner also argues that the self-care provision is a necessary adjunct to the family-care provision sustained in Hibbs. But his claim—that the provisions work in tandem to ensure the equal availability of total FMLA leave time to women and men despite their different leave-usage patterns―is unconvincing and does not comply with the requirements of City of Boerne. Also, there are no congressional findings of, or evidence on, how the self-care provision is necessary to the family-care provisions or how it reduces employer discrimination against women. Pp. 8–11.

(3) Finally, petitioner contends that the self-care provision helps single parents keep their jobs when they get ill. The fact that most single parents happen to be women demonstrates, at most, that the self-care provision was directed at remedying neutral leave restrictions that have a disparate effect on women. However, “[a]lthough disparate impact may be relevant evidence of . . . discrimination . . . such evidence is insufficient [to prove a constitutional violation] even where the Fourteenth Amendment subjects state action to strict scrutiny.” Board of Trustees of Univ. of Ala. v. Garrett, 531 U. S. 356. Because it is unlikely that many of the neutral leave policies affected by the self-care provision are unconstitutional, the scope of the self-care provision is out of proportion to its supposed remedial or preventive objectives. Pp. 11−12.

Justice Scalia adhered to his view that the Court should abandon the “congruence and proportionality” approach in favor of one that is properly tied to the text of §5, which grants Congress the power “to enforce, by appropriate legislation,” the other provisions of the Fourteenth Amendment. Outside the context of racial discrimination, Congress’s §5 power should be limited to the regulation of conduct that itself violates the Fourteenth Amendment and thus would not reach a State’s failure to grant self-care leave to its employees. Pp. 1−2.

Kennedy, J., announced the judgment of the Court and delivered an opinion, in which Roberts, C. J., and Thomas and Alito JJ., joined. Thomas, J., filed a concurring opinion. Scalia, J., filed an opinion concurring in the judgment. Ginsburg, J., filed a dissenting opinion, in which Breyer, J., joined, and in which Sotomayor and Kagan, JJ., joined as to all but footnote 1.

Although “laws of nature, natural phenomena, and abstract ideas” are not patentable subject matter under §101 of the Patent Act, Diamond v. Diehr, 450 U. S. 175, “an application of a law of nature . . . to a known structure or process may [deserve] patent protection,” id., at 187. But to transform an unpatentable law of nature into a patent-eligible application of such a law, a patent must do more than simply state the law of nature while adding the words “apply it.” See, e.g., Gottschalk v. Benson, 409 U. S. 63–72. It must limit its reach to a particular, inventive application of the law.

Respondent, Prometheus Laboratories, Inc. (Prometheus), is the sole and exclusive licensee of the two patents at issue, which concern the use of thiopurine drugs to treat autoimmune diseases. When ingested, the body metabolizes the drugs, producing metabolites in the bloodstream. Because patients metabolize these drugs differently, doctors have found it difficult to determine whether a particular patient’s dose is too high, risking harmful side effects, or too low, and so likely ineffective. The patent claims here set forth processes embodying researchers’ findings that identify correlations between metabolite levels and likely harm or ineffectiveness with precision. Each claim recites (1) an “administering” step—instructing a doctor to administer the drug to his patient—(2) a “determining” step—telling the doctor to measure the resulting metabolite levels in the patient’s blood—and (3) a “wherein” step—describing the metabolite concentrations above which there is a likelihood of harmful side-effects and below which it is likely that the drug dosage is ineffective, and informing the doctor that metabolite concentrations above or below these thresholds “indicate a need” to decrease or increase (respectively) the drug dosage.

Petitioners Mayo Collaborative Services and Mayo Clinic Rochester (Mayo) bought and used diagnostic tests based on Prometheus’ patents. But in 2004 Mayo announced that it intended to sell and market its own, somewhat different, diagnostic test. Prometheus sued Mayo contending that Mayo’s test infringed its patents. The District Court found that the test infringed the patents but granted summary judgment to Mayo, reasoning that the processes claimed by the patents effectively claim natural laws or natural phenomena—namely, the correlations between thiopurine metabolite levels and the toxicity and efficacy of thiopurine drugs—and therefore are not patentable. The Federal Circuit reversed, finding the processes to be patent eligible under the Circuit’s “machine or transformation test.” On remand from this Court for reconsideration in light of Bilski v. Kappos, 561 U. S. ___, which clarified that the “machine or transformation test” is not a definitive test of patent eligibility, id., at ___–___, the Federal Circuit reaffirmed its earlier conclusion.

Held: Prometheus’ process is not patent eligible. Pp. 8–24.

