Opinions
Opinions
Supreme Court
Sandra Day O'Connor served as a justice on the U.S. Supreme Court from 1981 to 2006. This page lists the opinions she wrote during her time on the court.
Post Retirement Opinions
After her retirement from the Supreme Court, Sandra Day O'Connor continued to hear cases in the U.S. Court of Appeals for the Ninth Circuit as a designated judge.
Arizona Appellate Court Opinions
Sandra Day O'Connor served as a judge on the Arizona Court of Appeals from 1980 to 1981. This page lists the opinions she wrote during her time on the state bench.
Filters
Justice O’Connor delivered the opinion of the Court.
This is a maritime case about a train wreck. A shipment of machinery from Australia was destined for Huntsville, Alabama. The intercontinental journey was uneventful, and the machinery reached the United States unharmed. But the train carrying the machinery on its final, inland leg derailed, causing extensive damage. The machinery’s owner sued the railroad. The railroad seeks shelter in two liability limitations contained in contracts that upstream carriers negotiated for the machinery’s delivery.
I
This controversy arises from two bills of lading (essentially, contracts) for the transportation of goods from Australia to Alabama. A bill of lading records that a carrier has received goods from the party that wishes to ship them, states the terms of carriage, and serves as evidence of the contract for carriage. See 2 T. Schoenbaum, Admiralty and Maritime Law 58–60 (3d ed. 2001) (hereinafter Schoenbaum); Carriage of Goods by Sea Act (COGSA), 49 Stat. 1208, 46 U. S. C. App. §1303. Respondent James N. Kirby, Pty Ltd. (Kirby), an Australian manufacturing company, sold 10 containers of machinery to the General Motors plant located outside Huntsville, Alabama. Kirby hired International Cargo Control (ICC), an Australian freight forwarding company, to arrange for delivery by “through” ( i.e., end-to-end) transportation. (A freight forwarding company arranges for, coordinates, and facilitates cargo transport, but does not itself transport cargo.) To formalize their contract for carriage, ICC issued a bill of lading to Kirby (ICC bill). The bill designates Sydney, Australia, as the port of loading, Savannah, Georgia, as the port of discharge, and Huntsville as the ultimate destination for delivery.
In negotiating the ICC bill, Kirby had the opportunity to declare the full value of the machinery and to have ICC assume liability for that value. Cf. New York, N. H. & H. R. Co. v. Nothnagle, 346 U. S. 128, 135 (1953) (a carrier must provide a shipper with a fair opportunity to declare value). Instead, and as is common in the industry, see Sturley, Carriage of Goods by Sea, 31 J. Mar. L. & Com. 241, 244 (2000), Kirby accepted a contractual liability limitation for ICC below the machinery’s true value, resulting, presumably, in lower shipping rates. The ICC bill sets various liability limitations for the journey from Sydney to Huntsville. For the sea leg, the ICC bill invokes the default liability rule set forth in the Carriage of Goods by Sea Act. The COGSA “package limitation” provides:
“Neither the carrier nor the ship shall in any event be or become liable for any loss or damage to or in connection with the transportation of goods in an amount exceeding $500 per package lawful money of the United States… unless the nature and value of such goods have been declared by the shipper before shipment and inserted into the bill of lading.” 46 U. S. C. App. §1304(5).
For the land leg, in turn, the bill limits the carrier’s liability to a higher amount.[ Footnote 1 ] So that other downstream parties expected to take part in the contract’s execution could benefit from the liability limitations, the bill also contains a so-called “Himalaya Clause.”[ Footnote 2 ] It provides:
“These conditions [for limitations on liability] apply whenever claims relating to the performance of the contract evidenced by this [bill of lading] are made against any servant, agent or other person (including any independent contractor) whose services have been used in order to perform the contract.” App. to Pet. for Cert. 59a, cl. 10.1.
Meanwhile, Kirby separately insured the cargo for its true value with its co-respondent in this case, Allianz Australia Insurance Ltd. (formerly MMI General Insurance, Ltd.).
Having been hired by Kirby, and because it does not itself actually transport cargo, ICC then hired Hamburg Südamerikanische Dampfschiflahrts-Gesellschafft Eggert & Amsinck (Hamburg Süd), a German ocean shipping company, to transport the containers. To formalize their contract for carriage, Hamburg Süd issued its own bill of lading to ICC (Hamburg Süd bill). That bill designates Sydney as the port of loading, Savannah as the port of discharge, and Huntsville as the ultimate destination for delivery. It adopts COGSA’s default rule in limiting the liability of Hamburg Süd, the bill’s designated carrier, to $500 per package. See 46 U. S. C. App. §1304(5). It also contains a clause extending that liability limitation beyond the “tackles”—that is, to potential damage on land as well as on sea. Finally, it too contains a Himalaya Clause extending the benefit of its liability limitation to “all agents… (including inland) carriers… and all independent contractors whatsoever.” App. 63, cl. 5(b).
Acting through a subsidiary, Hamburg Süd hired Petitioner Norfolk Southern Railroad (Norfolk) to transport the machinery from the Savannah port to Huntsville. Delivery failed. The Norfolk train carrying the machinery derailed en route, causing an alleged $1.5 million in damages. Kirby’s insurance company reimbursed Kirby for the loss. Kirby and its insurer then sued Norfolk in the United States District Court for the Northern District of Georgia, asserting diversity jurisdiction and alleging tort and contract claims. In its answer, Norfolk argued, among other things, that Kirby’s potential recovery could not exceed the amounts set forth in the liability limitations contained in the bills of lading for the machinery’s carriage.
The District Court granted Norfolk’s motion for partial summary judgment, holding that Norfolk’s liability was limited to $500 per container. Upon a joint motion from Norfolk and Kirby, the District Court certified its decision for interlocutory review pursuant to 28 U. S. C. §1292(b).
A divided panel of the Eleventh Circuit reversed. It held that Norfolk could not claim protection under the Himalaya Clause in the first contract, the ICC bill. It construed the language of the clause to exclude parties, like Norfolk, that had not been in privity with ICC when ICC issued the bill. 300 F. 3d 1300, 1308–1309 (2002). The majority also suggested that “a special degree of linguistic specificity is required to extend the benefits of a Himalaya clause to an inland carrier.” Id., at 1310. As for the Hamburg Süd bill, the court held that Kirby could be bound by the bill’s liability limitation “only if ICC was acting as Kirby’s agent when it received Hamburg Süd’s bill.” Id., at 1305. And, applying basic agency law principles, the Court of Appeals concluded that ICC had not been acting as Kirby’s agent when it received the bill. Ibid. Based on its opinion that Norfolk was not entitled to benefit from the liability limitation in either bill of lading, the Eleventh Circuit reversed the District Court’s grant of summary judgment for the railroad. We granted certiorari to decide whether Norfolk could take shelter in the liability limitations of either bill, 540 U. S. 1099 (2004), and now reverse.
II
The courts below appear to have decided this case on an assumption, shared by the parties, that federal rather than state law governs the interpretation of the two bills of lading. Respondents now object. They emphasize that, at bottom, this is a diversity case involving tort and contract claims arising out of a rail accident somewhere between Savannah and Huntsville. We think, however, borrowing from Justice Harlan, that “the situation presented here has a more genuinely salty flavor than that.” Kossick v. United Fruit Co., 365 U. S. 731, 742 (1961). When a contract is a maritime one, and the dispute is not inherently local, federal law controls the contract interpretation. Id., at 735.
Our authority to make decisional law for the interpretation of maritime contracts stems from the Constitution’s grant of admiralty jurisdiction to federal courts. See Art. III, §2, cl. 1 (providing that the federal judicial power shall extend to “all Cases of admiralty and maritime Jurisdiction”). See 28 U. S. C. §1333(1) (granting federal district courts original jurisdiction over “[a]ny civil case of admiralty or maritime jurisdiction”); R. Fallon, D. Meltzer, & D. Shapiro, Hart and Wechsler’s The Federal Courts and the Federal System 733–738 (5th ed. 2003). This suit was properly brought in diversity, but it could also be sustained under the admiralty jurisdiction by virtue of the maritime contracts involved. See Pope & Talbot, Inc. v. Hawn, 346 U. S. 406, 411 (1953) (“[S]ubstantial rights… are not to be determined differently whether [a] case is labelled ‘law side’ or ‘admiralty side’ on a district court’s docket”). Indeed, for federal common law to apply in these circumstances, this suit must also be sustainable under the admiralty jurisdiction. See Stewart Organization, Inc. v. Ricoh Corp., 487 U. S. 22, 28 (1988). Because the grant of admiralty jurisdiction and the power to make admiralty law are mutually dependent, the two are often intertwined in our cases.
Applying the two-step analysis from Kossick, we find that federal law governs this contract dispute. Our cases do not draw clean lines between maritime and non-maritime contracts. We have recognized that “[t]he boundaries of admiralty jurisdiction over contracts—as opposed to torts or crimes—being conceptual rather than spatial, have always been difficult to draw.” 365 U. S., at 735. To ascertain whether a contract is a maritime one, we cannot look to whether a ship or other vessel was involved in the dispute, as we would in a putative maritime tort case. Cf. Admiralty Extension Act, 46 U. S. C. App. §740 (“The admiralty and maritime jurisdiction of the United States shall extend to and include all cases of damage or injury… caused by a vessel on navigable water, notwithstanding that such damage or injury be done or consummated on land”); R. Force & M. Norris, 1 The Law of Seamen §1:15 (5th ed. 2003). Nor can we simply look to the place of the contract’s formation or performance. Instead, the answer “depends upon… the nature and character of the contract,” and the true criterion is whether it has “reference to maritime service or maritime transactions.” North Pacific S. S. Co. v. Hall Brothers Marine Railway & Shipbuilding Co., 249 U. S. 119, 125 (1919) (citing Insurance Co. v. Dunham, 11 Wall. 1, 26 (1871)). See also Exxon Corp. v. Central Gulf Lines, Inc., 500 U. S. 603, 611 (1991) (“[T]he trend in modern admiralty case law… is to focus the jurisdictional inquiry upon whether the nature of the transaction was maritime”).
The ICC and Hamburg Süd bills are maritime contracts because their primary objective is to accomplish the transportation of goods by sea from Australia to the eastern coast of the United States. See G. Gilmore & C. Black, Law of Admiralty 31 (2d ed. 1975) (“Ideally, the [admiralty] jurisdiction [over contracts ought] to include those and only those things principally connected with maritime transportation” (emphasis deleted)). To be sure, the two bills call for some performance on land; the final leg of the machinery’s journey to Huntsville was by rail. But under a conceptual rather than spatial approach, this fact does not alter the essentially maritime nature of the contracts.
In Kossick, for example, we held that a shipowner’s promise to assume responsibility for any improper treatment his seaman might receive at a New York hospital was a maritime contract. The seaman had asked the shipowner to pay for treatment by a private physician, but the shipowner, preferring the cheaper public hospital, offered to cover the costs of any complications that might arise from treatment there. We characterized his promise as a “fringe benefit” to a shipowner’s duty in maritime law to provide “ ‘maintenance and cure.’ ” 365 U. S., at 736–737. Because the promise was in furtherance of a “peculiarly maritime concer[n],” id., at 738, it folded into federal maritime law. It did not matter that the site of the inadequate treatment—which gave rise to the contract dispute—was in a hospital on land. Likewise, Norfolk’s rail journey from Savannah to Huntsville was a “fringe” portion of the intercontinental journey promised in the ICC and Hamburg Süd bills.
We have reiterated that the “ ‘fundamental interest giving rise to maritime jurisdiction is “the protection of maritime commerce.” ’ ” Exxon, supra, at 608 (emphasis added) (quoting Sisson v. Ruby, 497 U. S. 358, 367 (1990), in turn quoting Foremost Ins. Co. v. Richardson, 457 U. S. 668, 674 (1982)). The conceptual approach vindicates that interest by focusing our inquiry on whether the principal objective of a contract is maritime commerce. While it may once have seemed natural to think that only contracts embodying commercial obligations between the “tackles” ( i.e., from port to port) have maritime objectives, the shore is now an artificial place to draw a line. Maritime commerce has evolved along with the nature of transportation and is often inseparable from some land-based obligations. The international transportation industry “clearly has moved into a new era—the age of multimodalism, door-to-door transport based on efficient use of all available modes of transportation by air, water, and land.” 1 Schoenbaum 589 (4th ed. 2004). The cause is technological change: Because goods can now be packaged in standardized containers, cargo can move easily from one mode of transport to another. Ibid. See also NLRB v. Longshoremen, 447 U. S. 490, 494 (1980) (“ ‘[C]ontainerization may be said to constitute the single most important innovation in ocean transport since the steamship displaced the schooner’ ” (citation omitted)); G. Muller, Intermodal Freight Transportation 15–24 (3d ed. 1995).
Contracts reflect the new technology, hence the popularity of “through” bills of lading, in which cargo owners can contract for transportation across oceans and to inland destinations in a single transaction. See 1 Schoenbaum 595. Put simply, it is to Kirby’s advantage to arrange for transport from Sydney to Huntsville in one bill of lading, rather than to negotiate a separate contract—and to find an American railroad itself—for the land leg. The popularity of that efficient choice, to assimilate land legs into international ocean bills of lading, should not render bills for ocean carriage nonmaritime contracts.
Some lower federal courts appear to have taken a spatial approach when deciding whether intermodal transportation contracts for intercontinental shipping are maritime in nature. They have held that admiralty jurisdiction does not extend to contracts which require maritime and nonmaritime transportation, unless the nonmaritime transportation is merely incidental—and that long-distance land travel is not incidental. See, e.g., Hartford Fire Ins. Co. v. Orient Overseas Container Lines, 230 F. 3d 549, 555–556 (CA2 2000) (“[T]ransport by land under a bill of lading is not ‘incidental’ to transport by sea if the land segment involves great and substantial distances,” and land transport of over 850 miles across four countries is more than incidental); Sea-Land Serv., Inc. v. Danzig, 211 F. 3d 1373, 1378 (CA Fed. 2000) (holding that intermodal transport contracts were not maritime contracts because they called for “substantial transportation between inland locations and ports both in this country and the Middle East” that was not incidental to the transportation by sea); Kuehne & Nagel (AG & CO) v. Geosource, Inc., 874 F. 2d 283, 290 (CA5 1989) (holding that a through bill of lading calling for land transportation up to 1,000 miles was not a traditional maritime contract because such “extensive land-based operations cannot be viewed as merely incidental to the maritime operations”). As a preliminary matter, it seems to us imprecise to describe the land carriage required by an intermodal transportation contract as “incidental”; realistically, each leg of the journey is essential to accomplishing the contract’s purpose. In this case, for example, the bills of lading required delivery to Huntsville; the Savannah port would not do.
Furthermore, to the extent that these lower court decisions fashion a rule for identifying maritime contracts that depends solely on geography, they are inconsistent with the conceptual approach our precedent requires. See Kossick, 365 U. S., at 735. Conceptually, so long as a bill of lading requires substantial carriage of goods by sea, its purpose is to effectuate maritime commerce—and thus it is a maritime contract. Its character as a maritime contract is not defeated simply because it also provides for some land carriage. Geography, then, is useful in a conceptual inquiry only in a limited sense: If a bill’s sea components are insubstantial, then the bill is not a maritime contract.
Having established that the ICC and Hamburg Süd bills are maritime contracts, then, we must clear a second hurdle before applying federal law in their interpretation. Is this case inherently local? For not “every term in every maritime contract can only be controlled by some federally defined admiralty rule.” Wilburn Boat Co. v. Fireman’s Fund Ins. Co., 348 U. S. 310, 313 (1955) (applying state law to maritime contract for marine insurance because of state regulatory power over insurance industry). A maritime contract’s interpretation may so implicate local interests as to beckon interpretation by state law. See Kossick, supra, at 735. Respondents have not articulated any specific Australian or state interest at stake, though some are surely implicated. But when state interests cannot be accommodated without defeating a federal interest, as is the case here, then federal substantive law should govern. See Kossick, supra, at 739 (the process of deciding whether federal law applies “is surely… one of accommodation, entirely familiar in many areas of overlapping state and federal concern, or a process somewhat analogous to the normal conflict of laws situation where two sovereignties assert divergent interests in a transaction”); 2 Schoenbaum 61 (“Bills of lading issued outside the United States are governed by the general maritime law, considering relevant choice of law rules”).
Here, our touchstone is a concern for the uniform meaning of maritime contracts like the ICC and Hamburg Süd bills. We have explained that Article III’s grant of admiralty jurisdiction “ ‘must have referred to a system of law coextensive with, and operating uniformly in, the whole country. It certainly could not have been the intention to place the rules and limits of maritime law under the disposal and regulation of the several States, as that would have defeated the uniformity and consistency at which the Constitution aimed on all subjects of a commercial character affecting the intercourse of the States with each other or with foreign states.’ ” American Dredging Co. v. Miller, 510 U. S. 443, 451 (1994) (quoting The Lottawanna, 21 Wall. 558, 575 (1875)). See also Yamaha Motor Corp., U. S. A. v. Calhoun, 516 U. S. 199, 210 (1996) (“[I]n several contexts, we have recognized that vindication of maritime policies demanded uniform adherence to a federal rule of decision” (citing Kossick, supra, at 742; Pope & Talbot, 346 U. S., at 409; Garrett v. Moore-McCormack Co., 317 U.S. 239, 248–249 (1942))); Romero v. International Terminal Operating Co., 358 U.S. 354, 373 (1959) (“[S]tate law must yield to the needs of a uniform federal maritime law when this Court finds inroads on a harmonious system[,] [b]ut this limitation still leaves the States a wide scope”).
Applying state law to cases like this one would undermine the uniformity of general maritime law. The same liability limitation in a single bill of lading for international intermodal transportation often applies both to sea and to land, as is true of the Hamburg Süd bill. Such liability clauses are regularly executed around the world. See 1 Schoenbaum 595; Wood, Multimodal Transportation: An American Perspective on Carrier Liability and Bill of Lading Issues, 46 Am. J. Comp. L. 403, 407 (Supp. 1998). See also 46 U. S. C. App. §1307 (permitting parties to extend the COGSA default liability limit to damage done “prior to the loading on and subsequent to the discharge from the ship”). Likewise, a single Himalaya Clause can cover both sea and land carriers downstream, as is true of the ICC bill. See Part III–A, infra. Confusion and inefficiency will inevitably result if more than one body of law governs a given contract’s meaning. As we said in Kossick, when “a [maritime] contract … may well have been made anywhere in the world,” it “should be judged by one law wherever it was made.” 365 U. S., at 741. Here, that one law is federal.
In protecting the uniformity of federal maritime law, we also reinforce the liability regime Congress established in COGSA. By its terms, COGSA governs bills of lading for the carriage of goods “from the time when the goods are loaded on to the time when they are discharged from the ship.” 46 U. S. C. App. §1301(e). For that period, COGSA’s “package limitation” operates as a default rule. §1304(5). But COGSA also gives the option of extending its rule by contract. See §1307 (“Nothing contained in this chapter shall prevent a carrier or a shipper from entering into any agreement, stipulation, condition, reservation, or exemption as to the responsibility and liability of the carrier or the ship for the loss or damage to or in connection with the custody and care and handling of goods prior to the loading on and subsequent to the discharge from the ship on which the goods are carried by sea”). As COGSA permits, Hamburg Süd in its bill of lading chose to extend the default rule to the entire period in which the machinery would be under its responsibility, including the period of the inland transport. Hamburg Süd would not enjoy the efficiencies of the default rule if the liability limitation it chose did not apply equally to all legs of the journey for which it undertook responsibility. And the apparent purpose of COGSA, to facilitate efficient contracting in contracts for carriage by sea, would be defeated.
III
A
Turning to the merits, we begin with the ICC bill of lading, the first of the contracts at issue. Kirby and ICC made a contract for the carriage of machinery from Sydney to Huntsville, and agreed to limit the liability of ICC and other parties who would participate in transporting the machinery. The bill’s Himalaya Clause states:
“These conditions [for limitations on liability] apply whenever claims relating to the performance of the contract evidenced by this [bill of lading] are made against any servant, agent or other person (including any independent contractor) whose services have been used in order to perform the contract.” App. to Pet. for Cert. 59a, cl. 10.1 (emphasis added).
The question presented is whether the liability limitation in Kirby’s and ICC’s contract extends to Norfolk, which is ICC’s sub-subcontractor. The Circuits have split in answering this question. Compare, e.g., Akiyama Corp. of America v. M. V. Hanjin Marseilles, 162 F. 3d 571, 574 (CA9 1998) (privity of contract is not required in order to benefit from a Himalaya Clause), with Mikinberg v. Baltic S. S. Co., 988 F. 2d 327, 332 (CA2 1993) (a contractual relationship is required).
This is a simple question of contract interpretation. It turns only on whether the Eleventh Circuit correctly applied this Court’s decision in Robert C. Herd & Co. v. Krawill Machinery Corp., 359 U. S. 297 (1959). We conclude that it did not. In Herd, the bill of lading between a cargo owner and carrier said that, consistent with COGSA, “ ‘the Carrier’s liability, if any, shall be determined on the basis of $500 per package.’ ” Id., at 302. The carrier then hired a stevedoring company to load the cargo onto the ship, and the stevedoring company damaged the goods. The Court held that the stevedoring company was not a beneficiary of the bill’s liability limitation. Because it found no evidence in COGSA or its legislative history that Congress meant COGSA’s liability limitation to extend automatically to a carrier’s agents, like stevedores, the Court looked to the language of the bill of lading itself. It reasoned that a clause limiting “ ‘the Carrier’s liability’ ” did not “indicate that the contracting parties intended to limit the liability of stevedores or other agents…. If such had been a purpose of the contracting parties it must be presumed that they would in some way have expressed it in the contract.” Ibid. The Court added that liability limitations must be “strictly construed and limited to intended beneficiaries.” Id., at 305.
The Eleventh Circuit, like respondents, made much of the Herd decision. Deriving a principle of narrow construction from Herd, the Court of Appeals concluded that the language of the ICC bill’s Himalaya Clause is too vague to clearly include Norfolk. 300 F. 3d, at 1308. Moreover, the lower court interpreted Herd to require privity between the carrier and the party seeking shelter under a Himalaya Clause. Id., at 1308. But nothing in Herd requires the linguistic specificity or privity rules that the Eleventh Circuit attributes to it. The decision simply says that contracts for carriage of goods by sea must be construed like any other contracts: by their terms and consistent with the intent of the parties. If anything, Herd stands for the proposition that there is no special rule for Himalaya Clauses.
The Court of Appeals’ ruling is not true to the contract language or to the intent of the parties. The plain language of the Himalaya Clause indicates an intent to extend the liability limitation broadly—to “ any servant, agent or other person (including any independent contractor)” whose services contribute to performing the contract. App. to Pet. for Cert. 59a, cl.10.1 (emphasis added). “Read naturally, the word ‘any’ has an expansive meaning, that is, ‘one or some indiscriminately of whatever kind.’ ” United States v. Gonzales, 520 U. S. 1, 5 (1997) (quoting Webster’s Third New International Dictionary 97 (1976)). There is no reason to contravene the clause’s obvious meaning. See Green v. Biddle, 8 Wheat. 1, 89–90 (1823) (“[W]here the words of a law, treaty, or contract, have a plain and obvious meaning, all construction, in hostility with such meaning, is excluded”). The expansive contract language corresponds to the fact that various modes of transportation would be involved in performing the contract. Kirby and ICC contracted for the transportation of machinery from Australia to Huntsville, Alabama, and, as the crow flies, Huntsville is some 366 miles inland from the port of discharge. See G. Fitzpatrick & M. Modlin, Direct-Line Distances 168 (1986). Thus, the parties must have anticipated that a land carrier’s services would be necessary for the contract’s performance. It is clear to us that a railroad like Norfolk was an intended beneficiary of the ICC bill’s broadly written Himalaya Clause. Accordingly, Norfolk’s liability is limited by the terms of that clause.
B
The question arising from the Hamburg Süd bill of lading is more difficult. It requires us to set an efficient default rule for certain shipping contracts, a task that has been a challenge for courts for centuries. See, e.g., Hadley v. Baxendale, 9 Exch. 341, 156 Eng. Rep. 145 (1854). ICC and Hamburg Süd agreed that Hamburg Süd would transport the machinery from Sydney to Huntsville, and agreed to the COGSA “package limitation” on the liability of Hamburg Süd, its agents, and its independent contractors. The second question presented is whether that liability limitation, which ICC negotiated, prevents Kirby from suing Norfolk (Hamburg Süd’s independent contractor) for more. As we have explained, the liability limitation in the ICC bill, the first contract, sets liability for a land accident higher than this bill does. See n. 1, supra. Because Norfolk’s liability will be lower if it is protected by the Hamburg Süd bill too, we must reach this second question in order to give Norfolk the full relief for which it petitioned.
To interpret the Hamburg Süd bill, we turn to a rule drawn from our precedent about common carriage: When an intermediary contracts with a carrier to transport goods, the cargo owner’s recovery against the carrier is limited by the liability limitation to which the intermediary and carrier agreed. The intermediary is certainly not automatically empowered to be the cargo owner’s agent in every sense. That would be unsustainable. But when it comes to liability limitations for negligence resulting in damage, an intermediary can negotiate reliable and enforceable agreements with the carriers it engages.
We derive this rule from our decision about common carriage in Great Northern R. Co. v. O’Connor, 232 U. S. 508 (1914). In Great Northern, an owner hired a transfer company to arrange for the shipment of her goods. Without the owner’s express authority, the transfer company arranged for rail transport at a tariff rate that limited the railroad’s liability to less than the true value of the goods. The goods were lost en route, and the owner sued the railroad. The Court held that the railroad must be able to rely on the liability limitation in its tariff agreement with the transfer company. The railroad “had the right to assume that the Transfer Company could agree upon the terms of the shipment”; it could not be expected to know if the transfer company had any outstanding, conflicting obligation to another party. Id., at 514. The owner’s remedy, if necessary, was against the transfer company. Id., at 515.
Respondents object to our reading of Great Northern, and argue that this Court should fashion the federal rule of decision from general agency law principles. Like the Eleventh Circuit, respondents reason that Kirby cannot be bound by the bill of lading that ICC negotiated with Hamburg Süd unless ICC was then acting as Kirby’s agent. Other Courts of Appeals have also applied agency law to cases similar to this one. See, e.g., Kukje Hwajae Ins. Co. v. The M/V Hyundai Liberty, 294 F. 3d 1171, 1175–1177 (CA9 2002) (an intermediary acted as a cargo owner’s agent when negotiating a bill of lading with a downstream carrier).
We think reliance on agency law is misplaced here. It is undeniable that the traditional indicia of agency, a fiduciary relationship and effective control by the principal, did not exist between Kirby and ICC. See Restatement (Second) of Agency §1 (1957). But that is of no moment. The principle derived from Great Northern does not require treating ICC as Kirby’s agent in the classic sense. It only requires treating ICC as Kirby’s agent for a single, limited purpose: when ICC contracts with subsequent carriers for limitation on liability. In holding that an intermediary binds a cargo owner to the liability limitations it negotiates with downstream carriers, we do not infringe on traditional agency principles. We merely ensure the reliability of downstream contracts for liability limitations. In Great Northern, because the intermediary had been “entrusted with goods to be shipped by railway, and, nothing to the contrary appearing, the carrier had the right to assume that [the intermediary] could agree upon the terms of the shipment.” 232 U. S., at 514. Likewise, here we hold that intermediaries, entrusted with goods, are “agents” only in their ability to contract for liability limitations with carriers downstream.
Respondents also contend that any decision binding Kirby to the Hamburg Süd bill’s liability limitation will be disastrous for the international shipping industry. Various participants in the industry have weighed in as amici on both sides in this case, and we must make a close call. It would be idle to pretend that the industry can easily be characterized, or that efficient default rules can easily be discerned. In the final balance, however, we disagree with respondents for three reasons.
First, we believe that a limited agency rule tracks industry practices. In intercontinental ocean shipping, carriers may not know if they are dealing with an intermediary, rather than with a cargo owner. Even if knowingly dealing with an intermediary, they may not know how many other intermediaries came before, or what obligations may be outstanding among them. If the Eleventh Circuit’s rule were the law, carriers would have to seek out more information before contracting, so as to assure themselves that their contractual liability limitations provide true protection. That task of information gathering might be very costly or even impossible, given that goods often change hands many times in the course of intermodal transportation. See 1 Schoenbaum 589; Wood, 46 Am. J. Comp. L., at 404.
Second, if liability limitations negotiated with cargo owners were reliable while limitations negotiated with intermediaries were not, carriers would likely want to charge the latter higher rates. A rule prompting downstream carriers to distinguish between cargo owners and intermediary shippers might interfere with statutory and decisional law promoting nondiscrimination in common carriage. Cf. ICC v. Delaware, L. & W. R. Co., 220 U. S. 235, 251–256 (1911) (common carrier cannot “sit in judgment on the title of the prospective shipper”); Shipping Act, 46 U. S. C. App. §1709 (nondiscrimination rules). It would also, as we have intimated, undermine COGSA’s liability regime.
Finally, as in Great Northern, our decision produces an equitable result. See 232 U. S., at 515. Kirby retains the option to sue ICC, the carrier, for any loss that exceeds the liability limitation to which they agreed. And indeed, Kirby has sued ICC in an Australian court for damages arising from the Norfolk derailment. It seems logical that ICC—the only party that definitely knew about and was party to both of the bills of lading at issue here—should bear responsibility for any gap between the liability limitations in the bills. Meanwhile, Norfolk enjoys the benefit of the Hamburg Süd bill’s liability limitation.
IV
We hold that Norfolk is entitled to the protection of the liability limitations in the two bills of lading. Having undertaken this analysis, we recognize that our decision does no more than provide a legal backdrop against which future bills of lading will be negotiated. It is not, of course, this Court’s task to structure the international shipping industry. Future parties remain free to adapt their contracts to the rules set forth here, only now with the benefit of greater predictability concerning the rules for which their contracts might compensate.
The judgment of the United States Court of Appeals for the Eleventh Circuit is reversed, and the case is remanded for further proceedings consistent with this opinion.
It is so ordered.
Notes
Footnote 1
The bill provides that “the Freight Forwarder shall in no event be or become liable for any loss of or damage to the goods in an amount exceeding the equivalent of 666.67 SDR per package or unit or 2 SDR per kilogramme of gross weight of the goods lost or damaged, whichever is the higher, unless the nature and value of the goods shall have been declared by the Consignor.” App. to Pet. for Cert. 57a, cl. 8.3. An SDR, or Special Drawing Right, is a unit of account created by the International Monetary Fund and calculated daily on the basis of a basket of currencies. Liability computed per package for the 10 containers, for example, was approximately $17,373 when the bill of lading issued in June 1997, $17,231 when the goods were damaged on October 9, 1997, and $9,763 when the case was argued. See International Monetary Fund Exchange Rate Archives, http://www.imf.org/external/np/fin/rates/param_rms_mth.cfm (as visited Nov. 5, 2004, and available in Clerk of Court’s case file). Respondents claim that liability computed by weight is higher. The machinery’s weight is not in the record. In any case, because we conclude that Norfolk is also protected by the $500 per package limit in the second bill of lading at issue here, see Part III–B, infra, and thus cannot be liable for more than $5,000 for the 10 containers, each holding one machine, the precise liability under the ICC bill of lading does not matter.
Footnote 2
Clauses extending liability limitations take their name from an English case involving a steamship called Himalaya. See Adler v. Dickson, [1955] 1 Q. B. 158 (C. A.).
Justice O’Connor delivered the opinion of the Court.
These cases present the question whether conviction for conspiracy to commit money laundering, in violation of 18 U. S. C. §1956(h), requires proof of an overt act in furtherance of the conspiracy. We hold that it does not.
I
In March 1999, a federal grand jury returned a 20-count indictment against petitioners and five codefendants. As relevant here, Count II of the indictment charged petitioners with conspiracy to launder money, in violation of §1956(h). The indictment described, in general terms, the “manner and means” used to accomplish the objects of the money laundering conspiracy, but it did not charge the defendants with the commission of any overt act in furtherance thereof.
At trial, the Government presented evidence that petitioners were members of the executive board of an entity known as Greater Ministries International Church (GMIC). GMIC operated a “gifting” program that took in more than $400 million between 1996 and 1999. Under that program, petitioners and others induced unwary investors to give money to GMIC with promises that investors would receive double their money back within a year and a half. Petitioners marketed the program throughout the country, claiming that GMIC would generate returns on investors’ “gifts” through overseas investments in gold and diamond mining, commodities, and offshore banks. Investors were told that GMIC would use some of the profits for philanthropic purposes. Most of these claims were false. GMIC made none of the promised investments, had no assets, and gave virtually nothing to charity. Many participants in GMIC’s program received little or no return on their money, and their investments indeed largely turned out to be “gifts” to GMIC representatives. Petitioners together allegedly received more than $1.2 million in commissions on the money they solicited.
At the close of the evidence, petitioners asked the District Court to instruct the jury that the Government was required to prove beyond a reasonable doubt that at least one of the co-conspirators had committed an overt act in furtherance of the money laundering conspiracy. The court denied that request, and the jury returned a verdict of guilty on the money laundering conspiracy charge.
The Eleventh Circuit affirmed petitioners’ convictions, holding, in relevant part, that the jury instructions approved by the District Court were proper because §1956(h) does not require proof of an overt act. 349 F. 3d 1320, 1324 (2003). The Court of Appeals noted that some of its sister Circuits had taken the opposite position. Id., at 1323 (citing United States v. Wilson, 249 F. 3d 366, 379 (CA5 2001); United States v. Hildebrand, 152 F. 3d 756, 762 (CA8 1998)). It concluded, however, that those decisions were erroneously based on case law interpreting the general conspiracy statute, 18 U. S. C. §371, which, unlike §1956(h), expressly includes an overt-act requirement. 349 F. 3d, at 1323. The Eleventh Circuit instead relied upon United States v. Shabani, 513 U. S. 10 (1994), where we held that the drug conspiracy statute, 21 U. S. C. §846, does not require proof of an overt act. Because the language of 18 U. S. C. §1956(h) and 21 U. S. C. §846 is “nearly identical,” the Eleventh Circuit found itself compelled to follow the reasoning of Shabani in holding that §1956(h), too, requires no proof of an overt act. 349 F. 3d, at 1323–1324. We granted certiorari to resolve the conflict among the Circuits on the question presented, 542 U. S. ___ (2004), and we now affirm the decision below.
II
Congress enacted 18 U. S. C. §§1956 and 1957 (2000 ed. and Supp. II) as part of the Money Laundering Control Act of 1986, Pub. L. 99–570, 100 Stat. 3207–18. Section 1956 penalizes the knowing and intentional transportation or transfer of monetary proceeds from specified unlawful activities, while §1957 addresses transactions involving criminally derived property exceeding $10,000 in value. As originally enacted, neither section included a conspiracy provision. Accordingly, the Government relied on the general conspiracy statute, 18 U. S. C. §371, to prosecute conspiracies to commit the offenses set forth in §§1956 and 1957. In 1992, however, Congress enacted the money laundering conspiracy provision at issue in these cases, now codified at 18 U. S. C. §1956(h). See Annunzio-Wylie Anti-Money Laundering Act, Pub. L. 102–550, §1530, 106 Stat. 4066. Section 1956(h) provides: “Any person who conspires to commit any offense defined in [§1956] or section 1957 shall be subject to the same penalties as those prescribed for the offense the commission of which was the object of the conspiracy.”
In Shabani, we addressed whether the nearly identical language of the drug conspiracy statute, 21 U. S. C. §846, requires proof of an overt act. See ibid. (“Any person who attempts or conspires to commit any offense defined in this subchapter shall be subject to the same penalties as those prescribed for the offense, the commission of which was the object of the attempt or conspiracy”). We held that it does not, relying principally upon our earlier decisions in Nash v. United States, 229 U. S. 373 (1913), and Singer v. United States, 323 U. S. 338 (1945). See Shabani, supra, at 13–14. In each of those cases, the Court held that, where Congress had omitted from the relevant conspiracy provision any language expressly requiring an overt act, the Court would not read such a requirement into the statute. See Singer, supra, at 340 (Selective Training and Service Act of 1940); Nash, supra, at 378 (Sherman Act).
As we explained in Shabani, these decisions “follow the settled principle of statutory construction that, absent contrary indications, Congress intends to adopt the common law definition of statutory terms. See Molzof v. United States, 502 U. S. 301, 307–308 (1992). We have consistently held that the common law understanding of conspiracy ‘does not make the doing of any act other than the act of conspiring a condition of liability.’ ” 513 U. S., at 13–14 (quoting Nash, supra, at 378). In concluding that the drug conspiracy statute in Shabani did not require proof of an overt act, we found instructive the distinction between that statute and the general conspiracy statute, §371 , which supersedes the common law rule by expressly including an overt-act requirement. 513 U. S., at 14. See 18 U. S. C. §371 (“If two or more persons conspire either to commit any offense against the United States, or to defraud the United States, or any agency thereof in any manner or for any purpose, and one or more of such persons do any act to effect the object of the conspiracy, each shall be fined under this title or imprisoned not more than five years, or both” (emphasis added)).
Shabani distilled the governing rule for conspiracy statutes as follows: “ ‘ Nash and Singer give Congress a formulary: by choosing a text modeled on §371, it gets an overt-act requirement; by choosing a text modeled on the Sherman Act, 15 U. S. C. §1 [which, like 21 U. S. C. §846, omits any express overt-act requirement], it dispenses with such a requirement.’ ” 513 U. S., at 14 (quoting United States v. Sassi, 966 F. 2d 283, 284 (CA7 1992)). This rule dictates the outcome in the instant cases as well: Because the text of §1956(h) does not expressly make the commission of an overt act an element of the conspiracy offense, the Government need not prove an overt act to obtain a conviction.
III
Petitioners argue that the rule that governed Shabani is inapplicable here, because §1956(h) does not establish a new conspiracy offense; rather, they say, it merely increases the penalty for conviction of a money laundering conspiracy under §371. In other words, as we understand their argument, petitioners contend that the Government must continue to prosecute money laundering conspiracies under §371, but that §1956(h) now provides enhanced penalties for conviction. Since §371 contains an overt act requirement, the argument goes, the Government must prove an overt act in prosecutions ostensibly brought under §1956(h). This reading of §1956(h) is untenable for two principal reasons. First, petitioners concede—as they must—that §1956(h)’s text is sufficient to establish an offense. Indeed, its language is nearly identical to the drug conspiracy statute at issue in Shabani, which indisputably created an offense. Second, petitioners apparently read §1956(h) to supply an enhanced penalty for violation of §371 in cases where the object of the conspiracy is to violate the substantive money laundering offenses in §§1956(a) and 1957. But the text of §1956(h) fails to provide any cross-reference to §371. Mere use of the word “conspires” surely is not enough to establish the necessary link between these two separate statutes. In short, if Congress had intended to create the scheme petitioners envision, it would have done so in clearer terms.
Petitioners seek support for their construction of §1956(h) in the provision’s legislative history. They contend that this history contains no indication that Congress meant to create a new offense or to eliminate the pre-existing overt-act requirement for money laundering conspiracy prosecutions that hitherto had been brought under §371. They say that the history instead shows that §1956(h) was intended only to raise the penalty for money laundering conspiracy from the 5-year maximum sentence under §371 to the greater maximums available for substantive money laundering offenses under §§1956(a) and 1957. Petitioners also point out that, when Congress enacted §1956(h), it did so under the title “ Penalty for Money Laundering Conspiracies,” 106 Stat., at 4066 (emphasis added). Had Congress wanted to enact an “offense” provision, they argue, it would have titled it accordingly.
Because the meaning of §1956(h)’s text is plain and unambiguous, we need not accept petitioners’ invitation to consider the legislative history. But even were we to do so, we would reach the same conclusion. It is undisputed that Congress intended §1956(h) to increase the penalties for money laundering conspiracies. The provision’s text makes clear that Congress did so precisely by establishing a new offense. Given the clarity of the text, mere silence in the legislative history cannot justify reading an overt-act requirement, or a cross-reference to §371, into §1956(h). See, e.g., United States v. Wells, 519 U. S. 482, 496–497 (1997) (refusing to read a materiality element into the statute at issue based on silence in the legislative history); Harrison v. PPG Industries, Inc., 446 U. S. 578, 592 (1980) (“[I]t would be a strange canon of statutory construction that would require Congress to state in committee reports or elsewhere in its deliberations that which is obvious on the face of a statute”). Nor do we find it significant that Congress chose to label §1956(h) a “penalty” rather than an “offense” provision. See Pennsylvania Dept. of Corrections v. Yeskey, 524 U. S. 206, 212 (1998) (“ ‘[t]he title of a statute … cannot limit the plain meaning of the text’ ”); Castillo v. United States, 530 U. S. 120, 125 (2000) (although “[t]he title of the entirety of §924 is ‘Penalties’ … at least some portion of §924 … creates, not penalty enhancements, but entirely new crimes”).
Petitioners’ legislative history argument is particularly inapt here, we might add, because Congress is presumed to have knowledge of the governing rule described in Shabani. While Shabani was decided two years after §1956(h) was enacted, the rule it articulated was established decades earlier in Nash and Singer. These decisions establish a “formulary” that provides clear and predictable guidance to Congress. As the Government points out, Congress has included an express overt-act requirement in at least 22 other current conspiracy statutes, clearly demonstrating that it knows how to impose such a requirement when it wishes to do so. See Brief for United States 11, and n. 5 (citing statutes). Where Congress has chosen not to do so, we will not override that choice based on vague and ambiguous signals from legislative history.
We conclude by addressing two arguments raised by petitioners relating to the text and structure of §1956 as a whole. First, petitioners note that Congress placed each of the three substantive money laundering offenses in §1956 under subsection (a). Had the drafters intended §1956(h) to create a new offense, petitioners contend, they would have placed it with the other offenses in subsection (a) instead of in its own separate subsection. We fail to see why that should be so. The three offenses placed in subsection (a) share a common feature: All are substantive money-laundering crimes. We find nothing remarkable in Congress’ decision to place a qualitatively different conspiracy offense provision in a separate subsection.
Petitioners’ second textual argument is based on §1956(i) (2000 ed., Supp. II), a venue provision added to the statute in 2001. See USA PATRIOT ACT, Pub. L. 107–56, §1004, 115 Stat. 392. Section 1956(i)(2) (2000 ed., Supp. II) provides that “[a] prosecution for an attempt or conspiracy offense under [§1956 or §1957] may be brought in the district where venue would lie for the completed offense under [§1956(i)(1)], or in any other district where an act in furtherance of the attempt or conspiracy took place.” Petitioners contend that, by setting venue in the district where an overt act took place, Congress confirmed what (petitioners say) was the majority view of the Courts of Appeals at the time of §1956(i)’s enactment: that proof of an overt act was required under §1956(h). Moreover, petitioners argue, setting venue where an overt act took place makes little sense if such an act is not an element of the offense.
This argument fails for several reasons. As a preliminary matter, petitioners assume that §1956(i) is the sole provision setting venue in money laundering conspiracy prosecutions. Although we need not definitively construe that provision here, we note that its language appears permissive rather than exclusive—§1956(i) says a conspiracy prosecution “ may be brought” in a district meeting the specified criteria. (Emphasis added.) This suggests that the provision serves to supplement, rather than supplant, the default venue rule: “Unless a statute or these rules permit otherwise, the government must prosecute an offense in a district where the offense was committed.” Fed. Rule Crim. Proc. 18. For a conspiracy prosecution under the common law rule, the district in which the unlawful agreement was reached would satisfy this default venue rule. See Hyde v. Shine, 199 U. S. 62, 76 (1905).
But even if we assume, for the sake of argument, that §1956(i) is an exclusive venue provision, petitioners’ argument still fails. The provision authorizes two alternative venues for money laundering conspiracy prosecutions: (1) the district in which venue would lie if the completed substantive money laundering offense had been accomplished, or (2) any district in which an overt act in furtherance of the conspiracy was committed. The first venue option clearly does not require that any overt act have been committed, and the Government therefore need not allege or prove such an act for venue to be properly established under this portion of §1956(i). As to the second venue option, this Court has long held that venue is proper in any district in which an overt act in furtherance of the conspiracy was committed, even where an overt act is not a required element of the conspiracy offense. See, e.g., United States v. Socony-Vacuum Oil Co., 310 U. S. 150, 252 (1940); United States v. Trenton Potteries Co., 273 U. S. 392, 402–404 (1927). In light of this longstanding rule, §1956(i)(2)’s authorization of venue in a district where an overt act took place cannot be taken to indicate that Congress deemed such an act necessary for conviction under §1956(h). Instead, Congress appears merely to have confirmed the availability of this alternative venue option in money laundering conspiracy cases.
* * *
For the reasons set forth above, we hold that conviction for conspiracy to commit money laundering, in violation of 18 U. S. C. §1956(h), does not require proof of an overt act in furtherance of the conspiracy. Accordingly, the judgment of the Court of Appeals is affirmed.
It is so ordered.