(a) Because the laws of nature recited by Prometheus’ patent claims—the relationships between concentrations of certain metabolites in the blood and the likelihood that a thiopurine drug dosage will prove ineffective or cause harm—are not themselves patentable, the claimed processes are not patentable unless they have additional features that provide practical assurance that the processes are genuine applications of those laws rather than drafting efforts designed to monopolize the correlations. The three additional steps in the claimed processes here are not themselves natural laws but neither are they sufficient to transform the nature of the claims. The “administering” step simply identifies a group of people who will be interested in the correlations, namely, doctors who used thiopurine drugs to treat patients suffering from autoimmune disorders. Doctors had been using these drugs for this purpose long before these patents existed. And a “prohibition against patenting abstract ideas ‘cannot be circumvented by attempting to limit the use of the formula to a particular technological environment.’ ” Bilski, supra, at ___. The “wherein” clauses simply tell a doctor about the relevant natural laws, adding, at most, a suggestion that they should consider the test results when making their treatment decisions. The “determining” step tells a doctor to measure patients’ metabolite levels, through whatever process the doctor wishes to use. Because methods for making such determinations were well known in the art, this step simply tells doctors to engage in well-understood, routine, conventional activity previously engaged in by scientists in the field. Such activity is normally not sufficient to transform an unpatentable law of nature into a patent-eligible application of such a law. Parker v. Flook, 437 U. S. 584. Finally, considering the three steps as an ordered combination adds nothing to the laws of nature that is not already present when the steps are considered separately. Pp. 8–11.

(b) A more detailed consideration of the controlling precedents reinforces this conclusion. Pp. 11–19.

(1) Diehr and Flook, the cases most directly on point, both addressed processes using mathematical formulas that, like laws of nature, are not themselves patentable. In Diehr, the overall process was patent eligible because of the way the additional steps of the process integrated the equation into the process as a whole. 450 U. S., at 187. These additional steps transformed the process into an inventive application of the formula. But in Flook, the additional steps of the process did not limit the claim to a particular application, and the particular chemical processes at issue were all “well known,” to the point where, putting the formula to the side, there was no “inventive concept” in the claimed application of the formula. 437 U. S., at 594. Here, the claim presents a case for patentability that is weaker than Diehr’s patent-eligible claim and no stronger than Flook’s unpatentable one. The three steps add nothing specific to the laws of nature other than what is well-understood, routine, conventional activity, previously engaged in by those in the field. Pp. 11–13.

(2) Further support for the view that simply appending conventional steps, specified at a high level of generality, to laws of nature, natural phenomena, and abstract ideas cannot make those laws, phenomena, and ideas patentable is provided in O’Reilly v. Morse, 15 How. 62, 114–115; Neilson v. Harford, Webster’s Patent Cases 295, 371; Bilski, supra, at ___–___; and Benson, supra, at 64, 65, 67. Pp. 14–16.

(3) This Court has repeatedly emphasized a concern that patent law not inhibit future discovery by improperly tying up the use of laws of nature and the like. See, e.g., Benson, 409 U. S., at 67, 68. Rewarding with patents those who discover laws of nature might encourage their discovery. But because those laws and principles are “the basic tools of scientific and technological work,” id., at 67, there is a danger that granting patents that tie up their use will inhibit future innovation, a danger that becomes acute when a patented process is no more than a general instruction to “apply the natural law,” or otherwise forecloses more future invention than the underlying discovery could reasonably justify. The patent claims at issue implicate this concern. In telling a doctor to measure metabolite levels and to consider the resulting measurements in light of the correlations they describe, they tie up his subsequent treatment decision regardless of whether he changes his dosage in the light of the inference he draws using the correlations. And they threaten to inhibit the development of more refined treatment recommendations that combine Prometheus’ correlations with later discoveries. This reinforces the conclusion that the processes at issue are not patent eligible, while eliminating any temptation to depart from case law precedent. Pp. 16–19.

(c) Additional arguments supporting Prometheus’ position—that the process is patent eligible because it passes the “machine or transformation test”; that, because the particular laws of nature that the claims embody are narrow and specific, the patents should be upheld; that the Court should not invalidate these patents under §101 because the Patent Act’s other validity requirements will screen out overly broad patents; and that a principle of law denying patent coverage here will discourage investment in discoveries of new diagnostic laws of nature—do not lead to a different conclusion. Pp. 19–24.

628 F. 3d 1347, reversed.

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

The Longshore and Harbor Workers’ Compensation Act (LHWCA) creates a comprehensive scheme to pay compensation for an eligible employee’s disability or death resulting from injury occurring upon the navigable waters of the United States. Benefits for most types of disabilities are capped at twice the national average weekly wage for the fiscal year in which an injured employee is “newly awarded compensation.” 33 U. S. C. §906(c). The LHWCA requires employers to pay benefits voluntarily, without formal administrative proceedings. Typically, employers pay benefits without contesting liability, so no compensation orders are issued. However, if an employer controverts liability, or an employee contests his employer’s actions with respect to his benefits, the dispute proceeds to the Department of Labor’s Office of Workers’ Compensation Programs (OWCP) to be resolved, if possible, through informal procedures. An informal disposition may result in a compensation order. If not resolved informally, the dispute is referred to an administrative law judge (ALJ), who conducts a hearing and issues a compensation order.