Justice O’Connor, concurring.
I join the Court’s opinion. I write separately to emphasize that, even under the current statutory scheme, it is possible for the Government to detain inadmissible aliens for more than six months after they have been ordered removed. For one thing, the 6-month presumption we described in Zadvydas v. Davis, 533 U. S. 678 (2001), is just that—a presumption. The Court notes that the Government has not suggested here any reason why it takes longer to effect removal of inadmissible aliens than it does to effect removal of other aliens. It is conceivable, however, that a longer period is “reasonably necessary,” id., at 689, to effect removal of inadmissible aliens as a class. If the Government shows that to be true, then detention beyond six months will be lawful within the meaning we ascribed to 8 U. S. C. §1231(a)(6) in Zadvydas.
Moreover, the Government has other statutory means for detaining aliens whose removal is not foreseeable and whose presence poses security risks. Upon certifying that he has “reasonable grounds to believe” an alien has engaged in certain terrorist or other dangerous activity specified by statute, 8 U. S. C. §1226a(a)(3) (2000 ed., Supp. II), the Secretary of Homeland Security may detain that alien for successive six-month periods “if the release of the alien will threaten the national security of the United States or the safety of the community or any person,” §1226a(a)(6).
Finally, any alien released as a result of today’s holding remains subject to the conditions of supervised release. See 8 U. S. C. §1231(a)(3); 8 CFR §241.5 (2004). And, if he fails to comply with the conditions of release, he will be subject to criminal penalties—including further detention. See 8 U. S. C. §1253(b); Zadvydas, supra, at 695 (“We nowhere deny the right of Congress… to subject [aliens] to supervision within conditions when released from detention, or to incarcerate them where appropriate for violations of those conditions”)., with whom The Chief Justice joins as to Part I–A, dissenting.
Title 8 U. S. C. §1231(a)(6) states that aliens whom the Secretary of Homeland Security has ordered removed “may be detained beyond the removal period.” Nevertheless, in Zadvydas v. Davis, 533 U. S. 678 (2001), this Court construed this provision “to contain an implicit ‘reasonable time’ limitation” on the Secretary’s power to detain admitted aliens “[b]ased on our conclusion that indefinite detention of” those aliens “would raise serious constitutional concerns.” Id., at 682. “Aliens who have not yet gained initial admission to this country,” the Court assured us, “would present a very different question.” Ibid.
Today, the Court holds that this constitutional distinction—which “made all the difference” to the Zadvydas Court, id., at 693—is actually irrelevant, because “[t]he operative language of §1231(a)(6) … applies without differentiation to all three categories of aliens that are its subject.” Ante, at 6. While I wholeheartedly agree with the Court’s fidelity to the text of §1231(a)(6), the Court’s analysis cannot be squared with Zadvydas. And even if it could be so squared, Zadvydas was wrongly decided and should be overruled. I respectfully dissent.
I
I begin by addressing the majority’s interpretation of Zadvydas. The Court’s interpretation is not a fair reading of that case. It is also not required by any sound principle of statutory construction of which I am aware. To the contrary, what drives the majority’s reading is a novel “lowest common denominator” principle. Ante, at 8.
A
The majority’s reading of Zadvydas is implausible. Zadvydas held that interpreting §1231(a)(6) to authorize indefinite detention of admitted aliens later found removable would raise serious due process concerns. 533 U. S., at 690–696. The Court therefore read the statute to permit the Attorney General (now the Secretary of Homeland Security) to detain admitted aliens only as long as reasonably necessary to remove them from the country. Id., at 699.
The majority concedes that Zadvydas explicitly reserved the question whether its statutory holding as to admitted aliens applied equally to inadmissible aliens. Ante, at 7. This reservation was front and center in Zadvydas. It appeared in the introduction and is worth repeating in full:
“In these cases, we must decide whether [§1231(a)(6)] authorizes the Attorney General to detain a removable alien indefinitely beyond the removal period or only for a period reasonably necessary to secure the alien’s removal. We deal here with aliens who were admitted to the United States but subsequently ordered removed. Aliens who have not yet gained initial admission to this country would present a very different question. Based on our conclusion that indefinite detention of aliens in the former category would raise serious constitutional concerns, we construe the statute to contain an implicit ‘reasonable time’ limitation, the application of which is subject to federal-court review.” 533 U. S., at 682 (citation omitted; emphasis in original).
The Court reserved this question because the constitutional questions raised by detaining inadmissible aliens are different from those raised by detaining admitted aliens. It stated that the detention period in §1231(a)(6) was limited because it “read [the statute] in light of the Constitution’s demands.” Id., at 689. And it repeatedly emphasized constitutional distinctions among various groups of aliens, for which §1231(a)(6) makes no distinctions. See id., at 693–694 (noting the different constitutional considerations applicable to inadmissible and admissible aliens); id., at 695 (noting that “the cases before us [do not] require us to consider the political branches’ authority to control entry into the United States”); id., at 696 (noting that the opinion did not “consider terrorism or other special circumstances where special arguments might be made for forms of preventive detention and for heightened deference to the judgments of the political branches with respect to matters of national security”).
The majority’s reading of Zadvydas is inconsistent with these qualifications. If it were true that Zadvydas ’ interpretation of §1231(a)(6) applied to all aliens regardless of the constitutional concerns involved in each case, then the question of how §1231(a)(6) applies to them would not be “very different” depending on the alien before the Court. The question would be trivial, because the text of §1231(a)(6) plainly does not distinguish between admitted and nonadmitted aliens. There would also have been no need for the Court to go out of its way to leave aside “terrorism or other special circumstances,” id., at 696, or to disavow “considerat[ion of] the political branches’ authority to control entry into the United States,” id., at 695, for the construction the majority extracts from Zadvydas would have applied across the board, ibid. And the Court’s rationalization that its construction would therefore “leave no unprotected spot in the Nation’s armor,” id., at 695–696 (internal quotation marks omitted), would have been incorrect. The constitutional distinctions that pervade Zadvydas are evidence that the “very different” statutory question it reserved turned on them.
The Zadvydas Court thus tethered its reading of §1231(a)(6) to the specific class of aliens before it. The term this Court read into the statute was not simply a presumptive 6-month period, but a presumptive 6-month period for admitted aliens. Its reading of the statute “in light of the Constitution’s demands,” id., at 689, that is, depended on the constitutional considerations at work in “ the cases before [it], ” id., at 695 (emphasis added). One would expect the Court today, then, to follow the same two-step procedure it employed in Zadvydas. It should first ask whether the statute is ambiguous and, if so, whether one of the possible interpretations raises constitutional doubts as applied to Martinez and Benitez. Step one is dictated by Zadvydas: §1231(a)(6) is not clear on whether it permits indefinite detention. The Court should then move to the second step and ask whether either of the statute’s possible interpretations raises constitutional doubts as applied to Benitez and Martinez. If so, the Court would apply avoidance to adopt the interpretation free from constitutional doubt (as Zadvydas itself did).
The Court’s reasons for departing from this reading of Zadvydas are unpersuasive. The Court says that its reading is necessary to avoid “invent[ing] a statute rather than interpret[ing] one,” ante, at 6; to preclude “giving the same detention provision a different meaning” depending on the aliens before the Court, ante, at 8 (emphasis in original); and to forestall establishing “the dangerous principle that judges can give the same statutory text different meanings in different cases,” ante, at 15. I agree that we should adopt none of these principles, but this is no warrant for the reading of Zadvydas that the majority advocates. Zadvydas established a single and unchanging, if implausible, meaning of §1231(a)(6): that the detention period authorized by §1231(a)(6) depends not only on the circumstances surrounding a removal, but also on the type of alien ordered removed.
I grant that this understanding of Zadvydas could result in different detention periods for different classes of aliens—indefinite detention for some, limited detention for others. But it does not follow that this reads the meaning of the statute to “change” depending on the alien involved, any more than the meaning of the statute could be said to “change” simply because the time that is “reasonably necessary to effect removal” may differ depending on the type of alien involved, as both the Court’s opinion, ante, at 15, and Justice O’Connor’s concurring opinion, ante, at 1, concede it may. A statute’s sense is the same even if what it requires depends on factual context.
In support of its reading of Zadvydas, the Court relies on a statement in a dissent in Zadvydas that §1231(a)(6) could not be given a different reading for inadmissible aliens. Ante, at 8 (citing 533 U. S., at 710–711, 717 (opinion of Kennedy, J.)). That dissenting view, as the very quotation the majority stresses demonstrates, rested on the dissent’s premise that “it is not a plausible construction of §1231(a)(6) to imply a time limit as to one class and not to another.” Id., at 710 (opinion of Kennedy, J.). But the Zadvydas majority disagreed with that assumption and adopted a contrary interpretation of §1231(a)(6). For as the dissent recognized, Zadvydas ’ “logic might be that inadmissible and removable aliens might be treated differently.” Ibid. That was Zadvydas ’ logic precisely, as its repeated statements limiting its decision to inadmissible aliens show. To interpret Zadvydas properly, we must take its logic as given, not the logic of the reductio ad absurdum of Zadvydas that I joined in dissent.
B
The majority strains to recharacterize Zadvydas because it thinks that “[i]t is not at all unusual to give a statute’s ambiguous language a limiting construction called for by one of the statute’s applications, even though other of the statute’s applications, standing alone, would not support the same limitation.” Ante, at 8. In other words, it claims, “[t]he lowest common denominator, as it were, must govern.” Ibid. I disagree.
As an initial matter, this principle is inconsistent with Zadvydas itself. As explained above, the limiting construction Zadvydas adopted as to admitted aliens does not necessarily govern the other applications of §1231(a)(6). If the majority is correct that the “lowest common denominator” governs, then the careful distinction Zadvydas drew between admitted aliens and nonadmitted aliens was irrelevant at best and misleading at worst. Under this reading, Zadvydas would have come out the same way even if it had involved inadmissible aliens, for the “lowest common denominator” of the statute remains the same regardless of the identity of the alien before the Court. Again, this understanding of Zadvydas is implausible.
Beyond Zadvydas, the Court offers scant support for the idea that statutes should be stripped down to their “lowest common denominator[s].” It attempts to distinguish Jinks v. Richland County, 538 U. S. 456 (2003), and Raygor v. Regents of Univ. of Minn., 534 U. S. 533 (2002), ante, at 11–12, and n. 6, yet these cases employed exactly the procedure that the majority today says is impermissible. They construed 28 U. S. C. §1367(d),[ Footnote 1 ] a tolling provision, to apply to States and political subdivisions of States only to the extent that doing so would raise a constitutional doubt as applied to either entity. Jinks was explicit on this point:
“Although we held in [Raygor] that §1367(d) does not apply to claims filed in federal court against States but subsequently dismissed on sovereign immunity grounds, we did so to avoid interpreting the statute in a manner that would raise ‘serious constitutional doubt’ in light of our decisions protecting a State ’ s sovereign immunity from congressional abrogation …. [N]o such constitutional doubt arises from holding that petitioner’s claim against respondent—which is not a State, but a political subdivision of a State—falls under the definition of ‘ any claim asserted under subsection (a) [of §1367].’ ” 538 U. S., at 466 (citation omitted; emphasis in original).
This passage reads the meaning of §1367(d)—which applies to “any claim asserted under subsection (a)” of §1367—to hinge on the constitutional context. The Court is correct that Jinks and Raygor “hold that the single and unchanging disposition of §1367(d) … does not apply to claims against States.” Ante, at 12. But as the Court concedes, Jinks reached that holding only after analyzing whether the constitutional doubts at issue in Raygor applied to the county defendant. Ante, at 12, n. 6. The Court’s failure to do the same here cannot be reconciled with Jinks and Raygor: the Court should ask whether the constitutional concerns that justified the requirement of a clear statement in Zadvydas apply as well to inadmissible aliens.
The Court’s “lowest common denominator” principle is also in tension with Salinas v. United States, 522 U. S. 52 (1997). There, we rejected an argument that the federal bribery statute, 18 U. S. C. §666(a)(1)(B), should be construed to avoid constitutional doubts, in part on the ground that there was “no serious doubt about the constitutionality of §666(a)(1)(B) as applied to the facts of this case.” 522 U. S., at 60 (emphasis added). Unlike the Court’s approach to avoidance today, we disclaimed examination of the constitutionality of applications not before the Court: “Whatever might be said about §666(a)(1)(B)’s application in other cases, the application of §666(a)(1)(B) did not extend federal power beyond its proper bounds.” Id., at 61. The Court is mistaken that this passage in Salinas was a rejection of a constitutional argument on its merits. Ante, at 9, n. 5. Salinas, the petitioner, phrased his question presented solely in terms of the proper statutory interpretation of §666(a)(1)(B), Brief for Petitioner, O. T. 1996, No. 96–738, p. i, and never claimed that the statute was unconstitutional, see generally ibid.
C
More importantly, however, the Court’s “lowest common denominator” principle is inconsistent with the history of the canon of avoidance and is likely to have mischievous consequences. The modern canon of avoidance is a doctrine under which courts construe ambiguous statutes to avoid constitutional doubts, but this doctrine has its origins in a very different form of the canon. Traditionally, the avoidance canon was not a doctrine under which courts read statutes to avoid mere constitutional doubts. Instead, it commanded courts, when faced with two plausible constructions of a statute—one constitutional and the other unconstitutional—to choose the constitutional reading.[ Footnote 2 ] The traditional version of the canon thus requires courts to reach the issue whether the doubtful version of the statute is constitutional before adopting the construction that saves the statute from constitutional invalidity. A court faced with an ambiguous statute applies traditional avoidance by asking whether, given two plausible interpretations of that statute, one would be unconstitutional as applied to the plaintiff; and, if that interpretation is actually unconstitutional as applied to the plaintiff, the court picks the other (constitutional) reading. The court does not inquire whether either of the interpretations would be unconstitutional if applied to third parties not before the court, unless the challenge is facial or otherwise implicates third-party rights.
This history suggests that the “lowest common denominator” principle is mistaken. Courts applying the modern version of the canon of avoidance should no more look to the rights of third parties than do courts using the traditional version. Under modern avoidance, in other words, an ambiguous statute should be read to avoid a constitutional doubt only if the statute is constitutionally doubtful as applied to the litigant before the court (again, unless the constitutional challenge involves third-party rights). Yet the Court’s lowest common denominator principle allows a limiting construction of an ambiguous statute prompted by constitutional doubts to infect other applications of the statute—even if the statute raises no constitutional doubt as applied to the specific litigant in a given case and even if the constitutionally unproblematic application of the statute to the litigant is severable from the constitutionally dubious applications. The lowest common denominator principle thus allows an end run around black-letter constitutional doctrine governing facial and as-applied constitutional challenges to statutes: A litigant ordinarily cannot attack statutes as constitutionally invalid based on constitutional doubts concerning other litigants or factual circumstances.
The Court misses the point by answering that the canon of constitutional avoidance “is not a method of adjudicating constitutional questions by other means,” and that the canon rests on a presumption that “Congress did not intend the alternative which raises serious constitutional doubts.” Ante, at 10. That is true, but in deciding whether a plausible interpretation “raises serious constitutional doubts,” a court must employ the usual rules of constitutional adjudication. See ante, at 9 (noting that whether an interpretation is constitutionally doubtful turns on whether it raises “a multitude of constitutional problems”); Zadvydas, 533 U. S., at 690–696 (extensively employing constitutional analysis). Those rules include doctrines governing third-party constitutional challenges and the like. Moreover, the reason that courts perform avoidance at all, in any form, is that we assume “Congress intends statutes to have effect to the full extent the Constitution allows.” United States v. Booker, ante, at __ (Thomas, J., dissenting in part). Only my approach would extend §1231(a)(6) to its full constitutional bound consistent with modern avoidance, by narrowing the statute on a case-by-case basis only if constitutional concerns are actually present. By contrast, under the majority’s lowest common denominator principle, a statute like §1231(a)(6) must be narrowed once and for all based on constitutional concerns that may never materialize. In short, once narrowed in Zadvydas, §1231(a)(6) now limits the Executive’s power to detain unadmitted aliens—even though indefinite detention of unadmitted aliens may be perfectly constitutional.
All of this shows why the sole support the majority offers for its lowest common denominator principle can be squared with my analysis. That support is a plurality opinion of this Court (reaffirmed by footnote dictum in Leocal v. Ashcroft, ante, at ___, n. 8), that stated that the rule of lenity applies to statutes so long as they have some criminal applications. Ante, at 8 (citing United States v. Thompson/Center Arms Co., 504 U. S.505, 517 (1992)). To the extent that the rule of lenity is a constitutionally based clear statement rule, it is like vagueness doctrine, as its purpose is to ensure that those subjected to criminal prosecution have adequate notice of the conduct that the law prohibits. Cf., e.g., McBoyle v. United States, 283 U. S. 25, 27 (1931). Thompson/Center Arms is thus distinguishable, because our rules governing third-party challenges (rightly or wrongly) are more lenient in vagueness cases.[ Footnote 3 ] Zadvydas, by contrast, was a straightforward as-applied constitutional challenge. It concerned a constitutional doubt that arose from §1231(a)(6)’s application to Zadvydas himself, not its hypothetical application to other aliens, as its careful distinction between admitted and inadmissible aliens shows. To the extent that the rule of lenity is a nonconstitutionally based presumption about the interpretation of criminal statutes, the Thompson/Center Arms interpretive principle is fundamentally different from the canon of constitutional avoidance, because the rule of lenity is wholly independent of the rules governing constitutional adjudication. Either way, this case does not support the majority’s restatement of modern avoidance principles.
The cases at bar illustrate well the exception to the normal operation of as-applied constitutional adjudication that the Court’s approach creates. Congress explicitly provided that unconstitutional applications of §1231(a)(6) should be severed from constitutional applications.[ Footnote 4 ] Congress has thus indicated that courts should examine whether §1231(a)(6) raises a constitutional doubt application by application. After all, under the severability clause, if Zadvydas had held unconstitutional the indefinite detention of Zadvydas and Ho Ma, the constitutionality of the Secretary’s indefinite detention of Benitez and Martinez would remain an open question. Although Zadvydas did not formally hold §1231(a)(6) to be unconstitutional as applied to the aliens before it, the same procedure should be followed when analyzing whether §1231(a)(6) raises a constitutional doubt.[ Footnote 5 ] The Court today limits applications of §1231(a)(6) that may well be constitutional solely on the basis of constitutional doubts as to other applications, and despite that the severability clause contemplates application-by-application examination of the statute’s constitutionality.
The Court misapprehends my interpretive approach. It suggests that I would “spare [us] the necessity of ever finding a statute unconstitutional as applied,” ante, at 13, and “would render every statute a chameleon, its meaning subject to change depending on the presence or absence of constitutional concerns in each individual case,” ante, at 10. My approach does none of this. I simply would read ambiguous statutes to avoid as-applied constitutional doubts only if those doubts are present in the case before the Court. This leaves plenty of room for as-applied invalidation of statutes that are unambiguously unconstitutional. Nor would I permit a court to read every statute’s meaning to depend on constitutional concerns. That is permissible, in my view, only if the statute is ambiguous. Granted, I am thereby guilty of leaving courts free to interpret ambiguous statutes “as becoming inoperative when they ‘approach constitutional limits.’ ” Ante, at 13 . That is hardly an absurd result—unless one considers the modern canon of constitutional avoidance itself to be absurd. Every application of that canon, by rejecting a plausible interpretation of a statute, reads the statute to be inoperative to the extent it raises a constitutional doubt or “limit.”
In truth, the Court’s aggressive application of modern constitutional avoidance doctrine poses the greater danger. A disturbing number of this Court’s cases have applied the canon of constitutional doubt to statutes that were on their face clear. See, e.g., INS v. St. Cyr, 533 U. S. 289, 327–336 (2001) (Scalia, J., dissenting); Public Citizen v. Department of Justice, 491 U. S. 440, 481–482 (1989) (Kennedy, J., concurring in judgment); Lowe v. SEC, 472 U. S. 181, 212–213 (1985) (White, J., concurring in result). This Court and others may now employ the “lowest common denominator” approach to limit the application of statutes wholesale by searching for hypothetical unconstitutional applications of them—or, worse yet, hypothetical constitutional doubts —despite the absence of any facial constitutional problem (at least, so long as those hypothetical doubts pose “a multitude of constitutional problems,” ante, at 9). This is so even if Congress has expressed its clear intent that unconstitutional applications should be severed from constitutional applications, regardless of whether the challenger has third-party standing to raise the constitutional issue, and without the need to engage in full-fledged constitutional analysis.
This danger is real. In St. Cyr, this Court held that the Immigration and Nationality Act (INA) did not divest district courts of jurisdiction under 28 U. S. C. §2241 over habeas actions filed by criminal aliens to challenge removal orders, 533 U. S., at 314. The Court did so because it thought that otherwise the statute would preclude any avenue of judicial review of removal orders of criminal aliens, thus raising a serious Suspension Clause question. Id., at 305. This was a construction of (among other provisions) 8 U. S. C. §§1252(a)(1) and 1252(b)(9), and 28 U. S. C. §2241, none of which distinguishes between criminal and noncriminal aliens. 533 U. S., at 308–314. The INA, however, clearly allows noncriminal aliens, unlike criminal aliens, a right to judicial review of removal decisions in the courts of appeals under the review provisions of §1252(a)(1), and St. Cyr involved only criminal aliens. After St. Cyr, therefore, one would have thought that “noncriminal aliens seeking to challenge their removal orders… [would] still presumably be required to proceed directly to the court of appeals by way of petition for review, under the restrictive modified Hobbs Act review provisions set forth in §1252(a)(1),” rather than sue directly under the habeas statute. Id., at 335 (Scalia, J., joined by Rehnquist, C. J., and O’Connor and Thomas, JJ., dissenting). Yet lower courts, relying on a version of the Court’s “lowest common denominator” principle, have held just the opposite: They have entertained noncriminal aliens’ habeas actions challenging removal orders. Chmakov v. Blackman, 266 F. 3d 210, 214–215 (CA3 2001); see also Riley v. INS, 310 F. 3d 1253, 1256 (CA10 2002); Liu v. INS, 293 F. 3d 36, 38–41 (CA2 2002). The logic in allowing noncriminal aliens, who have a right to judicial review of removal decisions, to take advantage of constitutional doubt that arises from precluding any avenue of judicial review for criminal aliens, see St. Cyr, supra, at 305, escapes me.
II
The Court is also mistaken in affording Zadvydas stare decisis effect. Zadvydas was wrong in both its statutory and its constitutional analysis for the reasons expressed well by the dissents in that case. See 533 U. S., at 705–718 (opinion of Kennedy, J.); id., at 702–705 (opinion of Scalia, J.). I continue to adhere to those views and will not repeat the analysis of my colleagues. I write only to explain why I do not consider Zadvydas to bind us.
Zadvydas cast itself as a statutory case, but that fact should not prevent us from overruling it. It is true that we give stronger stare decisis effect to our holdings in statutory cases than in constitutional cases. See, e.g., Hilton v. South Carolina Public Railways Comm’n, 502 U. S. 197, 205 (1991). This rule, however, is not absolute, and we should not hesitate to allow our precedent to yield to the true meaning of an Act of Congress when our statutory precedent is “unworkable” or “badly reasoned.” Holder v. Hall, 512 U. S. 874, 936 (1994) (Thomas, J., concurring in judgment) (quoting Payne v. Tennessee, 501 U. S. 808, 827 (1991) (internal quotation marks omitted)). “[W]e have never applied stare decisis mechanically to prohibit overruling our earlier decisions determining the meaning of statutes.” Monell v. New York City Dept. of Social Servs., 436 U. S. 658, 695 (1978). The mere fact that Congress can overturn our cases by statute is no excuse for failing to overrule a statutory precedent of ours that is clearly wrong, for the realities of the legislative process often preclude readopting the original meaning of a statute that we have upset.
Zadvydas ’ reading of §1231(a)(6) is untenable. Section 1231(a)(6) provides that aliens whom the Secretary of Homeland Security has ordered removed “may be detained beyond the removal period.” There is no qualification to this authorization, and no reference to a “reasonable time” limitation. Just as we exhaust the aid of the “traditional tools of statutory construction,” Chevron U. S. A. Inc. v. Natural Resources Defense Council, Inc., 467 U. S. 837, 843, n. 9 (1984), before deferring to an agency’s interpretation of a statute, so too should we exhaust those tools before deciding that a statute is ambiguous and that an alternative plausible construction of the statute should be adopted.
Application of those traditional tools begins and ends with the text of §1231(a)(6). Zadvydas ’ observation that “if Congress had meant to authorize long-term detention of unremovable aliens, it certainly could have spoken in clearer terms,” 533 U. S., at 697, proves nothing. Congress could have spoken more clearly in any statutory case in which the statute does not mention the particular factual scenario before the Court. Congress provided for a “reasonable time” limit to detentions pending removal in other portions of §1231. Id., at 708 (Kennedy, J., dissenting). Its failure to do the same in §1231(a)(6) confirms what is unmistakable from its terms: that there is no time limit on the Secretary’s power to detain aliens. There is no textually evident alternative construction that would avoid the constitutional doubts identified by the majority.
Even apart from the Court’s incredible reading of §1231(a)(6), the normal reason for affording our statutory holdings strong stare decisis effect—that Congress is free to overrule them if it disagrees—does not apply to Zadvydas. Zadvydas is a statutory case in name only. Although the Zadvydas majority purported to find indefinite detention only constitutionally doubtful, its lengthy analysis strongly signaled to Congress that indefinite detention of admitted aliens would be unconstitutional. Indeed, far from avoiding that constitutional question in Zadvydas, the Court took it head on, giving it extended treatment. Id., at 690–697; but see ante, at 10 (noting the “fundamenta[l]” tenet that “[t]he canon [of constitutional avoidance] is not a method of adjudicating constitutional questions by other means”). Zadvydas makes clear that the Court thought indefinite detention to be more than constitutionally suspect, and there is evidence that some Members of Congress understood as much.[ Footnote 6 ] This is why the Court’s assurance that if “the security of our borders will be compromised if [the United States] must release into the country inadmissible aliens who cannot be removed…. Congress can attend to it,” ante, at 14, rings hollow. Short of constitutional amendment, it is only within the power of this Court to correct Zadvydas ’ error.
The Court points to 8 U. S. C. §1226a(a)(6) (2000 ed., Supp. II), a statute that Congress passed shortly after Zadvydas, as evidence that Congress can correct Zadvydas ’ mistake. Ante, at 14–15, n. 8. This statute only confirms my concern that Zadvydas is legislatively uncorrectable. Section 1226a(a)(6) authorizes detention for a period of six months beyond the removal period of aliens who present a national security threat, but only to the extent that those aliens’ removal is not reasonably foreseeable. Ante, at 14–15, n. 8. Yet Zadvydas conceded that indefinite detention might not violate due process in “certain special and narrow nonpunitive circumstances … where a special justification, such as harm-threatening mental illness, outweighs the individual’s constitutionally protected interest in avoiding physical restraint.” 533 U. S., at 690 (internal quotation marks and citations omitted). Moreover, Zadvydas set a 6-month presumptive outer limit on the detention power. Id., at 701. Congress crafted §1226a(a)(6) to operate within the boundaries Zadvydas set. This provision says nothing about whether Congress may authorize detention of aliens for greater lengths of time or for reasons the Court found constitutionally problematic in Zadvydas.
***
For the foregoing reasons, I would affirm the judgment of the Eleventh Circuit and reverse the judgment of the Ninth Circuit. I therefore respectfully dissent.
Notes
Footnote 1
Section 1367(d) provides that “[t]he period of limitations for any claim asserted under [§1367(a)], and for any other claim in the same action that is voluntarily dismissed at the same time as or after the dismissal of the claim under [§1367(a)], shall be tolled while the claim is pending and for a period of 30 days after it is dismissed unless State law provides for a longer tolling period.”
Footnote 2
See Rust v. Sullivan, 500 U. S. 173, 190–191 (1991) (distinguishing the classic and modern versions of the canon and citing cases); Hooper v. California, 155 U. S. 648, 657 (1895) (“The elementary rule is that every reasonable construction must be resorted to, in order to save a statute from unconstitutionality”); Mossman v. Higginson, 4 Dall. 12, 14 (1800) (reasoning that the statute under review “can, and must, receive a construction, consistent with the constitution”); Ex parte Randolph, 20 F. Cas. 242, 254 (No. 11,558) (CC Va. 1833) (Marshall, J.); Vermeule, Saving Constructions, 85 Geo. L. J. 1945, 1949 (1997); H. Black, Handbook on the Construction and Interpretation of the Laws 113–114 (2d ed. 1911). The modern version seems to have originated in United States ex rel. Attorney General v. Delaware & Hudson Co., 213 U. S. 366, 408 (1909).
Footnote 3
See, e.g., Chicago v. Morales, 527 U. S. 41, 55, and n. 22 (1999) (plurality opinion); Kolender v. Lawson, 461 U. S. 352, 358–359, n. 8 (1983); Papachristou v. Jacksonville, 405 U. S. 156 (1972).
Footnote 4
“If any provision of this division … or the application of such provision to any person or circumstances is held to be unconstitutional, the remainder of this division and the application of the provisions of this division to any person or circumstance shall not be affected thereby.” Note following 8 U. S. C. §1101, p. 840 (Separability).
Footnote 5
Crowell v. Benson, 285 U. S. 22 (1932), bolsters my approach. Employing the canon of avoidance, the Court construed a statute in that case to allow judicial review of jurisdictional facts but not legislative facts. It did so even though the terms of the statute itself did not distinguish between the two sorts of facts. Id., at 62–63. The presence of a severability provision in the statute gave “assurance that there [was] no violation of the purpose of the Congress in sustaining the determinations of fact of the deputy commissioner where he acts within his authority in passing upon compensation claims while denying finality to his conclusions as to the jurisdictional facts upon which the valid application of the statute depends.” Ibid. So too here, the presence of a severability provision should reassure the Court that applying Zadvydas ’ limiting construction of §1231(a)(6) to some aliens and not others is consistent with the statute.
Footnote 6
See H. R. Conf. Rep. No. 108–10, p. 600 (2003) (“A recent Supreme Court decision held that criminal aliens cannot be detained indefinitely,” no doubt referring to Zadvydas ); H. R. Rep. No. 108–724, pt. 5, p. 191 (2004) (“The danger posed by the requirement that these aliens be allowed to remain in the U. S. was increased exponentially by the 2001 Supreme Court decision of Zadvydas v. Davis, in which the Court made clear that it would strike down as unconstitutional the indefinite detention by [the Secretary] of aliens with removal orders whose countries will not take them back, except in the most narrow of circumstances” (footnotes omitted)); 147 Cong. Rec. S11047 (Oct. 25, 2001) (“Indefinite detention of aliens is permitted only in extraordinary circumstances,” citing Zadvydas ).
Justice O’Connor delivered the opinion of the Court.
The California Department of Corrections (CDC) has an unwritten policy of racially segregating prisoners in double cells in reception centers for up to 60 days each time they enter a new correctional facility. We consider whether strict scrutiny is the proper standard of review for an equal protection challenge to that policy.
I
A
CDC institutions house all new male inmates and all male inmates transferred from other state facilities in reception centers for up to 60 days upon their arrival. During that time, prison officials evaluate the inmates to determine their ultimate placement. Double-cell assignments in the reception centers are based on a number of factors, predominantly race. In fact, the CDC has admitted that the chances of an inmate being assigned a cellmate of another race are “ ‘[p]retty close’ ” to zero percent. App. to Pet. for Cert. 3a. The CDC further subdivides prisoners within each racial group. Thus, Japanese-Americans are housed separately from Chinese-Americans, and Northern California Hispanics are separated from Southern California Hispanics.
The CDC’s asserted rationale for this practice is that it is necessary to prevent violence caused by racial gangs. Brief for Respondents 1–6. It cites numerous incidents of racial violence in CDC facilities and identifies five major prison gangs in the State: Mexican Mafia, Nuestra Familia, Black Guerilla Family, Aryan Brotherhood, and Nazi Low Riders. Id., at 2. The CDC also notes that prison-gang culture is violent and murderous. Id., at 3. An associate warden testified that if race were not considered in making initial housing assignments, she is certain there would be racial conflict in the cells and in the yard. App. 215a. Other prison officials also expressed their belief that violence and conflict would result if prisoners were not segregated. See, e.g., id., at 305a–306a. The CDC claims that it must therefore segregate all inmates while it determines whether they pose a danger to others. See Brief for Respondents 29.
With the exception of the double cells in reception areas, the rest of the state prison facilities—dining areas, yards, and cells—are fully integrated. After the initial 60-day period, prisoners are allowed to choose their own cellmates. The CDC usually grants inmate requests to be housed together, unless there are security reasons for denying them.
B
Garrison Johnson is an African-American inmate in the custody of the CDC. He has been incarcerated since 1987 and, during that time, has been housed at a number of California prison facilities. Fourth Amended Complaint 3, Record, Doc. No. 78. Upon his arrival at Folsom prison in 1987, and each time he was transferred to a new facility thereafter, Johnson was double-celled with another African-American inmate. See ibid.
Johnson filed a complaint pro se in the United States District Court for the Central District of California on February 24, 1995, alleging that the CDC’s reception-center housing policy violated his right to equal protection under the Fourteenth Amendment by assigning him cellmates on the basis of his race. He alleged that, from 1987 to 1991, former CDC Director James Rowland instituted and enforced an unconstitutional policy of housing inmates according to race. Second Amended Complaint 2–4, Record, Doc. No. 21. Johnson made the same allegations against former Director James Gomez for the period from 1991 until the filing of his complaint. Ibid. The District Court dismissed his complaint for failure to state a claim. The Court of Appeals for the Ninth Circuit reversed and remanded, holding that Johnson had stated a claim for racial discrimination in violation of the Equal Protection Clause of the Fourteenth Amendment. Johnson v. California, 207 F. 3d 650, 655 (2000).
On remand, Johnson was appointed counsel and granted leave to amend his complaint. On July 5, 2000, he filed his Fourth Amended Complaint. Record, Doc. No. 81. Johnson claimed that the CDC’s policy of racially segregating all inmates in reception-center cells violated his rights under the Equal Protection Clause. Johnson sought damages, alleging that former CDC Directors Rowland and Gomez, in their individual capacities, violated his constitutional rights by formulating and implementing the CDC’s housing policy. He also sought injunctive relief against former CDC Director Stephen Cambra.
Johnson has consistently challenged, and the CDC has consistently defended, the policy as a whole—as it relates to both new inmates and inmates transferred from other facilities. Johnson was first segregated in 1987 as a new inmate when he entered the CDC facility at Folsom. Since 1987, he has been segregated each time he has been transferred to a new facility. Thus, he has been subject to the CDC’s policy both as a new inmate and as an inmate transferred from one facility to another.
After discovery, the parties moved for summary judgment. The District Court granted summary judgment to the defendants on grounds that they were entitled to qualified immunity because their conduct was not clearly unconstitutional. The Court of Appeals for the Ninth Circuit affirmed. 321 F. 3d 791 (2003). It held that the constitutionality of the CDC’s policy should be reviewed under the deferential standard we articulated in Turner v. Safley, 482 U. S. 78 (1987)—not strict scrutiny. 321 F. 3d, at 798–799. Applying Turner, it held that Johnson had the burden of refuting the “common-sense connection” between the policy and prison violence. 321 F. 3d, at 802. Though it believed this was a “close case,” id., at 798, the Court of Appeals concluded that the policy survived Turner ’s deferential standard, 321 F. 3d, at 807.
The Court of Appeals denied Johnson’s petition for rehearing en banc. Judge Ferguson, joined by three others, dissented on grounds that “[t]he panel’s decision ignore[d] the Supreme Court’s repeated and unequivocal command that all racial classifications imposed by the government must be analyzed by a reviewing court under strict scrutiny, and fail[ed] to recognize that [the] Turner analysis is inapplicable in cases, such as this one, in which the right asserted is not inconsistent with legitimate penological objectives.” 336 F. 3d 1117 (2003) (internal quotation marks and citations omitted). We granted certiorari to decide which standard of review applies. 540 U. S. 1217 (2004).
II
A
We have held that “ all racial classifications [imposed by government] … must be analyzed by a reviewing court under strict scrutiny.” Adarand Constructors, Inc. v. Peńa, 515 U. S. 200, 227 (1995) (emphasis added). Under strict scrutiny, the government has the burden of proving that racial classifications “are narrowly tailored measures that further compelling governmental interests.” Ibid. We have insisted on strict scrutiny in every context, even for so-called “benign” racial classifications, such as race-conscious university admissions policies, see Grutter v. Bollinger, 539 U. S. 306, 326 (2003), race-based preferences in government contracts, see Adarand, supra, at 226, and race-based districting intended to improve minority representation, see Shaw v. Reno, 509 U. S. 630, 650 (1993).
The reasons for strict scrutiny are familiar. Racial classifications raise special fears that they are motivated by an invidious purpose. Thus, we have admonished time and again that, “[a]bsent searching judicial inquiry into the justification for such race-based measures, there is simply no way of determining … what classifications are in fact motivated by illegitimate notions of racial inferiority or simple racial politics.” Richmond v. J. A. Croson Co., 488 U. S. 469, 493 (1989) (plurality opinion). We therefore apply strict scrutiny to all racial classifications to “ ‘smoke out’ illegitimate uses of race by assuring that [government] is pursuing a goal important enough to warrant use of a highly suspect tool.” Ibid.[ Footnote 1 ]
The CDC claims that its policy should be exempt from our categorical rule because it is “neutral”—that is, it “neither benefits nor burdens one group or individual more than any other group or individual.” Brief for Respondents 16. In other words, strict scrutiny should not apply because all prisoners are “equally” segregated. The CDC’s argument ignores our repeated command that “racial classifications receive close scrutiny even when they may be said to burden or benefit the races equally.” Shaw, supra, at 651. Indeed, we rejected the notion that separate can ever be equal—or “neutral”—50 years ago in Brown v. Board of Education, 347 U. S. 483 (1954), and we refuse to resurrect it today. See also Powers v . Ohio, 499 U. S. 400, 410 (1991) (rejecting the argument that race-based peremptory challenges were permissible because they applied equally to white and black jurors and holding that “[i]t is axiomatic that racial classifications do not become legitimate on the assumption that all persons suffer them in equal degree”).
We have previously applied a heightened standard of review in evaluating racial segregation in prisons. In Lee v. Washington, 390 U. S. 333 (1968) (per curiam), we upheld a three-judge court’s decision striking down Alabama’s policy of segregation in its prisons. Id., at 333–334. Alabama had argued that desegregation would undermine prison security and discipline, id., at 334, but we rejected that contention. Three Justices concurred “to make explicit something that is left to be gathered only by implication from the Court’s opinion”—“that prison authorities have the right, acting in good faith and in particularized circumstances, to take into account racial tensions in maintaining security, discipline, and good order in prisons and jails.” Ibid. (emphasis added). The concurring Justices emphasized that they were “unwilling to assume that state or local prison authorities might mistakenly regard such an explicit pronouncement as evincing any dilution of this Court’s firm commitment to the Fourteenth Amendment’s prohibition of racial discrimination.” Ibid.
The need for strict scrutiny is no less important here, where prison officials cite racial violence as the reason for their policy. As we have recognized in the past, racial classifications “threaten to stigmatize individuals by reason of their membership in a racial group and to incite racial hostility.” Shaw, supra, at 643 (citing J. A. Croson Co., supra, at 493 (plurality opinion) (emphasis added)). Indeed, by insisting that inmates be housed only with other inmates of the same race, it is possible that prison officials will breed further hostility among prisoners and reinforce racial and ethnic divisions. By perpetuating the notion that race matters most, racial segregation of inmates “may exacerbate the very patterns of [violence that it is] said to counteract.” Shaw, supra, at 648; see also Trulson & Marquart, The Caged Melting Pot: Toward an Understanding of the Consequences of Desegregation in Prisons, 36 Law & Soc. Rev. 743, 774 (2002) (in a study of prison desegregation, finding that “over [10 years] the rate of violence between inmates segregated by race in double cells surpassed the rate among those racially integrated”). See also Brief for Former State Corrections Officials as Amici Curiae 19 (opinion of former corrections officials from six States that “racial integration of cells tends to diffuse racial tensions and thus diminish interracial violence” and that “a blanket policy of racial segregation of inmates is contrary to sound prison management”).
The CDC’s policy is unwritten. Although California claimed at oral argument that two other States follow a similar policy, see Tr. of Oral Arg. 30–31, this assertion was unsubstantiated, and we are unable to confirm or deny its accuracy.[ Footnote 2 ] Virtually all other States and the Federal Government manage their prison systems without reliance on racial segregation. See Brief for United States as Amicus Curiae 24. Federal regulations governing the Federal Bureau of Prisons (BOP) expressly prohibit racial segregation. 28 CFR §551.90 (2004) (“[BOP] staff shall not discriminate against inmates on the basis of race, religion, national origin, sex, disability, or political belief. This includes the making of administrative decisions and providing access to work, housing and programs”). The United States contends that racial integration actually “leads to less violence in BOP’s institutions and better prepares inmates for re-entry into society.” Brief for United States as Amicus Curiae 25. Indeed, the United States argues, based on its experience with the BOP, that it is possible to address “concerns of prison security through individualized consideration without the use of racial segregation, unless warranted as a necessary and temporary response to a race riot or other serious threat of race-related violence.” Id., at 24. As to transferees, in particular, whom the CDC has already evaluated at least once, it is not clear why more individualized determinations are not possible.
Because the CDC’s policy is an express racial classification, it is “immediately suspect.” Shaw, 509 U. S., at 642; see also Washington v. Seattle School Dist. No. 1, 458 U. S. 457, 485 (1982). We therefore hold that the Court of Appeals erred when it failed to apply strict scrutiny to the CDC’s policy and to require the CDC to demonstrate that its policy is narrowly tailored to serve a compelling state interest.
B
The CDC invites us to make an exception to the rule that strict scrutiny applies to all racial classifications, and instead to apply the deferential standard of review articulated in Turner v. Safley, 482 U. S. 78 (1987), because its segregation policy applies only in the prison context. We decline the invitation. In Turner, we considered a claim by Missouri prisoners that regulations restricting inmate marriages and inmate-to-inmate correspondence were unconstitutional. Id., at 81. We rejected the prisoners’ argument that the regulations should be subject to strict scrutiny, asking instead whether the regulation that burdened the prisoners’ fundamental rights was “reasonably related” to “legitimate penological interests.” Id., at 89.
We have never applied Turner to racial classifications. Turner itself did not involve any racial classification, and it cast no doubt on Lee. We think this unsurprising, as we have applied Turner ’s reasonable-relationship test only to rights that are “inconsistent with proper incarceration.” Overton v. Bazzetta, 539 U. S. 126, 131 (2003); see also Pell v. Procunier, 417 U. S. 817, 822 (1974) (“[A] prison inmate retains those First Amendment rights that are not inconsistent with his status as a prisoner or with the legitimate penological objectives of the corrections system”). This is because certain privileges and rights must necessarily be limited in the prison context. See O’Lone v. Estate of Shabazz, 482 U. S. 342, 348 (1987) (“ ‘[L]awful incarceration brings about the necessary withdrawal or limitation of many privileges and rights, a retraction justified by the considerations underlying our penal system’ ” (quoting Price v. Johnston, 334 U. S.266, 285 (1948))). Thus, for example, we have relied on Turner in addressing First Amendment challenges to prison regulations, including restrictions on freedom of association, Overton, supra; limits on inmate correspondence, Shaw v. Murphy, 532 U. S. 223 (2001); restrictions on inmates’ access to courts, Lewis v. Casey, 518 U. S. 343 (1996); restrictions on receipt of subscription publications, Thornburgh v. Abbott, 490 U. S. 401 (1989); and work rules limiting prisoners’ attendance at religious services, Shabazz, supra. We have also applied Turner to some due process claims, such as involuntary medication of mentally ill prisoners, Washington v. Harper, 494 U. S. 210 (1990); and restrictions on the right to marry, Turner, supra.
The right not to be discriminated against based on one’s race is not susceptible to the logic of Turner. It is not a right that need necessarily be compromised for the sake of proper prison administration. On the contrary, compliance with the Fourteenth Amendment’s ban on racial discrimination is not only consistent with proper prison administration, but also bolsters the legitimacy of the entire criminal justice system. Race discrimination is “especially pernicious in the administration of justice.” Rose v. Mitchell, 443 U. S. 545, 555 (1979). And public respect for our system of justice is undermined when the system discriminates based on race. Cf. Batson v. Kentucky, 476 U. S. 79, 99 (1986) (“[P]ublic respect for our criminal justice system and the rule of law will be strengthened if we ensure that no citizen is disqualified from jury service because of his race”). When government officials are permitted to use race as a proxy for gang membership and violence without demonstrating a compelling government interest and proving that their means are narrowly tailored, society as a whole suffers. For similar reasons, we have not used Turner to evaluate Eighth Amendment claims of cruel and unusual punishment in prison. We judge violations of that Amendment under the “deliberate indifference” standard, rather than Turner ’s “reasonably related” standard. See Hope v. Pelzer, 536 U. S. 730, 738 (2002) (asking whether prison officials displayed “ ‘deliberate indifference’ to the inmate’s health or safety” where an inmate claimed that they violated his rights under the Eighth Amendment (quoting Hudson v. McMillian, 503 U. S. 1, 8 (1992))). This is because the integrity of the criminal justice system depends on full compliance with the Eighth Amendment. See Spain v. Procunier, 600 F. 2d 189, 193–194 (CA9 1979) (Kennedy, J.) (“[T]he full protections of the eighth amendment most certainly remain in force [in prison]. The whole point of the amendment is to protect persons convicted of crimes. … Mechanical deference to the findings of state prison officials in the context of the eighth amendment would reduce that provision to a nullity in precisely the context where it is most necessary” (internal quotation marks omitted)).