In fiscal year 2002, petitioner Roberts was injured at an Alaska marine terminal while working for respondent Sea-Land Services, Inc. Sea-Land (except for six weeks in 2003) voluntarily paid Roberts benefits until fiscal year 2005. Roberts then filed an LHWCA claim, and Sea-Land controverted. In fiscal year 2007, an ALJ awarded Roberts benefits at the fiscal year 2002 statutory maximum rate. Roberts sought reconsideration, contending that the award should have been set at the higher statutory maximum rate for fiscal year 2007, when, he argued, he was “newly awarded compensation” by order of the ALJ. The ALJ denied his motion, and the Department of Labor’s Benefits Review Board affirmed, concluding that the pertinent maximum rate is determined by the date disability commences. The Ninth Circuit affirmed.

Held: An employee is “newly awarded compensation” when he first becomes disabled and thereby becomes statutorily entitled to benefits, no matter whether, or when, a compensation order issues on his behalf. Pp. 5−18.

(a) Roberts contends that the statutory term “awarded compensation” means “awarded compensation in a formal order,” while Sea-Land and the Director, OWCP, maintain that it means “statutorily entitled to compensation because of disability.” Although §906 can be interpreted either way, only Sea-Land and the Director’s interpretation makes §906 a working part of the statutory scheme. Under Roberts’ interpretation, no employee receiving voluntary payments has been “awarded compensation,” so none is subject to an identifiable maximum rate of compensation. That result is incompatible with the LHWCA’s design. Section 906(b)(1) caps compensation at twice the applicable national average weekly wage, as determined by the Secretary of Labor. Section 906(b)(3), in turn, directs the Secretary to determine that wage before each fiscal year begins, at which time it becomes the “applicable national average weekly wage” for the coming fiscal year. And §906(c), in its turn, provides that the Secretary’s determination shall apply to those “newly awarded compensation” during such fiscal year. Through a series of cross-references, the three provisions work together to cap disability benefits. By its terms, and subject to one express exception, §906(b)(1) specifies that the cap applies globally, to all disability claims. Because all three provisions interlock, the cap functions as Congress intended only if §906(c) also applies globally, to all such cases. Roberts’ interpretation would give §906(c) no application in the many cases in which no formal orders issue.

Using the national average weekly wage for the fiscal year in which an employee becomes disabled coheres with the LHWCA’s administrative structure. An employer must be able to calculate the cap in order to pay benefits within 14 days of notice of an employee’s disability, see §914(b), and in order to certify to the Department of Labor whether the maximum rate is being paid. Similarly, an OWCP claims examiner must verify the compensation rate in light of the applicable cap. It is difficult to see how an employer or claims examiner can use a national average weekly wage other than the one in effect at the time an employee becomes disabled. Moreover, applying the national average weekly wage for the fiscal year in which an employee becomes disabled advances the LHWCA’s purpose to compensate disability, defined as “incapacity because of injury to earn the wages which the employee was receiving at the time of injury.” §902(10). It also avoids disparate treatment of similarly situated employees; Roberts’ reading would permit two employees who earn the same salary and suffer the same injury on the same day to receive different maximum compensation rates based on the happenstance of their obtaining orders in different fiscal years. Finally, applying the national average weekly wage for the fiscal year in which disability commences discourages gamesmanship in the claims process. If the fiscal year in which an order issues were to determine the cap, the fact that the national average wage rises each year with inflation would be unduly significant. Roberts’ rule would reward employees who receive voluntary payments with windfalls for initiating unnecessary administrative proceedings to secure a higher rate, while simultaneously punishing employers who have complied fully with their statutory obligations to make voluntary payments. Pp. 5−13.

(b) Roberts’ counterarguments are unconvincing. First, although the LHWCA sometimes uses “award” to mean “award in a formal order,” the presumption that identical words used in different parts of the same Act are intended to have the same meaning, readily yields whenever, as here, the variation in the word’s use in the LHWCA reasonably warrants the conclusion that it was employed in different parts of the Act with different intent. See General Dynamics Land Systems, Inc. v. Cline, 540 U. S. 581.

Second, Roberts argues that, because this Court has refused to read the statutory phrase “person entitled to compensation” in §933(g) to mean “person awarded compensation,” Estate of Cowart v. Nicklos Drilling Co., 505 U. S. 469, the converse must also be true: “awarded compensation” in §906(c) cannot mean “entitled to compensation.” But Cowart’s reasoning does not work in reverse. Cowart did not construe §906(c) or “award,” see id., at 478–479, and it did not hold that the groups of “employees entitled to compensation” and “employees awarded compensation” were mutually exclusive, see id., at 477.

Finally, Roberts contends that his interpretation furthers the LHWCA’s purpose of providing employees with prompt compensation by encouraging employers to avoid delay and expedite administrative proceedings. But his remedy would also punish employers who voluntarily pay benefits at the proper rate from the time of their employees’ injuries, because they would owe benefits under the higher cap applicable in any future fiscal year when their employees chose to file claims. The more measured deterrent to employer delay is interest that accrues from the date an unpaid benefit came due. Pp. 13–18.

625 F. 3d 1204, affirmed.

Sotomayor, J., delivered the opinion of the Court, in which Roberts, C. J., and Scalia, Kennedy, Thomas, Breyer, Alito, and Kagan, JJ., joined. Ginsburg, J., filed an opinion concurring in part and dissenting in part.