In the prison context, when the government’s power is at its apex, we think that searching judicial review of racial classifications is necessary to guard against invidious discrimination. Granting the CDC an exemption from the rule that strict scrutiny applies to all racial classifications would undermine our “unceasing efforts to eradicate racial prejudice from our criminal justice system.” McCleskey v. Kemp, 481 U. S. 279, 309 (1987) (internal quotation marks omitted).
The CDC argues that “[d]eference to the particular expertise of prison officials in the difficult task of managing daily prison operations” requires a more relaxed standard of review for its segregation policy. Brief for Respondents 18. But we have refused to defer to state officials’ judgments on race in other areas where those officials traditionally exercise substantial discretion. For example, we have held that, despite the broad discretion given to prosecutors when they use their peremptory challenges, using those challenges to strike jurors on the basis of their race is impermissible. See Batson, supra, at 89–96. Similarly, in the redistricting context, despite the traditional deference given to States when they design their electoral districts, we have subjected redistricting plans to strict scrutiny when States draw district lines based predominantly on race. Compare generally Vieth v. Jubilier, 541 U. S. 267 (2004) (partisan gerrymandering), with Shaw v. Reno, 509 U. S. 630 (1993) (racial gerrymandering).
We did not relax the standard of review for racial classifications in prison in Lee, and we refuse to do so today. Rather, we explicitly reaffirm what we implicitly held in Lee: The “necessities of prison security and discipline,” 390 U. S., at 334, are a compelling government interest justifying only those uses of race that are narrowly tailored to address those necessities. See Grutter, 539 U. S., at 353 (Thomas, J., concurring in part and dissenting in part) (citing Lee for the principle that “protecting prisoners from violence might justify narrowly tailored racial discrimination”); J. A. Croson Co., 488 U. S., at 521 (Scalia, J., concurring) (citing Lee for the proposition that “only a social emergency rising to the level of imminent danger to life or limb—for example, a prison race riot, requiring temporary segregation of inmates—can justify an exception to the principle embodied in the Fourteenth Amendment that ‘[o]ur Constitution is color-blind, and neither knows nor tolerates classes among citizens’ ” (quoting Plessy v. Ferguson, 163 U. S. 537, 559 (1896) (Harlan, J., dissenting))); see also Pell, 417 U. S., at 823 (“[C]entral to all other corrections goals is the institutional consideration of internal security within the correctional facilities themselves”).
Justice Thomas would subject race-based policies in prisons to Turner ’s deferential standard of review because, in his view, judgments about whether race-based policies are necessary “are better left in the first instance to the officials who run our Nation’s prisons.” Post, at 20. But Turner is too lenient a standard to ferret out invidious uses of race. Turner requires only that the policy be “reasonably related” to “legitimate penological interests.” 482 U. S., at 89. Turner would allow prison officials to use race-based policies even when there are race-neutral means to accomplish the same goal, and even when the race-based policy does not in practice advance that goal. See, e.g., 321 F. 3d, at 803 (case below) (reasoning that, under Turner, the Court of Appeals did “not have to agree that the policy actually advances the CDC’s legitimate interest, but only [that] ‘defendants might reasonably have thought that the policy would advance its interests’ ”). See also Turner, supra, at 90 (warning that Turner is not a “least restrictive alternative test” (internal quotation marks omitted)).
For example, in Justice Thomas’ world, prison officials could segregate visiting areas on the ground that racial mixing would cause unrest in the racially charged prison atmosphere. Under Turner, “[t]he prisoner would have to prove that there would not be a riot. [But] [i]t is certainly ‘plausible’ that such a riot could ensue: our society, as well as our prisons, contains enough racists that almost any interracial interaction could potentially lead to conflict.” 336 F. 3d, at 1120 (case below) (Ferguson, J., dissenting from denial of rehearing en banc). Indeed, under Justice Thomas’ view, there is no obvious limit to permissible segregation in prisons. It is not readily apparent why, if segregation in reception centers is justified, segregation in the dining halls, yards, and general housing areas is not also permissible. Any of these areas could be the potential site of racial violence. If Justice Thomas’ approach were to carry the day, even the blanket segregation policy struck down in Lee might stand a chance of survival if prison officials simply asserted that it was necessary to prison management. We therefore reject the Turner standard for racial classifications in prisons because it would make rank discrimination too easy to defend.
The CDC protests that strict scrutiny will handcuff prison administrators and render them unable to address legitimate problems of race-based violence in prisons. See also post, at 9, 24–25 (Thomas, J., dissenting). Not so. Strict scrutiny is not “strict in theory, but fatal in fact.” Adarand, 515 U. S., at 237 (internal quotation marks omitted); Grutter, 539 U. S., at 326–327 (“Although all governmental uses of race are subject to strict scrutiny, not all are invalidated by it”). Strict scrutiny does not preclude the ability of prison officials to address the compelling interest in prison safety. Prison administrators, however, will have to demonstrate that any race-based policies are narrowly tailored to that end. See id., at 327 (“When race-based action is necessary to further a compelling governmental interest, such action does not violate the constitutional guarantee of equal protection so long as the narrow-tailoring requirement is also satisfied”).[ Footnote 3 ]
The fact that strict scrutiny applies “says nothing about the ultimate validity of any particular law; that determination is the job of the court applying strict scrutiny.” Adarand, supra, at 229–230. At this juncture, no such determination has been made. On remand, the CDC will have the burden of demonstrating that its policy is narrowly tailored with regard to new inmates as well as transferees. Prisons are dangerous places, and the special circumstances they present may justify racial classifications in some contexts. Such circumstances can be considered in applying strict scrutiny, which is designed to take relevant differences into account.
III
We do not decide whether the CDC’s policy violates the Equal Protection Clause. We hold only that strict scrutiny is the proper standard of review and remand the case to allow the Court of Appeals for the Ninth Circuit, or the District Court, to apply it in the first instance. See Consolidated Rail Corporation v. Gottshal, 512 U. S. 532, 557–558 (1994) (reversing and remanding for the lower court to apply the correct legal standard in the first instance); Lucas v. South Carolina Coastal Council, 505 U. S. 1003, 1031–1032 (1992) (same). The judgment of the Court of Appeals is reversed, and the case is remanded for further proceedings consistent with this opinion.
It is so ordered.
The Chief Justice took no part in the decision of this case.
Notes
Footnote 1
Justice Thomas takes a hands-off approach to racial classifications in prisons, suggesting that a “compelling showing [is] needed to overcome the deference we owe to prison administrators.” Post, at 21 (dissenting opinion). But such deference is fundamentally at odds with our equal protection jurisprudence. We put the burden on state actors to demonstrate that their race-based policies are justified.
Footnote 2
Though, as Justice Thomas points out, see post, at 22–23, and n. 12, inmates in reception centers in Oklahoma and Texas “are not generally assigned randomly to racially integrated cells,” it is also the case that “these inmates are not precluded from integrated cell assignments.” Oklahoma Dept. of Corrections, Policies and Procedures, Operations Memorandum No. OP–030102, Inmate Housing (Sept. 16, 2004), available at http://www.doc.state.ok.us/docs/policies.htm (as visited Jan. 21, 2005, and available in the Clerk of Court’s case file); Texas Dept. of Criminal Justice, Security Memorandum No. SM–01.28, Assignment to General Population Two-Person Cells (June 15, 2002). See also Brief for Former State Corrections Officials as Amici Curiae 20, n. 10 (“To the extent that race is considered in the assignment calculus in Oklahoma, it appears to be one factor among many, and as a result, individualized consideration is given to all inmates”). We therefore have no way of knowing whether, in practice, inmates in Oklahoma and Texas, like those in California, have close to no chance, App. to Pet. for Cert. 3a, of being celled with a person of a different race. See also Brief for Former State Corrections Officials as Amici Curiae 19–20 (“[W]e are aware of no state other than California that assumes that every incoming prisoner is incapable of getting along with a cell mate of a different race. And we are aware of no state other than California that has acted on such an assumption by adopting an inflexible and absolute policy of racial segregation of double cells in reception centers”).
Footnote 3
Justice Thomas characterizes the CDC’s policy as a “limited” one, see post, at 2, but the CDC’s policy is in fact sweeping in its application. It applies to all prisoners housed in double cells in reception centers, whether newly admitted or transferred from one facility to another. Moreover, despite Justice Thomas’ suggestion that the CDC considers other nonracial factors in determining housing placements, the CDC itself has admitted that, in practice, there is a “ ‘[p]retty close’ ” to zero percent chance that an inmate will be housed with a person of a different race. App. to Pet. for Cert. 3a. See also generally post, at 1–2, and n. 1 (Stevens, J., dissenting). Thus, despite an inmate’s “age, physical size, mental health, medical needs, [and] criminal history,” post, at 13 (Thomas, J., dissenting), the fact that he is black categorically precludes him from being celled with a white inmate. As we explain, see infra, at 15, we do not decide whether the threat of violence in California prisons is sufficient to justify such a broad policy.
Justice O’Connor, dissenting.
The Court’s decision today establishes a categorical rule forbidding the execution of any offender for any crime committed before his 18th birthday, no matter how deliberate, wanton, or cruel the offense. Neither the objective evidence of contemporary societal values, nor the Court’s moral proportionality analysis, nor the two in tandem suffice to justify this ruling.
Although the Court finds support for its decision in the fact that a majority of the States now disallow capital punishment of 17-year-old offenders, it refrains from asserting that its holding is compelled by a genuine national consensus. Indeed, the evidence before us fails to demonstrate conclusively that any such consensus has emerged in the brief period since we upheld the constitutionality of this practice in Stanford v. Kentucky, 492 U. S. 361 (1989).
Instead, the rule decreed by the Court rests, ultimately, on its independent moral judgment that death is a disproportionately severe punishment for any 17-year-old offender. I do not subscribe to this judgment. Adolescents as a class are undoubtedly less mature, and therefore less culpable for their misconduct, than adults. But the Court has adduced no evidence impeaching the seemingly reasonable conclusion reached by many state legislatures: that at least some 17-year-old murderers are sufficiently mature to deserve the death penalty in an appropriate case. Nor has it been shown that capital sentencing juries are incapable of accurately assessing a youthful defendant’s maturity or of giving due weight to the mitigating characteristics associated with youth.
On this record—and especially in light of the fact that so little has changed since our recent decision in Stanford —I would not substitute our judgment about the moral propriety of capital punishment for 17-year-old murderers for the judgments of the Nation’s legislatures. Rather, I would demand a clearer showing that our society truly has set its face against this practice before reading the Eighth Amendment categorically to forbid it.
I
A
Let me begin by making clear that I agree with much of the Court’s description of the general principles that guide our Eighth Amendment jurisprudence. The Amendment bars not only punishments that are inherently “ ‘barbaric,’ ” but also those that are “ ‘ excessive’ in relation to the crime committed. ” Coker v. Georgia, 433 U. S. 584, 592 (1977) (plurality opinion). A sanction is therefore beyond the state’s authority to inflict if it makes “no measurable contribution” to acceptable penal goals or is “grossly out of proportion to the severity of the crime.” Ibid. The basic “precept of justice that punishment for crime should be … proportioned to [the] offense,” Weems v. United States, 217 U. S. 349, 367 (1910), applies with special force to the death penalty. In capital cases, the Constitution demands that the punishment be tailored both to the nature of the crime itself and to the defendant’s “personal responsibility and moral guilt.” Enmund v. Florida, 458 U. S. 782, 801 (1982); see also id., at 825 (O’Connor, J., dissenting); Tison v. Arizona, 481 U. S. 137, 149 (1987); Eddings v. Oklahoma, 455 U. S. 104, 111–112 (1982).
It is by now beyond serious dispute that the Eighth Amendment’s prohibition of “cruel and unusual punishments” is not a static command. Its mandate would be little more than a dead letter today if it barred only those sanctions—like the execution of children under the age of seven—that civilized society had already repudiated in 1791. See ante, at 1 (Stevens, J., concurring); cf. Stanford, supra, at 368 (discussing the common law rule at the time the Bill of Rights was adopted). Rather, because “[t]he basic concept underlying the Eighth Amendment is nothing less than the dignity of man,” the Amendment “must draw its meaning from the evolving standards of decency that mark the progress of a maturing society.” Trop v. Dulles, 356 U. S. 86, 100–101 (1958) (plurality opinion). In discerning those standards, we look to “objective factors to the maximum possible extent.” Coker, supra, at 592 (plurality opinion). Laws enacted by the Nation’s legislatures provide the “clearest and most reliable objective evidence of contemporary values.” Penry v. Lynaugh, 492 U. S. 302, 331 (1989). And data reflecting the actions of sentencing juries, where available, can also afford “ ‘a significant and reliable objective index’ ” of societal mores. Coker, supra, at 596 (plurality opinion) (quoting Gregg v. Georgia, 428 U. S. 153, 181 (1976)) (joint opinion of Stewart, Powell, and Stevens, JJ.).
Although objective evidence of this nature is entitled to great weight, it does not end our inquiry. Rather, as the Court today reaffirms, see ante, at 9, 20–21, “the Constitution contemplates that in the end our own judgment will be brought to bear on the question of the acceptability of the death penalty under the Eighth Amendment.” Coker, supra, at 597 (plurality opinion). “[P]roportionality—at least as regards capital punishment—not only requires an inquiry into contemporary standards as expressed by legislators and jurors, but also involves the notion that the magnitude of the punishment imposed must be related to the degree of the harm inflicted on the victim, as well as to the degree of the defendant’s blameworthiness.” Enmund, supra, at 815 (O’Connor, J., dissenting). We therefore have a “constitutional obligation” to judge for ourselves whether the death penalty is excessive punishment for a particular offense or class of offenders. See Stanford, 492 U. S., at 382 (O’Connor, J., concurring in part and concurring in judgment); see also Enmund, supra, at 797 (“[I]t is for us ultimately to judge whether the Eighth Amendment permits imposition of the death penalty”).
B
Twice in the last two decades, the Court has applied these principles in deciding whether the Eighth Amendment permits capital punishment of adolescent offenders. In Thompson v. Oklahoma, 487 U. S. 815 (1988), a plurality of four Justices concluded that the Eighth Amendment barred capital punishment of an offender for a crime committed before the age of 16. I concurred in that judgment on narrower grounds. At the time, 32 state legislatures had “definitely concluded that no 15-year-old should be exposed to the threat of execution,” and no legislature had affirmatively endorsed such a practice. Id., at 849 (O’Connor, J., concurring in judgment). While acknowledging that a national consensus forbidding the execution of 15-year-old offenders “very likely” did exist, I declined to adopt that conclusion as a matter of constitutional law without clearer evidentiary support. Ibid. Nor, in my view, could the issue be decided based on moral proportionality arguments of the type advanced by the Court today. Granting the premise “that adolescents are generally less blameworthy than adults who commit similar crimes,” I wrote, “it does not necessarily follow that all 15-year-olds are incapable of the moral culpability that would justify the imposition of capital punishment.” Id., at 853. Similarly, we had before us no evidence “that 15-year-olds as a class are inherently incapable of being deterred from major crimes by the prospect of the death penalty.” Ibid. I determined instead that, in light of the strong but inconclusive evidence of a national consensus against capital punishment of under-16 offenders, concerns rooted in the Eighth Amendment required that we apply a clear statement rule. Because the capital punishment statute in Thompson did not specify the minimum age at which commission of a capital crime would be punishable by death, I concluded that the statute could not be read to authorize the death penalty for a 15-year-old offender. Id., at 857–858.
The next year, in Stanford v. Kentucky, supra, the Court held that the execution of 16or 17-year-old capital murderers did not violate the Eighth Amendment. I again wrote separately, concurring in part and concurring in the judgment. At that time, 25 States did not permit the execution of under-18 offenders, including 13 that lacked the death penalty altogether. See id., at 370. While noting that “[t]he day may come when there is such general legislative rejection of the execution of 16or 17-year-old capital murderers that a clear national consensus can be said to have developed,” I concluded that that day had not yet arrived. Id., at 381–382 (opinion concurring in part and concurring in judgment). I reaffirmed my view that, beyond assessing the actions of legislatures and juries, the Court has a constitutional obligation to judge for itself whether capital punishment is a proportionate response to the defendant’s blameworthiness. Id., at 382. Nevertheless, I concluded that proportionality arguments similar to those endorsed by the Court today did not justify a categorical Eighth Amendment rule against capital punishment of 16and 17-year-old offenders. See ibid. (citing Thompson, supra, at 853–854 (O’Connor, J., concurring in judgment)).
The Court has also twice addressed the constitutionality of capital punishment of mentally retarded offenders. In Penry v. Lynaugh, 492 U. S. 302 (1989), decided the same year as Stanford, we rejected the claim that the Eighth Amendment barred the execution of the mentally retarded. At that time, only two States specifically prohibited the practice, while 14 others did not have capital punishment at all. 492 U. S., at 334. Much had changed when we revisited the question three Terms ago in Atkins v. Virginia, 536 U. S. 304 (2002). In Atkins, the Court reversed Penry and held that the Eighth Amendment forbids capital punishment of mentally retarded offenders. 536 U. S., at 321. In the 13 years between Penry and Atkins, there had been a wave of legislation prohibiting the execution of such offenders. By the time we heard Atkins, 30 States barred the death penalty for the mentally retarded, and even among those States theoretically permitting such punishment, very few had executed a mentally retarded offender in recent history. 536 U. S., at 314–316. On the basis of this evidence, the Court determined that it was “fair to say that a national consensus ha[d] developed against” the practice. Id., at 316.
But our decision in Atkins did not rest solely on this tentative conclusion. Rather, the Court’s independent moral judgment was dispositive. The Court observed that mentally retarded persons suffer from major cognitive and behavioral deficits, i.e., “subaverage intellectual functioning” and “significant limitations in adaptive skills such as communication, self-care, and self-direction that became manifest before age 18.” Id., at 318. “Because of their impairments, [such persons] by definition … have diminished capacities to understand and process information, to communicate, to abstract from mistakes and learn from experience, to engage in logical reasoning, to control impulses, and to understand the reactions of others.” Ibid. We concluded that these deficits called into serious doubt whether the execution of mentally retarded offenders would measurably contribute to the principal penological goals that capital punishment is intended to serve—retribution and deterrence. Id., at 319–321. Mentally retarded offenders’ impairments so diminish their personal moral culpability that it is highly unlikely that such offenders could ever deserve the ultimate punishment, even in cases of capital murder. Id., at 319. And these same impairments made it very improbable that the threat of the death penalty would deter mentally retarded persons from committing capital crimes. Id., at 319–320. Having concluded that capital punishment of the mentally retarded is inconsistent with the Eighth Amendment, the Court “ ‘le[ft] to the State[s] the task of developing appropriate ways to enforce the constitutional restriction upon [their] execution of sentences.’ ” Id., at 317 (quoting Ford v. Wainwright, 477 U. S. 399, 416–417 (1986)).
II
A
Although the general principles that guide our Eighth Amendment jurisprudence afford some common ground, I part ways with the Court in applying them to the case before us. As a preliminary matter, I take issue with the Court’s failure to reprove, or even to acknowledge, the Supreme Court of Missouri’s unabashed refusal to follow our controlling decision in Stanford. The lower court concluded that, despite Stanford ’s clear holding and historical recency, our decision was no longer binding authority because it was premised on what the court deemed an obsolete assessment of contemporary values. Quite apart from the merits of the constitutional question, this was clear error.
Because the Eighth Amendment “draw[s] its meaning from … evolving standards of decency,” Trop, 356 U. S., at 101 (plurality opinion), significant changes in societal mores over time may require us to reevaluate a prior decision. Nevertheless, it remains “ this Court’s prerogative alone to overrule one of its precedents.” State Oil Co. v. Khan, 522 U. S. 3, 20 (1997) (emphasis added). That is so even where subsequent decisions or factual developments may appear to have “significantly undermined” the rationale for our earlier holding. United States v. Hatter, 532 U. S. 557, 567 (2001); see also State Oil Co., supra, at 20; Rodriguez de Quijas v. Shearson/American Express, Inc., 490 U. S. 477, 484 (1989). The Eighth Amendment provides no exception to this rule. On the contrary, clear, predictable, and uniform constitutional standards are especially desirable in this sphere. By affirming the lower court’s judgment without so much as a slap on the hand, today’s decision threatens to invite frequent and disruptive reassessments of our Eighth Amendment precedents.
B
In determining whether the juvenile death penalty comports with contemporary standards of decency, our inquiry begins with the “clearest and most reliable objective evidence of contemporary values”—the actions of the Nation’s legislatures. Penry, supra, at 331. As the Court emphasizes, the overall number of jurisdictions that currently disallow the execution of under-18 offenders is the same as the number that forbade the execution of mentally retarded offenders when Atkins was decided. Ante, at 10. At present, 12 States and the District of Columbia do not have the death penalty, while an additional 18 States and the Federal Government authorize capital punishment but prohibit the execution of under-18 offenders. See ante, at 27–28 (Appendix A). And here, as in Atkins, only a very small fraction of the States that permit capital punishment of offenders within the relevant class has actually carried out such an execution in recent history: Six States have executed under-18 offenders in the 16 years since Stanford, while five States had executed mentally retarded offenders in the 13 years prior to Atkins. See Atkins, 536 U. S., at 316; V. Streib, The Juvenile Death Penalty Today: Death Sentences and Executions for Juvenile Crimes, January 1, 1973–December 31, 2004, No. 76, pp. 15–23 (2005), available at http://www.law.onu.edu/faculty/streib/documents/Juv DeathDec2004.pdf (last updated Jan. 31, 2005) (as visited Feb. 25, 2005, and available in the Clerk of the Court’s case file) (hereinafter Streib). In these respects, the objective evidence in this case is, indeed, “similar, and in some respects parallel to” the evidence upon which we relied in Atkins. Ante, at 10.
While the similarities between the two cases are undeniable, the objective evidence of national consensus is marginally weaker here. Most importantly, in Atkins there was significant evidence of opposition to the execution of the mentally retarded, but there was virtually no countervailing evidence of affirmative legislative support for this practice. Cf. Thompson, 487 U. S., at 849 (O’Connor, J., concurring in judgment) (attributing significance to the fact that “no legislature in this country has affirmatively and unequivocally endorsed” capital punishment of 15-year-old offenders). The States that permitted such executions did so only because they had not enacted any prohibitory legislation. Here, by contrast, at least eight States have current statutes that specifically set 16 or 17 as the minimum age at which commission of a capital crime can expose the offender to the death penalty. See ante, at 26 (Appendix A).* Five of these eight States presently have one or more juvenile offenders on death row (six if respondent is included in the count), see Streib 24–31, and four of them have executed at least one under-18 offender in the past 15 years, see id., at 15–23. In all, there are currently over 70 juvenile offenders on death row in 12 different States (13 including respondent). See id., at 11, 24–31. This evidence suggests some measure of continuing public support for the availability of the death penalty for 17-year-old capital murderers.
Moreover, the Court in Atkins made clear that it was “not so much the number of [States forbidding execution of the mentally retarded] that [was] significant, but the consistency of the direction of change.” 536 U. S., at 315. In contrast to the trend in Atkins, the States have not moved uniformly towards abolishing the juvenile death penalty. Instead, since our decision in Stanford, two States have expressly reaffirmed their support for this practice by enacting statutes setting 16 as the minimum age for capital punishment. See Mo. Rev. Stat. §565.020.2 (2000); Va. Code Ann. §18.2–10(a) (Lexis 2004). Furthermore, as the Court emphasized in Atkins itself, 536 U. S., at 315, n. 18, the pace of legislative action in this context has been considerably slower than it was with regard to capital punishment of the mentally retarded. In the 13 years between our decisions in Penry and Atkins, no fewer than 16 States banned the execution of mentally retarded offenders. See Atkins, supra, at 314–315. By comparison, since our decision 16 years ago in Stanford, only four States that previously permitted the execution of under-18 offenders, plus the Federal Government, have legistlatively reversed course, and one additional State’s high court has construed the State’s death penalty statute not to apply to under-18 offenders, see State v. Furman, 122 Wash. 2d 440, 458, 858 P. 2d 1092, 1103 (1993) (en banc). The slower pace of change is no doubt partially attributable, as the Court says, to the fact that 11 States had already imposed a minimum age of 18 when Stanford was decided. See ante, at 12–13. Nevertheless, the extraordinary wave of legislative action leading up to our decision in Atkins provided strong evidence that the country truly had set itself against capital punishment of the mentally retarded. Here, by contrast, the halting pace of change gives reason for pause.
To the extent that the objective evidence supporting today’s decision is similar to that in Atkins, this merely highlights the fact that such evidence is not dispositive in either of the two cases. After all, as the Court today confirms, ante, at 9, 20–21, the Constitution requires that “ ‘in the end our own judgment … be brought to bear’ ” in deciding whether the Eighth Amendment forbids a particular punishment. Atkins, supra, at 312 (quoting Coker, 433 U. S., at 597 (plurality opinion)). This judgment is not merely a rubber stamp on the tally of legislative and jury actions. Rather, it is an integral part of the Eighth Amendment inquiry—and one that is entitled to independent weight in reaching our ultimate decision.
Here, as in Atkins, the objective evidence of a national consensus is weaker than in most prior cases in which the Court has struck down a particular punishment under the Eighth Amendment. See Coker, supra, at 595–596 (plurality opinion) (striking down death penalty for rape of an adult woman, where only one jurisdiction authorized such punishment); Enmund, 458 U. S., at 792 (striking down death penalty for certain crimes of aiding and abetting felony-murder, where only eight jurisdictions authorized such punishment); Ford v. Wainwright, 477 U. S., at 408 (striking down capital punishment of the insane, where no jurisdiction permitted this practice). In my view, the objective evidence of national consensus, standing alone, was insufficient to dictate the Court’s holding in Atkins. Rather, the compelling moral proportionality argument against capital punishment of mentally retarded offenders played a decisive role in persuading the Court that the practice was inconsistent with the Eighth Amendment. Indeed, the force of the proportionality argument in Atkins significantly bolstered the Court’s confidence that the objective evidence in that case did, in fact, herald the emergence of a genuine national consensus. Here, by contrast, the proportionality argument against the juvenile death penalty is so flawed that it can be given little, if any, analytical weight—it proves too weak to resolve the lingering ambiguities in the objective evidence of legislative consensus or to justify the Court’s categorical rule.
C
Seventeen-year-old murderers must be categorically exempted from capital punishment, the Court says, because they “cannot with reliability be classified among the worst offenders.” Ante, at 15. That conclusion is premised on three perceived differences between “adults,” who have already reached their 18th birthdays, and “juveniles,” who have not. See ante, at 15–16. First, juveniles lack maturity and responsibility and are more reckless than adults. Second, juveniles are more vulnerable to outside influences because they have less control over their surroundings. And third, a juvenile’s character is not as fully formed as that of an adult. Based on these characteristics, the Court determines that 17-year-old capital murderers are not as blameworthy as adults guilty of similar crimes; that 17-year-olds are less likely than adults to be deterred by the prospect of a death sentence; and that it is difficult to conclude that a 17-year-old who commits even the most heinous of crimes is “irretrievably depraved.” Ante, at 16–18. The Court suggests that “a rare case might arise in which a juvenile offender has sufficient psychological maturity, and at the same time demonstrates sufficient depravity, to merit a sentence of death.” Ante, at 18. However, the Court argues that a categorical age-based prohibition is justified as a prophylactic rule because “[t]he differences between juvenile and adult offenders are too marked and well understood to risk allowing a youthful person to receive the death penalty despite insufficient culpability.” Ante, at 19.
It is beyond cavil that juveniles as a class are generally less mature, less responsible, and less fully formed than adults, and that these differences bear on juveniles’ comparative moral culpability. See, e.g., Johnson v. Texas, 509 U. S. 350, 367 (1993) (“There is no dispute that a defendant’s youth is a relevant mitigating circumstance”); id., at 376 (O’Connor, J., dissenting) (“[T]he vicissitudes of youth bear directly on the young offender’s culpability and responsibility for the crime”); Eddings, 455 U. S., at 115–116 (“Our history is replete with laws and judicial recognition that minors, especially in their earlier years, generally are less mature and responsible than adults”). But even accepting this premise, the Court’s proportionality argument fails to support its categorical rule.
First, the Court adduces no evidence whatsoever in support of its sweeping conclusion, see ante, at 18, that it is only in “rare” cases, if ever, that 17-year-old murderers are sufficiently mature and act with sufficient depravity to warrant the death penalty. The fact that juveniles are generally less culpable for their misconduct than adults does not necessarily mean that a 17-year-old murderer cannot be sufficiently culpable to merit the death penalty. At most, the Court’s argument suggests that the average 17-year-old murderer is not as culpable as the average adult murderer. But an especially depraved juvenile offender may nevertheless be just as culpable as many adult offenders considered bad enough to deserve the death penalty. Similarly, the fact that the availability of the death penalty may be less likely to deter a juvenile from committing a capital crime does not imply that this threat cannot effectively deter some 17-year-olds from such an act. Surely there is an age below which no offender, no matter what his crime, can be deemed to have the cognitive or emotional maturity necessary to warrant the death penalty. But at least at the margins between adolescence and adulthood—and especially for 17-year-olds such as respondent—the relevant differences between “adults” and “juveniles” appear to be a matter of degree, rather than of kind. It follows that a legislature may reasonably conclude that at least some 17-year-olds can act with sufficient moral culpability, and can be sufficiently deterred by the threat of execution, that capital punishment may be warranted in an appropriate case.
Indeed, this appears to be just such a case. Christopher Simmons’ murder of Shirley Crook was premeditated, wanton, and cruel in the extreme. Well before he committed this crime, Simmons declared that he wanted to kill someone. On several occasions, he discussed with two friends (ages 15 and 16) his plan to burglarize a house and to murder the victim by tying the victim up and pushing him from a bridge. Simmons said they could “ ‘get away with it’ ” because they were minors. Brief for Petitioners 3. In accord with this plan, Simmons and his 15-year-old accomplice broke into Mrs. Crook’s home in the middle of the night, forced her from her bed, bound her, and drove her to a state park. There, they walked her to a railroad trestle spanning a river, “hog-tied” her with electrical cable, bound her face completely with duct tape, and pushed her, still alive, from the trestle. She drowned in the water below. Id., at 4. One can scarcely imagine the terror that this woman must have suffered throughout the ordeal leading to her death. Whatever can be said about the comparative moral culpability of 17-year-olds as a general matter, Simmons’ actions unquestionably reflect “ ‘a consciousness materially more “depraved” than that of’ … the average murderer.” See Atkins, 536 U. S., at 319 (quoting Godfrey v. Georgia, 446 U. S. 420, 433 (1980)). And Simmons’ prediction that he could murder with impunity because he had not yet turned 18—though inaccurate—suggests that he did take into account the perceived risk of punishment in deciding whether to commit the crime. Based on this evidence, the sentencing jury certainly had reasonable grounds for concluding that, despite Simmons’ youth, he “ha[d] sufficient psychological maturity” when he committed this horrific murder, and “at the same time demonstrate[d] sufficient depravity, to merit a sentence of death.” See ante, at 18.
The Court’s proportionality argument suffers from a second and closely related defect: It fails to establish that the differences in maturity between 17-year-olds and young “adults” are both universal enough and significant enough to justify a bright-line prophylactic rule against capital punishment of the former. The Court’s analysis is premised on differences in the aggregate between juveniles and adults, which frequently do not hold true when comparing individuals. Although it may be that many 17-year-old murderers lack sufficient maturity to deserve the death penalty, some juvenile murderers may be quite mature. Chronological age is not an unfailing measure of psychological development, and common experience suggests that many 17-year-olds are more mature than the average young “adult.” In short, the class of offenders exempted from capital punishment by today’s decision is too broad and too diverse to warrant a categorical prohibition. Indeed, the age-based line drawn by the Court is indefensibly arbitrary—it quite likely will protect a number of offenders who are mature enough to deserve the death penalty and may well leave vulnerable many who are not.
For purposes of proportionality analysis, 17-year-olds as a class are qualitatively and materially different from the mentally retarded. “Mentally retarded” offenders, as we understood that category in Atkins, are defined by precisely the characteristics which render death an excessive punishment. A mentally retarded person is, “by definition,” one whose cognitive and behavioral capacities have been proven to fall below a certain minimum. See Atkins, 536 U. S., at 318; see also id., at 308, n. 3 (discussing characteristics of mental retardation); id., at 317, and n. 22 (leaving to the States the development of mechanisms to determine which offenders fall within the class exempt from capital punishment). Accordingly, for purposes of our decision in Atkins, the mentally retarded are not merely less blameworthy for their misconduct or less likely to be deterred by the death penalty than others. Rather, a mentally retarded offender is one whose demonstrated impairments make it so highly unlikely that he is culpable enough to deserve the death penalty or that he could have been deterred by the threat of death, that execution is not a defensible punishment. There is no such inherent or accurate fit between an offender’s chronological age and the personal limitations which the Court believes make capital punishment excessive for 17-year-old murderers. Moreover, it defies common sense to suggest that 17-year-olds as a class are somehow equivalent to mentally retarded persons with regard to culpability or susceptibility to deterrence. Seventeen-year-olds may, on average, be less mature than adults, but that lesser maturity simply cannot be equated with the major, lifelong impairments suffered by the mentally retarded.
The proportionality issues raised by the Court clearly implicate Eighth Amendment concerns. But these concerns may properly be addressed not by means of an arbitrary, categorical age-based rule, but rather through individualized sentencing in which juries are required to give appropriate mitigating weight to the defendant’s immaturity, his susceptibility to outside pressures, his cognizance of the consequences of his actions, and so forth. In that way the constitutional response can be tailored to the specific problem it is meant to remedy. The Eighth Amendment guards against the execution of those who are “insufficiently culpable,” see ante, at 19, in significant part, by requiring sentencing that “reflect[s] a reasoned moral response to the defendant’s background, character, and crime.” California v. Brown, 479 U. S. 538, 545 (1987) (O’Connor, J., concurring). Accordingly, the sentencer in a capital case must be permitted to give full effect to all constitutionally relevant mitigating evidence. See Tennard v. Dretke, 542 U. S. ___, ___ (2004) (slip op., at 9–10); Lockett v. Ohio, 438 U. S. 586, 604 (1978) (plurality opinion). A defendant’s youth or immaturity is, of course, a paradigmatic example of such evidence. See Eddings, 455 U. S., at 115–116.
Although the prosecutor’s apparent attempt to use respondent’s youth as an aggravating circumstance in this case is troubling, that conduct was never challenged with specificity in the lower courts and is not directly at issue here. As the Court itself suggests, such “overreaching” would best be addressed, if at all, through a more narrowly tailored remedy. See ante, at 19. The Court argues that sentencing juries cannot accurately evaluate a youthful offender’s maturity or give appropriate weight to the mitigating characteristics related to youth. But, again, the Court presents no real evidence—and the record appears to contain none—supporting this claim. Perhaps more importantly, the Court fails to explain why this duty should be so different from, or so much more difficult than, that of assessing and giving proper effect to any other qualitative capital sentencing factor. I would not be so quick to conclude that the constitutional safeguards, the sentencing juries, and the trial judges upon which we place so much reliance in all capital cases are inadequate in this narrow context.
D
I turn, finally, to the Court’s discussion of foreign and international law. Without question, there has been a global trend in recent years towards abolishing capital punishment for under-18 offenders. Very few, if any, countries other than the United States now permit this practice in law or in fact. See ante, at 22–23. While acknowledging that the actions and views of other countries do not dictate the outcome of our Eighth Amendment inquiry, the Court asserts that “the overwhelming weight of international opinion against the juvenile death penalty … does provide respected and significant confirmation for [its] own conclusions.” Ante, at 24. Because I do not believe that a genuine national consensus against the juvenile death penalty has yet developed, and because I do not believe the Court’s moral proportionality argument justifies a categorical, age-based constitutional rule, I can assign no such confirmatory role to the international consensus described by the Court. In short, the evidence of an international consensus does not alter my determination that the Eighth Amendment does not, at this time, forbid capital punishment of 17-year-old murderers in all cases.
Nevertheless, I disagree with Justice Scalia’s contention, post, at 15–22 (dissenting opinion), that foreign and international law have no place in our Eighth Amendment jurisprudence. Over the course of nearly half a century, the Court has consistently referred to foreign and international law as relevant to its assessment of evolving standards of decency. See Atkins, 536 U. S., at 317, n. 21; Thompson, 487 U. S., at 830–831, and n. 31 (plurality opinion); Enmund, 458 U. S., at 796–797, n. 22; Coker, 433 U. S., at 596, n. 10 (plurality opinion); Trop, 356 U. S., at 102–103 (plurality opinion). This inquiry reflects the special character of the Eighth Amendment, which, as the Court has long held, draws its meaning directly from the maturing values of civilized society. Obviously, American law is distinctive in many respects, not least where the specific provisions of our Constitution and the history of its exposition so dictate. Cf. post, at 18–19 (Scalia, J., dissenting) (discussing distinctively American rules of law related to the Fourth Amendment and the Establishment Clause). But this Nation’s evolving understanding of human dignity certainly is neither wholly isolated from, nor inherently at odds with, the values prevailing in other countries. On the contrary, we should not be surprised to find congruence between domestic and international values, especially where the international community has reached clear agreement—expressed in international law or in the domestic laws of individual countries—that a particular form of punishment is inconsistent with fundamental human rights. At least, the existence of an international consensus of this nature can serve to confirm the reasonableness of a consonant and genuine American consensus. The instant case presents no such domestic consensus, however, and the recent emergence of an otherwise global consensus does not alter that basic fact.
***
In determining whether the Eighth Amendment permits capital punishment of a particular offense or class of offenders, we must look to whether such punishment is consistent with contemporary standards of decency. We are obligated to weigh both the objective evidence of societal values and our own judgment as to whether death is an excessive sanction in the context at hand. In the instant case, the objective evidence is inconclusive; standing alone, it does not demonstrate that our society has repudiated capital punishment of 17-year-old offenders in all cases. Rather, the actions of the Nation’s legislatures suggest that, although a clear and durable national consensus against this practice may in time emerge, that day has yet to arrive. By acting so soon after our decision in Stanford, the Court both pre-empts the democratic debate through which genuine consensus might develop and simultaneously runs a considerable risk of inviting lower court reassessments of our Eighth Amendment precedents.
To be sure, the objective evidence supporting today’s decision is similar to (though marginally weaker than) the evidence before the Court in Atkins. But Atkins could not have been decided as it was based solely on such evidence. Rather, the compelling proportionality argument against capital punishment of the mentally retarded played a decisive role in the Court’s Eighth Amendment ruling. Moreover, the constitutional rule adopted in Atkins was tailored to this proportionality argument: It exempted from capital punishment a defined group of offenders whose proven impairments rendered it highly unlikely, and perhaps impossible, that they could act with the degree of culpability necessary to deserve death. And Atkins left to the States the development of mechanisms to determine which individual offenders fell within this class.
In the instant case, by contrast, the moral proportionality arguments against the juvenile death penalty fail to support the rule the Court adopts today. There is no question that “the chronological age of a minor is itself a relevant mitigating factor of great weight,” Eddings, 455 U. S., at 116, and that sentencing juries must be given an opportunity carefully to consider a defendant’s age and maturity in deciding whether to assess the death penalty. But the mitigating characteristics associated with youth do not justify an absolute age limit. A legislature can reasonably conclude, as many have, that some 17-year-old murderers are mature enough to deserve the death penalty in an appropriate case. And nothing in the record before us suggests that sentencing juries are so unable accurately to assess a 17-year-old defendant’s maturity, or so incapable of giving proper weight to youth as a mitigating factor, that the Eighth Amendment requires the bright-line rule imposed today. In the end, the Court’s flawed proportionality argument simply cannot bear the weight the Court would place upon it.
Reasonable minds can differ as to the minimum age at which commission of a serious crime should expose the defendant to the death penalty, if at all. Many jurisdictions have abolished capital punishment altogether, while many others have determined that even the most heinous crime, if committed before the age of 18, should not be punishable by death. Indeed, were my office that of a legislator, rather than a judge, then I, too, would be inclined to support legislation setting a minimum age of 18 in this context. But a significant number of States, including Missouri, have decided to make the death penalty potentially available for 17-year-old capital murderers such as respondent. Without a clearer showing that a genuine national consensus forbids the execution of such offenders, this Court should not substitute its own “inevitably subjective judgment” on how best to resolve this difficult moral question for the judgments of the Nation’s democratically elected legislatures. See Thompson, supra, at 854 (O’Connor, J., concurring in judgment). I respectfully dissent.
Notes
* In 12 other States that have capital punishment, under-18 offenders can be subject to the death penalty as a result of transfer statutes that permit such offenders to be tried as adults for certain serious crimes. See ante, at 26 (Appendix A). As I observed in Thompson v. Oklahoma, 487 U. S. 815, 850–852 (1988) (opinion concurring in judgment): “There are many reasons, having nothing whatsoever to do with capital punishment, that might motivate a legislature to provide as a general matter for some [minors] to be channeled into the adult criminal justice process.” Accordingly, while these 12 States clearly cannot be counted as opposing capital punishment of under-18 offenders, the fact that they permit such punishment through this indirect mechanism does not necessarily show affirmative and unequivocal legislative support for the practice. See ibid.
Justice O’Connor, with whom Justice Kennedy and Justice Breyer join, dissenting.
The Court today adopts a rule that is not compelled by statute or by this Court’s precedent, that makes little sense as a practical matter, and that will substantially frustrate Congress’ scheme for punishing repeat violent offenders who violate federal gun laws. The Court is willing to acknowledge that the petitioner’s prior state burglary convictions occurred, and that they involved unpermitted entries with intent to commit felonies. But the Court refuses to accept one additional, commonsense inference, based on substantial documentation and without any evidence to the contrary: that petitioner was punished for his entries into buildings.
The petitioner, Reginald Shepard, has never actually denied that the prior crimes at issue were burglaries of buildings. Nor has he denied that, in pleading guilty to those crimes, he understood himself to be accepting punishment for burglarizing buildings. Instead, seeking to benefit from the unavailability of certain old court records and from a minor ambiguity in the prior crimes’ charging documents, petitioner asks us to foreclose any resort to the clear and uncontradicted background documents that gave rise to and supported his earlier convictions.
The Court acquiesces in that wish and instructs the federal courts to ignore all but the narrowest evidence regarding an Armed Career Criminal Act defendant’s prior guilty pleas. I respectfully dissent.
I
The Armed Career Criminal Act of 1984 (ACCA) mandates a 15-year minimum sentence for certain federal firearms violations where the defendant has three prior convictions for a “violent felony.” 18 U. S. C. §924(e). In defining violent felonies for this purpose, Congress has specified that the term includes any crime, punishable by more than one year’s imprisonment, that “is burglary, arson, or extortion, involves use of explosives, or otherwise involves conduct that presents a serious potential risk of physical injury to another.” §924(e)(2)(B)(ii). We held in Taylor v. United States, 495 U. S. 575 (1990), that the statute’s use of the term “burglary” was meant to encompass only what we described as “generic” burglary, a crime with three elements: (i) “unlawful or unprivileged entry into, or remaining in,” (ii) “a building or structure,” (iii) “with intent to commit a crime.” Id., at 598–599.
That left the problem of how to determine whether a defendant’s past conviction qualified as a conviction for generic burglary. The most formalistic approach would have been to find the ACCA requirement satisfied only when the statute under which the defendant was convicted was one limited to “generic” burglary. But Taylor wisely declined to follow that course. The statutes which some States—like Massachusetts here, or Missouri in Taylor —use to prosecute generic burglary are overbroad for ACCA purposes: They are not limited to “generic” burglary, but also punish the nongeneric kind. Restricting the sentencing court’s inquiry to the face of the statute would have frustrated the purposes of the ACCA by allowing some violent recidivists convicted of federal gun crimes to escape the ACCA’s heightened punishment based solely on the fortuity of where they had committed their previous crimes.
Instead, Taylor adopted a more “pragmatic” approach. Ante, at 6 (majority opinion). Every statute punishes a certain set of criminalized actions; the problem with some burglary statutes, for purposes of the ACCA, is that they are overinclusive. But Taylor permitted a federal court to “go beyond the mere fact of conviction”—and to determine, by using other sources, whether the defendant’s prior crime was in the subset of the statutory crime qualifying as generic burglary. For example, where a defendant’s prior conviction occurred by jury trial, Taylor instructed the federal court to review “the indictment or information and jury instructions” from the earlier conviction, to see whether they had “required the jury to find all the elements of generic burglary in order to convict.” 495 U. S., at 602.
As the Court recognizes, however, Taylor ’s use of that one example did not purport to be exhaustive. See ante, at 6. See also United States v. Harris, 964 F. 2d 1234, 1236 (CA1 1992) (Breyer, C. J.). Rather, Taylor left room for courts to determine which other reliable and simple sources might aid in determining whether a defendant had in fact been convicted of generic burglary. The Court identifies several such sources that a sentencing judge may consult under the ACCA: the “charging document, written plea agreement, transcript of plea colloquy, and any explicit factual finding by the trial judge to which the defendant assented.” Ante, at 1–2. I would expand that list to include any uncontradicted, internally consistent parts of the record from the earlier conviction. That would include the two sources the First Circuit relied upon in this case.
Shepard’s four prior convictions all occurred by guilty pleas to charges under Massachusetts’ two burglary statutes—statutes that punish “[w]hoever … breaks and enters a building, ship, vessel or vehicle, with intent to commit a felony.” Mass. Gen. Laws Ann., ch. 266, §16 (West 2000) (emphasis added); see also §18. The criminal complaints used as charging documents for the convictions at issue did not specify that Shepard’s offenses had involved a building, but instead closely copied the more inclusive language of the appropriate statute. If these complaints were the only evidence of the factual basis of Shepard’s guilty pleas, then I would agree with the majority that there was no way to know whether those convictions were for burglarizing a building. But the Government did have additional evidence. For each of the convictions, the Government had both the applications by which the police had secured the criminal complaints and the police reports attached to those applications. Those documents decisively show that Shepard’s illegal act in each prior conviction was the act of entering a building. Moreover, they make inescapable the conclusion that, at each guilty plea, Shepard understood himself to be admitting the crime of breaking into a building.
Consider, for instance, the first burglary conviction at issue. The complaint for that conviction alleged that, on May 6, 1989, Shepard “did break and enter in the night time the building, ship, vessel or vehicle, the property of Jerri Cothran, with intent to commit a felony therein” in violation of §16. 3 App. 5. The place of the offense was alleged as “30 Harlem St.,” and the complaint contained a cross-reference to “CC#91–394783.”
The majority would have us stop there. Since both the statute and the charging document name burglary of a “building, ship, vessel or vehicle,” the majority concludes that there is no way to tell whether Massachusetts punished Shepard for transgressing its laws by burglarizing a building, or for doing so by burglarizing a vehicle, ship, or vessel. (Although the majority would also allow a look at Shepard’s written plea agreement or a transcript of the plea proceedings, those items are no longer available in Shepard’s case, since Massachusetts has apparently seen little need to preserve the miscellany of long-past convictions.)
I would look as well to additional portions of the record from that plea—the complaint application and police report. The complaint application lists the same statute, describes (in abbreviated form) the same offense, names the same victim and address, and contains the same reference number (though differently hyphenated) as the complaint itself. In addition, the application specifies as relevant “PROPERTY” (meaning “Goods stolen, what destroyed, etc.”) a “Cellar Door.” Id., at 6. The police report (which also names the same victim, date, and place of offense, and contains the same reference number as the other two documents) gives substantially more detail about why Massachusetts began criminal proceedings against Shepard:
“[R]esponded to [radio call] to 30 Harlem St. for B&E; in progress. On arrival observed cellar door in rear had been broken down. Spoke to victim who stated that approx 3:00 a.m. she heard noises downstairs. She then observed suspect … in her pantry.” Id., at 7.
Three points need to be made about the relationship between the complaint (whose use the majority finds completely unobjectionable) and the application and police report (which I would also consider). First, all of the documents concern the same crime. Second, the three documents are entirely consistent—nothing in any of them casts doubt on the veracity of the others. Finally, and most importantly, the common understanding behind all three documents was that, whatever the range of conduct punishable by the state statute, this defendant was being prosecuted for burglary of a building. See 348 F. 3d 308, 314 (CA1 2003) (“[T]here is a compelling inference that the plea was to the complaint and that the complaint embodied the events described in the application or police report in the case file”).
There certainly is no evidence in the record contradicting that understanding. Notably, throughout these proceedings, Shepard has never denied that the four guilty pleas at issue involved breaking into buildings. Nor has he denied that his contemporaneous understanding of each plea was that, as a result of his admission, he would be punished for having broken into a building. During his federal sentencing hearings, Shepard did submit an affidavit about his prior convictions. But that affidavit carefully dances around the key issues of what Shepard actually did to run afoul of the law and what he thought was the substance of his guilty plea. Rather, the affidavit focuses on what the judge said to Shepard at the hearing and what Shepard said in response. Even in that regard, the affidavit is strangely ambiguous. In discussing the first conviction, for instance, the affidavit states that “the judge [who took the plea] did not read” the police report to Shepard, “and did not ask me whether or not the information contained in the … report was true.” 1 App. 100. See also ibid. (“I did not admit the truth of the information contained in the … report as part of my plea and I have never admitted in court that the facts alleged in the reports are true”). The affidavit’s statements about the other three prior convictions are similar.
Those statements could be taken as Shepard’s denial that he was ever asked about (or ever admitted to) any of the specific facts of his crime that happen to be mentioned in the police reports—facts like the date and place of the offense, whether he entered through a cellar door and proceeded to the pantry, and so on. But to believe that, we would have to presume that all four Massachusetts courts violated their duty under state law to ensure themselves of the factual basis for Shepard’s plea. In Massachusetts, “[a] defendant’s choice to plead guilty will not alone support conviction; the defendant’s guilt in fact must be established.” Commonwealth v. DelVerde, 398 Mass. 288, 296, 496 N. E. 2d 1357, 1362 (1986). As a result, even if “the defendant admits to the crime in open court, … a court may not convict unless there are sufficient facts on the record to establish each element of the offense.” Id., at 297, 496 N. E. 2d, at 1363. See also Commonwealth v. Colon, 439 Mass. 519, 529, n. 13, 789 N. E. 2d 566, 573, n. 13 (2003) (guilty plea requires admission to the facts); 2 E. Blumenson, S. Fisher, & D. Kanstroom, Massachusetts Criminal Practice §37.7B, p. 288 (1998) (“Usually this is accomplished by the recitation of either the grand jury minutes or police reports, but defendant’s admissions during the plea, or trial evidence, can also support the factual basis” (footnote omitted)). Cf. Commonwealth v. Forde, 392 Mass. 453, 458, 466 N. E. 2d 510, 513 (1984) (conviction cannot be based on uncorroborated confession; rather, there must be some evidence that the crime was “real and not imaginary”). It is thus unlikely that Shepard really intended his affidavit as a statement that none of the various facts found in the police reports were ever admitted by him or discussed in his presence during his guilty pleas.
More likely, Shepard’s attorney carefully phrased the affidavit so that it would admit of a different meaning: that the plea courts never asked, and Shepard never answered, the precise question: “Is what the police report says true?” But I fail to see how that is relevant, so long as Shepard understood that, in pleading guilty, he was agreeing to be punished for the building break-in that was the subject of the entire proceeding.
There may be some scenarios in which—as the result of charge bargaining, for instance, or due to unexpected twists in an investigation—a defendant’s guilty plea is premised on substantially different facts than those that were the basis for the original police investigation. In such a case, a defendant might well be confused about the practical meaning of his admission of guilt. Cf. Taylor, 495 U. S., at 601–602 (“[I]f a guilty plea to a lesser, nonburglary offense was the result of a plea bargain, it would seem unfair to impose a sentence enhancement as if the defendant had pleaded guilty to burglary”). But there is no claim of such circumstances here: All signs are that everyone involved in each prior plea—from the judge, to the prosecutor, to the defense lawyer, to Shepard himself—understood each plea as Shepard’s admission that he had broken into the building where the police caught him. Given each police report’s never-superseded allegation that Shepard had burglarized a building, it strains credulity beyond the breaking point to assert that, in each case, Shepard was actually prosecuted for and pleaded guilty to burglarizing a ship or a car. The lower court was surely right to detect “an air of make-believe” about Shepard’s case. 348 F. 3d, at 311.
The majority’s rule, which forces the federal sentencing court to feign agnosticism about clearly knowable facts, cannot be squared with the ACCA’s twin goals of incapacitating repeat violent offenders, and of doing so consistently notwithstanding the peculiarities of state law. Cf. Taylor, supra, at 582 (“ ‘[I]n terms of fundamental fairness, the Act should ensure, to the extent that it is consistent with the prerogatives of the States in defining their own offenses, that that same type of conduct is punishable on the Federal level in all cases’ ” (quoting S. Rep. No. 98–190, p. 20 (1983))). The Court’s overscrupulous regard for formality leads it not only to an absurd result, but also to a result that Congress plainly hoped to avoid.
II
The Court gives two principal reasons for today’s ruling: adherence to the Court’s decision in Taylor, and constitutional concerns about the defendant’s right to a jury trial.
The first is hardly convincing. As noted above, Taylor itself set no rule for guilty pleas, and its list of sources for a sentencing court to consider was not intended to be exhaustive. Supra, at 3. The First Circuit’s disposition of this case, therefore, was not in direct conflict with Taylor. Nor did it conflict with the spirit of Taylor. Taylor was in part about “[f]air[ness]” to defendants. 495 U. S., at 602. But there is nothing unfair (and a great deal that is positively just) about recognizing and acting upon plain and uncontradicted evidence that a defendant, in entering his prior plea, knew he was being prosecuted for and was pleading guilty to burglary of a building. Taylor also sought to avoid the impracticality of mini-sentencing-trials featuring opposing witnesses perusing lengthy transcripts of prior proceedings. Id., at 601. But no such problem presents itself in this case: The Government proposed using only the small documentary record behind Shepard’s pleas. Those documents relate to facts that Shepard does not dispute, and Shepard has not indicated any desire to submit counterevidence.
The issue most central to Taylor was the need to effectuate Congress’ “categorical approach” to sentencing recidivist federal offenders—an approach which responds to the reality of a defendant’s prior crimes, rather than the happenstance of how those crimes “were labeled by state law.” Id., at 589. But rather than promote this goal, the majority opinion today injects a new element of arbitrariness into the ACCA: A defendant’s sentence will now depend not only on the peculiarities of the statutes particular States use to prosecute generic burglary, but also on whether those States’ record retention policies happen to preserve the musty “written plea agreement[s]” and recordings of “plea colloqu[ies]” ancillary to long-past convictions. Ante, at 1. In other words, with respect to this most critical issue, the majority’s rule is not consistent with Taylor at all.
That is why I strongly suspect that the driving force behind today’s decision is not Taylor itself, but rather “[d]evelopments in the law since Taylor.” Ante, at 9. A majority of the Court defends its rule as necessary to avoid a result that might otherwise be unconstitutional under Apprendi v. New Jersey, 530 U. S. 466 (2000), and related cases. Ante, at 10–12 (plurality opinion); ante, at 2–3 (Thomas, J., concurring in part and concurring in judgment). I have criticized that line of cases from the beginning, and I need not repeat my reasoning here. See Id., at 523 (dissenting opinion); Ring v. Arizona, 536 U. S. 584, 619 (2002) (dissenting opinion); Blakely v. Washington, 542 U. S. ___, ___ (2004) (slip op., at 8–10) (dissenting opinion). See also Jones v. United States, 526 U. S. 227, 254 (1999) (Kennedy, J., dissenting); Blakely, supra, at ___ (slip op., at 13–17) (Breyer, J., dissenting); United States v. Booker, 543 U. S. ___, ___ (2005) (slip op., at 2–6) (Breyer, J., dissenting). It is a battle I have lost.
But it is one thing for the majority to apply its Apprendi rule within that rule’s own bounds, and quite another to extend the rule into new territory that Apprendi and succeeding cases had expressly and consistently disclaimed. Yet today’s decision reads Apprendi to cast a shadow possibly implicating recidivism determinations, which until now had been safe from such formalism. See Blakely, supra, at ___ (slip op., at 5) (“ ‘ Other than the fact of a prior conviction, any fact that increases the penalty of a crime beyond the prescribed statutory maximum must be submitted to a jury, and proved beyond a reasonable doubt’ ”) (emphasis added; quoting Apprendi, supra, at 490)). See also Booker, supra, at ___ (slip op., at 20) (opinion of the Court by Stevens, J.) (similar).
Even in a post Apprendi world, I cannot understand how today’s case raises any reasonable constitutional concern. To the contrary, this case presents especially good reasons for respecting Congress’ long “tradition of treating recidivism as a sentencing factor” determined by the judge, Almendarez-Torres v. United States, 523 U. S. 224, 243 (1998), rather than as a substantive offense element determined by the jury. First, Shepard’s prior convictions were themselves “established through procedures satisfying the fair notice, reasonable doubt, and jury trial guarantees.” Jones, supra, at 249. Second, as with most recidivism determinations, see Almendarez-Torres, supra, at 235, the burglary determination in Shepard’s case concerned an extremely narrow issue, with the relevant facts not seriously contested. See supra, at 6–7 (discussing shortcomings of Shepard’s affidavit). Finally, today’s hint at extending the Apprendi rule to the issue of ACCA prior crimes surely will do no favors for future defendants in Shepard’s shoes. When ACCA defendants in the future go to trial rather than plead guilty, the majority’s ruling in effect invites the Government, in prosecuting the federal gun charge, also “to prove to the jury” the defendant’s prior burglaries. Almendarez-Torres, 523 U. S., at 234–235. “[T]he introduction of evidence of a defendant’s prior crimes risks significant prejudice,” id., at 235, and that prejudice is likely to be especially strong in ACCA cases, where the relevant prior crimes are, by definition, “violent,” 18 U. S. C. §924(e). In short, whatever the merits of the Apprendi doctrine, that doctrine does not currently bear on, and should not be extended to bear on, determinations of a defendant’s past crimes, like the ACCA predicates at issue in Shepard’s case. The plurality’s concern about constitutional doubt, ante, at 10–12, and Justice Thomas’ concern about constitutional error, ante, at 2–3, are therefore misplaced.
***
For the reasons explained above, I would find that the First Circuit properly established the applicability of the ACCA sentence by looking to the complaint applications and police reports from the prior convictions. Because the Court concludes otherwise, I respectfully dissent.
Justice O’Connor delivered the opinion of the Court.
Roderick Jackson, a teacher in the Birmingham, Alabama, public schools, brought suit against the Birmingham Board of Education (Board) alleging that the Board retaliated against him because he had complained about sex discrimination in the high school’s athletic program. Jackson claimed that the Board’s retaliation violated Title IX of the Education Amendments of 1972, Pub. L. 92–318, 86 Stat. 373, as amended, 20 U. S. C. §1681 et seq. The District Court dismissed Jackson’s complaint on the ground that Title IX does not prohibit retaliation, and the Court of Appeals for the Eleventh Circuit affirmed. 309 F. 3d 1333 (2002). We consider here whether the private right of action implied by Title IX encompasses claims of retaliation. We hold that it does where the funding recipient retaliates against an individual because he has complained about sex discrimination.
I
Because Jackson’s Title IX claim was dismissed under Federal Rule of Civil Procedure 12(b)(6) for failure to state a claim upon which relief can be granted, “we must assume the truth of the material facts as alleged in the complaint.” Summit Health, Ltd. v. Pinhas, 500 U. S. 322, 325 (1991).
According to the complaint, Jackson has been an employee of the Birmingham school district for over 10 years. In 1993, the Board hired Jackson to serve as a physical education teacher and girls’ basketball coach. Jackson was transferred to Ensley High School in August 1999. At Ensley, he discovered that the girls’ team was not receiving equal funding and equal access to athletic equipment and facilities. The lack of adequate funding, equipment, and facilities made it difficult for Jackson to do his job as the team’s coach.
In December 2000, Jackson began complaining to his supervisors about the unequal treatment of the girls’ basketball team, but to no avail. Jackson’s complaints went unanswered, and the school failed to remedy the situation. Instead, Jackson began to receive negative work evaluations and ultimately was removed as the girls’ coach in May 2001. Jackson is still employed by the Board as a teacher, but he no longer receives supplemental pay for coaching.
After the Board terminated Jackson’s coaching duties, he filed suit in the United States District Court for the Northern District of Alabama. He alleged, among other things, that the Board violated Title IX by retaliating against him for protesting the discrimination against the girls’ basketball team. Amended Complaint 2–3, App. 10–11. The Board moved to dismiss on the ground that Title IX’s private cause of action does not include claims of retaliation. The District Court granted the motion to dismiss.
The Court of Appeals for the Eleventh Circuit affirmed. 309 F. 3d 1333 (2002). It assumed, for purposes of the appeal, that the Board retaliated against Jackson for complaining about Title IX violations. It then held that Jackson’s suit failed to state a claim because Title IX does not provide a private right of action for retaliation, reasoning that “[n]othing in the text indicates any congressional concern with retaliation that might be visited on those who complain of Title IX violations.” Id., at 1344. Relying on our decision in Alexander v. Sandoval, 532 U. S. 275 (2001), the Court of Appeals also concluded that a Department of Education regulation expressly prohibiting retaliation does not create a private cause of action for retaliation: “Because Congress has not created a right through Title IX to redress harms resulting from retaliation, [the regulation] may not be read to create one either.” 309 F. 3d, at 1346. Finally, the court held that, even if Title IX prohibits retaliation, Jackson would not be entitled to relief because he is not within the class of persons protected by the statute.
We granted certiorari, 542 U. S. __ (2004), to resolve a conflict in the Circuits over whether Title IX’s private right of action encompasses claims of retaliation for complaints about sex discrimination. Compare Lowrey v. Texas A & M Univ. System, 117 F. 3d 242, 252 (CA5 1997) (“[T]itle IX affords an implied cause of action for retaliation”); Preston v. Virginia ex rel. New River Community College, 31 F. 3d 203, 206 (CA4 1994) (same); with the case below, supra.
II
A
Title IX prohibits sex discrimination by recipients of federal education funding. The statute provides that “[n]o person in the United States shall, on the basis of sex, be excluded from participation in, be denied the benefits of, or be subjected to discrimination under any education program or activity receiving Federal financial assistance.” 20 U. S. C. §1681(a). More than 25 years ago, in Cannon v. University of Chicago, 441 U. S. 677, 690–693 (1979), we held that Title IX implies a private right of action to enforce its prohibition on intentional sex discrimination. In subsequent cases, we have defined the contours of that right of action. In Franklin v. Gwinnett County Public Schools, 503 U. S. 60 (1992), we held that it authorizes private parties to seek monetary damages for intentional violations of Title IX. We have also held that the private right of action encompasses intentional sex discrimination in the form of a recipient’s deliberate indifference to a teacher’s sexual harassment of a student, Gebser v. Lago Vista Independent School Dist., 524 U. S. 274, 290–291 (1998), or to sexual harassment of a student by another student, Davis v. Monroe County Bd. of Ed., 526 U. S. 629, 642 (1999).
In all of these cases, we relied on the text of Title IX, which, subject to a list of narrow exceptions not at issue here, broadly prohibits a funding recipient from subjecting any person to “discrimination” “on the basis of sex.” 20 U. S. C. §1681. Retaliation against a person because that person has complained of sex discrimination is another form of intentional sex discrimination encompassed by Title IX’s private cause of action. Retaliation is, by definition, an intentional act. It is a form of “discrimination” because the complainant is being subjected to differential treatment. See generally Olmstead v. L. C., 527 U. S. 581, 614 (1999) (Kennedy, J., concurring in judgment) (the “normal definition of discrimination” is “differential treatment”); see also Newport News Shipbuilding & Dry Dock Co. v. EEOC, 462 U. S. 669, 682, n. 22 (1983) (discrimination means “less favorable” treatment). Moreover, retaliation is discrimination “on the basis of sex” because it is an intentional response to the nature of the complaint: an allegation of sex discrimination. We conclude that when a funding recipient retaliates against a person because he complains of sex discrimination, this constitutes intentional “discrimination” “on the basis of sex,” in violation of Title IX.
The Court of Appeals’ conclusion that Title IX does not prohibit retaliation because the “statute makes no mention of retaliation,” 309 F. 3d, at 1344, ignores the import of our repeated holdings construing “discrimination” under Title IX broadly. Though the statute does not mention sexual harassment, we have held that sexual harassment is intentional discrimination encompassed by Title IX’s private right of action. Franklin, 503 U. S., at 74–75; see also id., at 75 (noting that, under Meritor Savings Bank, FSB v. Vinson, 477 U. S. 57, 64 (1986), “ ‘when a supervisor sexually harasses a subordinate because of the subordinate’s sex, that supervisor “discriminate[s]” on the basis of sex,’ ” and holding that “the same rule should apply when a teacher sexually harasses … a student”). Thus, a recipient’s deliberate indifference to a teacher’s sexual harassment of a student also “violate[s] Title IX’s plain terms.” Davis, supra, at 643 (citing Gebser, supra, at 290–291). Likewise, a recipient’s deliberate indifference to sexual harassment of a student by another student also squarely constitutes “discrimination” “on the basis of sex.” Davis, 526 U. S., at 643; see also id., at 650 (“Having previously determined that ‘sexual harassment’ is ‘discrimination’… under Title IX, we are constrained to conclude that student-on-student sexual harassment, if sufficiently severe, can likewise rise to the level of discrimination actionable under the statute”). “Discrimination” is a term that covers a wide range of intentional unequal treatment; by using such a broad term, Congress gave the statute a broad reach. See North Haven Bd. of Ed. v. Bell, 456 U. S. 512, 521 (1982) (Courts “ ‘must accord’ ” Title IX “ ‘a sweep as broad as its language’ ”).
Congress certainly could have mentioned retaliation in Title IX expressly, as it did in §704 of Title VII of the Civil Rights Act of 1964, 78 Stat. 257, as amended, 86 Stat. 109, 42 U. S. C. §2000e–3(a) (providing that it is an “unlawful employment practice” for an employer to retaliate against an employee because he has “opposed any practice made an unlawful employment practice by [Title VII], or because he has made a charge, testified, assisted, or participated in any manner in an investigation, proceeding, or hearing under [Title VII]”). Title VII, however, is a vastly different statute from Title IX, see Gebser, 524 U. S., at 283–284, 286–287, and the comparison the Board urges us to draw is therefore of limited use. Title IX’s cause of action is implied, while Title VII’s is express. See id., at 283–284. Title IX is a broadly written general prohibition on discrimination, followed by specific, narrow exceptions to that broad prohibition. See 20 U. S. C. §1681. By contrast, Title VII spells out in greater detail the conduct that constitutes discrimination in violation of that statute. See 42 U. S. C. §§2000e–2 (giving examples of unlawful employment practices), 2000e–3 (prohibiting “[o]ther unlawful employment practices,” including (a) “discrimination” in the form of retaliation; and (b) the discriminatory practice of “[p]rinting or publication of notices or advertisements indicating prohibited preference …”). Because Congress did not list any specific discriminatory practices when it wrote Title IX, its failure to mention one such practice does not tell us anything about whether it intended that practice to be covered.
Title IX was enacted in 1972, three years after our decision in Sullivan v. Little Hunting Park, Inc., 396 U. S. 229 (1969). In Sullivan, we held that Rev. Stat. §1978, 42 U. S. C. §1982, which provides that “[a]ll citizens of the United States shall have the same right … as is enjoyed by white citizens … to inherit, purchase, lease, sell, hold, and convey real and personal property,” protected a white man who spoke out against discrimination toward one of his tenants and who suffered retaliation as a result. Sullivan had rented a house to a black man and assigned him a membership share and use rights in a private park. The corporation that owned the park would not approve the assignment to the black lessee. Sullivan protested, and the corporation retaliated against him by expelling him and taking his shares. Sullivan sued the corporation, and we upheld Sullivan’s cause of action under 42 U. S. C. §1982 for “[retaliation] for the advocacy of [the black person’s] cause.” 396 U. S., at 237. Thus, in Sullivan we interpreted a general prohibition on racial discrimination to cover retaliation against those who advocate the rights of groups protected by that prohibition.[ Footnote 1 ]
Congress enacted Title IX just three years after Sullivan was decided, and accordingly that decision provides a valuable context for understanding the statute. As we recognized in Cannon, “it is not only appropriate but also realistic to presume that Congress was thoroughly familiar with [Sullivan] and that it expected its enactment [of Title IX] to be interpreted in conformity with [it].” 441 U. S., at 699; see also id., at 698, n. 22. Retaliation for Jackson’s advocacy of the rights of the girls’ basketball team in this case is “discrimination” “on the basis of sex,” just as retaliation for advocacy on behalf of a black lessee in Sullivan was discrimination on the basis of race.
B
The Board contends that our decision in Alexander v. Sandoval, 532 U. S. 275 (2001), compels a holding that Title IX’s private right of action does not encompass retaliation. Sandoval involved an interpretation of Title VI of the Civil Rights Act of 1964, 78 Stat. 252, as amended, 42 U. S. C. §2000d et seq., which provides in §601 that no person shall, “on the ground of race, color, or national origin, be excluded from participation in, be denied the benefits of, or be subjected to discrimination under any program or activity” covered by Title VI. 42 U. S. C. §2000d. Section 602 of Title VI authorizes federal agencies to effectuate the provisions in §601 by enacting regulations. Pursuant to that authority, the Department of Justice promulgated regulations forbidding funding recipients from adopting policies that had “the effect of subjecting individuals to discrimination because of their race, color, or national origin.” 28 CFR §42.104(b)(2) (1999). The Sandoval petitioners brought suit to enjoin an English-only policy of the Alabama Department of Public Safety on grounds that it disparately impacted non-English speakers in violation of the regulations. Though we assumed that the regulations themselves were valid, see 532 U. S., at 281, we rejected the contention that the private right of action to enforce intentional violations of Title VI encompassed suits to enforce the disparate-impact regulations. We did so because “[i]t is clear… that the disparate-impact regulations do not simply apply §601—since they indeed forbid conduct that §601 permits—and therefore clear that the private right of action to enforce §601 does not include a private right to enforce these regulations.” Id., at 285. See also Central Bank of Denver, N. A. v. First Interstate Bank of Denver, N. A., 511 U. S. 164, 173 (1994) (A “private plaintiff may not bring a [suit based on a regulation] against a defendant for acts not prohibited by the text of [the statute]”). Thus, Sandoval held that private parties may not invoke Title VI regulations to obtain redress for disparate-impact discrimination because Title VI itself prohibits only intentional discrimination.
The Board cites a Department of Education regulation prohibiting retaliation “against any individual for the purpose of interfering with any right or privilege secured by [Title IX],” 34 CFR §100.7(e) (2004) (incorporated by reference by §106.71), and contends that Jackson, like the petitioners in Sandoval, seeks an “impermissible extension of the statute” when he argues that Title IX’s private right of action encompasses retaliation. Brief for Respondent 45. This argument, however, entirely misses the point. We do not rely on regulations extending Title IX’s protection beyond its statutory limits; indeed, we do not rely on the Department of Education’s regulation at all, because the statute itself contains the necessary prohibition. As we explain above, see supra, at 4–5, the text of Title IX prohibits a funding recipient from retaliating against a person who speaks out against sex discrimination, because such retaliation is intentional “discrimination” “on the basis of sex.” We reach this result based on the statute’s text. In step with Sandoval, we hold that Title IX’s private right of action encompasses suits for retaliation, because retaliation falls within the statute’s prohibition of intentional discrimination on the basis of sex. [ Footnote 2 ]
C
Nor are we convinced by the Board’s argument that, even if Title IX’s private right of action encompasses discrimination, Jackson is not entitled to invoke it because he is an “indirect victi[m]” of sex discrimination. Brief for Respondent 33. The statute is broadly worded; it does not require that the victim of the retaliation must also be the victim of the discrimination that is the subject of the original complaint. If the statute provided instead that “no person shall be subjected to discrimination on the basis of such individual’s sex,” then we would agree with the Board. Cf. 42 U. S. C. §2000e–2(a)(1) (“It shall be an unlawful employment practice for an employer… to discriminate against any individual … because of such individual’s race, color, religion, sex, or national origin” (emphasis added)). However, Title IX contains no such limitation. Where the retaliation occurs because the complainant speaks out about sex discrimination, the “on the basis of sex” requirement is satisfied. The complainant is himself a victim of discriminatory retaliation, regardless of whether he was the subject of the original complaint.[ Footnote 3 ] As we explain above, see supra, at 6–7, this is consistent with Sullivan, which formed an important part of the backdrop against which Congress enacted Title IX. Sullivan made clear that retaliation claims extend to those who oppose discrimination against others. See 396 U. S., at 237 (holding that a person may bring suit under 42 U. S. C. §1982 if he can show that he was “punished for trying to vindicate the rights of minorities”).
Congress enacted Title IX not only to prevent the use of federal dollars to support discriminatory practices, but also “to provide individual citizens effective protection against those practices.” Cannon, 441 U. S., at 704. We agree with the United States that this objective “would be difficult, if not impossible, to achieve if persons who complain about sex discrimination did not have effective protection against retaliation.” Brief for United States as Amicus Curiae 13. If recipients were permitted to retaliate freely, individuals who witness discrimination would be loathe to report it, and all manner of Title IX violations might go unremedied as a result. See Sullivan, supra, at 237 (noting that without protection against retaliation, the underlying discrimination is perpetuated).
Reporting incidents of discrimination is integral to Title IX enforcement and would be discouraged if retaliation against those who report went unpunished. Indeed, if retaliation were not prohibited, Title IX’s enforcement scheme would unravel. Recall that Congress intended Title IX’s private right of action to encompass claims of a recipient’s deliberate indifference to sexual harassment. See generally Davis, 526 U. S. 629. Accordingly, if a principal sexually harasses a student, and a teacher complains to the school board but the school board is indifferent, the board would likely be liable for a Title IX violation. See generally Gebser, 524 U. S. 274. But if Title IX’s private right of action does not encompass retaliation claims, the teacher would have no recourse if he were subsequently fired for speaking out. Without protection from retaliation, individuals who witness discrimination would likely not report it, indifference claims would be short-circuited, and the underlying discrimination would go unremedied.
Title IX’s enforcement scheme also depends on individual reporting because individuals and agencies may not bring suit under the statute unless the recipient has received “actual notice” of the discrimination. Id., at 288, 289–290 (holding that an appropriate official of the recipient must have actual knowledge of discrimination and respond with deliberate indifference before a private party may bring suit); 20 U. S. C. §1682 (providing that a federal agency may terminate funding only after it “has advised the appropriate person or persons of the failure to comply with the requirement and has determined that compliance cannot be secured by voluntary means”). If recipients were able to avoid such notice by retaliating against all those who dare complain, the statute’s enforcement scheme would be subverted. We should not assume that Congress left such a gap in its scheme.
Moreover, teachers and coaches such as Jackson are often in the best position to vindicate the rights of their students because they are better able to identify discrimination and bring it to the attention of administrators. Indeed, sometimes adult employees are “the only effective adversar[ies]” of discrimination in schools. See Sullivan, supra, at 237 (“[A] white owner is at times ‘the only effective adversary’ of the unlawful restrictive covenant” (citing Barrows v. Jackson, 346 U. S. 249, 259 (1953))).
D
The Board is correct in pointing out that, because Title IX was enacted as an exercise of Congress’ powers under the Spending Clause, see, e.g., Davis, supra, at 640; Gebser, supra, at 287; Franklin, 503 U. S., at 74–75, and n. 8., “private damages actions are available only where recipients of federal funding had adequate notice that they could be liable for the conduct at issue,” Davis, supra, at 640. When Congress enacts legislation under its spending power, that legislation is “in the nature of a contract: in return for federal funds, the States agree to comply with federally imposed conditions.” Pennhurst State School and Hospital v. Halderman, 451 U. S. 1, 17 (1981). As we have recognized, “[t]here can … be no knowing acceptance [of the terms of the contract] if a State is unaware of the conditions [imposed by the legislation on its receipt of funds].” Ibid.
The Board insists that we should not interpret Title IX to prohibit retaliation because it was not on notice that it could be held liable for retaliating against those who complain of Title IX violations. We disagree. Funding recipients have been on notice that they could be subjected to private suits for intentional sex discrimination under Title IX since 1979, when we decided Cannon. Pennhurst does not preclude private suits for intentional acts that clearly violate Title IX. Davis, supra, at 642.
Indeed, in Davis, we held that Pennhurst did not pose an obstacle to private suits for damages in cases of a recipient’s deliberate indifference to one student’s sexual harassment of another, because the deliberate indifference constituted intentional discrimination on the basis of sex. Davis, supra, at 650. See also Franklin, supra, at 75 (“Congress surely did not intend for federal monies to be expended to support the intentional actions it sought by statute to proscribe”). Similarly, we held in Gebser that a recipient of federal funding could be held liable for damages under Title IX for deliberate indifference to a teacher’s harassment of a student. 524 U. S., at 287–288. In Gebser, as in Davis, we acknowledged that federal funding recipients must have notice that they will be held liable for damages. See Davis, supra, at 642; Gebser, supra, at 287. But we emphasized that “this limitation on private damages actions is not a bar to liability where a funding recipient intentionally violates the statute.” Davis, supra, at 642 (citing Franklin, 503 U. S., at 74–75). See also ibid. (“[T]he [Pennhurst] notice problem does not arise in a case such as this, in which intentional discrimination is alleged”); Bennett v. Kentucky Dept. of Ed., 470 U. S. 656, 665–666 (1985) (holding that there was sufficient notice under Pennhurst where a statute made clear that some conditions were placed on the receipt of federal funds, and stating that Congress need not “specifically identif[y] and proscrib[e]” each condition in the legislation). Simply put, “ Pennhurst does not bar a private damages action under Title IX where the funding recipient engages in intentional conduct that violates the clear terms of the statute.” Davis, 526 U. S., at 642.
Thus, the Board should have been put on notice by the fact that our cases since Cannon, such as Gebser and Davis, have consistently interpreted Title IX’s private cause of action broadly to encompass diverse forms of intentional sex discrimination. Indeed, retaliation presents an even easier case than deliberate indifference. It is easily attributable to the funding recipient, and it is always—by definition—intentional. We therefore conclude that retaliation against individuals because they complain of sex discrimination is “intentional conduct that violates the clear terms of the statute,” Davis, 526 U. S., at 642, and that Title IX itself therefore supplied sufficient notice to the Board that it could not retaliate against Jackson after he complained of discrimination against the girls’ basketball team.
The regulations implementing Title IX clearly prohibit retaliation and have been on the books for nearly 30 years. Cf., e.g., id., at 643 (holding that Title IX’s regulatory scheme “has long provided funding recipients with notice that they may be liable for their failure to respond to the discriminatory acts of certain nonagents”). More importantly, the Courts of Appeals that had considered the question at the time of the conduct at issue in this case all had already interpreted Title IX to cover retaliation. See, e.g., Lowrey, 117 F. 3d, at 252; Preston, 31 F. 3d, at 206. The Board could not have realistically supposed that, given this context, it remained free to retaliate against those who reported sex discrimination. Cf. Davis, supra, at 644 (stating that the common law of torts “has put schools on notice that they may be held responsible under state law for their failure to protect students from the tortious acts of third parties”). A reasonable school board would realize that institutions covered by Title IX cannot cover up violations of that law by means of discriminatory retaliation.
To prevail on the merits, Jackson will have to prove that the Board retaliated against him because he complained of sex discrimination. The amended complaint alleges that the Board retaliated against Jackson for complaining to his supervisor, Ms. Evelyn Baugh, about sex discrimination at Ensley High School. At this stage of the proceedings, “[t]he issue is not whether a plaintiff will ultimately prevail but whether the claimant is entitled to offer evidence to support the claims.” Scheuer v. Rhodes, 416 U. S. 232, 236 (1974). Accordingly, the judgment of the Court of Appeals for the Eleventh Circuit is reversed, and the case is remanded for further proceedings consistent with this opinion.
It is so ordered.
Notes
Footnote 1
Justice Thomas contends that Sullivan merely decided that the white owner had standing to assert the rights of the black lessee. Post, at 11 (dissenting opinion). But Sullivan ’s holding was not so limited. It plainly held that the white owner could maintain his own private cause of action under §1982 if he could show that he was “punished for trying to vindicate the rights of minorities.” 396 U. S., at 237.
Footnote 2
We agree with Justice Thomas that plaintiffs may not assert claims under Title IX for conduct not prohibited by that statute. Post, at 10–11 (dissenting opinion). See also Central Bank of Denver, N. A. v. First Interstate Bank of Denver, N. A., 511 U. S. 164, 173 (1994) (“[T]he private plaintiff may not bring a 10b–5 suit against a defendant for acts not prohibited by the text of §10(b)”). But we part ways with regard to our reading of the statute. We interpret Title IX’s text to clearly prohibit retaliation for complaints about sex discrimination.
Footnote 3
Justice Thomas contends that “extending the implied cause of action under Title IX to claims of retaliation expands the class of people the statute protects beyond the specific beneficiaries.” Post, at 11 (dissenting opinion). But Title IX’s beneficiaries plainly include all those who are subjected to “discrimination” “on the basis of sex.” 20 U. S. C. §1681(a). Because, as we explain above, see supra, at 4–5, retaliation in response to a complaint about sex discrimination is “discrimination” “on the basis of sex,” the statute clearly protects those who suffer such retaliation. The following hypothetical, offered by petitioner at oral argument, illustrates this point: If the male captain of the boys’ basketball team and the female captain of the girls’ basketball team together approach the school principal to complain about discrimination against the girls’ team, and the principal retaliates by expelling them both from the honor society, then both the female and the male captains have been “discriminated” against “on the basis of sex.” Tr. of Oral Arg. at 53–54.
Justice O’Connor, with whom Justice Kennedy and Justice Thomas join, concurring in the judgment.
“Disparate treatment … captures the essence of what Congress sought to prohibit in the [Age Discrimination in Employment Act of 1967 (ADEA), 29 U. S. C. §621 et seq. ] It is the very essence of age discrimination for an older employee to be fired because the employer believes that productivity and competence decline with old age.” Hazen Paper Co. v. Biggins, 507 U. S. 604, 610 (1993). In the nearly four decades since the ADEA’s enactment, however, we have never read the statute to impose liability upon an employer without proof of discriminatory intent. See ibid.; Markham v. Geller, 451 U. S. 945 (1981) (Rehnquist, J., dissenting from denial of certiorari). I decline to join the Court in doing so today.
I would instead affirm the judgment below on the ground that disparate impact claims are not cognizable under the ADEA. The ADEA’s text, legislative history, and purposes together make clear that Congress did not intend the statute to authorize such claims. Moreover, the significant differences between the ADEA and Title VII of the Civil Rights Act of 1964 counsel against transposing to the former our construction of the latter in Griggs v. Duke Power Co., 401 U. S. 424 (1971). Finally, the agencies charged with administering the ADEA have never authoritatively construed the statute’s prohibitory language to impose disparate impact liability. Thus, on the precise question of statutory interpretation now before us, there is no reasoned agency reading of the text to which we might defer.
I
A
Our starting point is the statute’s text. Section 4(a) of the ADEA makes it unlawful for an employer:
“(1) to fail or refuse to hire or to discharge any individual or otherwise discriminate against any individual with respect to his compensation, terms, conditions, or privileges of employment, because of such individual’s age; [or]
“(2) to limit, segregate, or classify his employees in any way which would deprive or tend to deprive any individual of employment opportunities or otherwise adversely affect his status as an employee, because of such individual’s age ….” 29 U. S. C. §623(a).
Neither petitioners nor the plurality contend that the first paragraph, §4(a)(1), authorizes disparate impact claims, and I think it obvious that it does not. That provision plainly requires discriminatory intent, for to take an action against an individual “ because of such individual’s age” is to do so “by reason of” or “on account of” her age. See Webster’s Third New International Dictionary 194 (1961); see also Teamsters v. United States, 431 U. S. 324, 335–336, n. 15 (1977) (“ ‘Disparate treatment’ … is the most easily understood type of discrimination. The employer simply treats some people less favorably than others because of their [protected characteristic]. Proof of discriminatory motive is critical” (emphasis added)).
Petitioners look instead to the second paragraph, §4(a)(2), as the basis for their disparate impact claim. But petitioners’ argument founders on the plain language of the statute, the natural reading of which requires proof of discriminatory intent. Section 4(a)(2) uses the phrase “because of … age” in precisely the same manner as does the preceding paragraph—to make plain that an employer is liable only if its adverse action against an individual is motivated by the individual’s age.
Paragraphs (a)(1) and (a)(2) do differ in one informative respect. The employer actions targeted by paragraph (a)(1)— i.e., refusing to hire, discharging, or discriminating against—are inherently harmful to the targeted individual. The actions referred to in paragraph (a)(2), on the other hand— i.e., limiting, segregating, or classifying—are facially neutral. Accordingly, paragraph (a)(2) includes additional language which clarifies that, to give rise to liability, the employer’s action must actually injure someone: The decision to limit, segregate, or classify employees must “deprive or tend to deprive [an] individual of employment opportunities or otherwise adversely affect his status as an employee.” That distinction aside, the structures of paragraphs (a)(1) and (a)(2) are otherwise identical. Each paragraph prohibits an employer from taking specified adverse actions against an individual “because of such individual’s age.”
The plurality instead reads paragraph (a)(2) to prohibit employer actions that “adversely affect [an individual’s] status as an employe[e] because of such individual’s age.” Under this reading, “because of … age” refers to the cause of the adverse effect rather than the motive for the employer’s action. See ante, at 6. This reading is unpersuasive for two reasons. First, it ignores the obvious parallel between paragraphs (a)(1) and (a)(2) by giving the phrase “because of such individual’s age” a different meaning in each of the two paragraphs. And second, it ignores the drafters’ use of a comma separating the “because of … age” clause from the preceding language. That comma makes plain that the “because of … age” clause should not be read, as the plurality would have it, to modify only the “adversely affect” phrase. See, e.g., United States v. Ron Pair Enterprises, Inc., 489 U. S. 235, 241 (1989) (interpreting statute in light of the drafters’ use of a comma to set aside a particular phrase from the following language); see also B. Garner, A Dictionary of Modern Legal Usage 101 (2d ed. 1995) (“Generally, the word because should not follow a comma”). Rather, the “because of … age” clause is set aside to make clear that it modifies the entirety of the preceding paragraph: An employer may not, because of an individual’s age, limit, segregate, or classify his employees in a way that harms that individual.
The plurality also argues that its reading is supported by the supposed “incongruity” between paragraph (a)(2)’s use of the plural in referring to the employer’s actions (“limit, segregate, or classify his employees ”) and its use of the singular in the “because of such individual’s age” clause. (Emphases added.) Ante, at 7, n. 6. Not so. For the reasons just stated, the “because of… age” clause modifies all of the preceding language of paragraph (a)(2). That preceding language is phrased in both the plural (insofar as it refers to the employer’s actions relating to employees ) and the singular (insofar as it requires that such action actually harm an individual ). The use of the singular in the “because of… age” clause simply makes clear that paragraph (a)(2) forbids an employer to limit, segregate, or classify his employees if that decision is taken because of even one employee’s age and that individual (alone or together with others) is harmed.
B
While §4(a)(2) of the ADEA makes it unlawful to intentionally discriminate because of age, §4(f)(1) clarifies that “[i]t shall not be unlawful for an employer … to take any action otherwise prohibited under subsections (a), (b), (c), or (e) of this section … where the differentiation is based on reasonable factors other than age ….” 29 U. S. C. §623(f)(1). This “reasonable factors other than age” (RFOA) provision “insure[s] that employers [are] permitted to use neutral criteria” other than age, EEOC v. Wyoming, 460 U. S. 226, 232–233 (1983), even if this results in a disparate adverse impact on older workers. The provision therefore expresses Congress’ clear intention that employers not be subject to liability absent proof of intentional age-based discrimination. That policy, in my view, cannot easily be reconciled with the plurality’s expansive reading of §4(a)(2).
The plurality however, reasons that the RFOA provision’s language instead confirms that §4(a) authorizes disparate impact claims. If §4(a) prohibited only intentional discrimination, the argument goes, then the RFOA provision would have no effect because any action based on a factor other than age would not be “ ‘otherwise prohibited’ ” under §4(a). See ante, at 9–10. Moreover, the plurality says, the RFOA provision applies only to employer actions based on reasonable factors other than age—so employers may still be held liable for actions based on un reasonable nonage factors. See ante, at 10.
This argument misconstrues the purpose and effect of the RFOA provision. Discriminatory intent is required under §4(a), for the reasons discussed above. The role of the RFOA provision is to afford employers an independent safe harbor from liability. It provides that, where a plaintiff has made out a prima facie case of intentional age discrimination under §4(a)—thus “creat[ing] a presumption that the employer unlawfully discriminated against the employee,” Texas Dept. of Community Affairs v. Burdine, 450 U. S. 248, 254 (1981)—the employer can rebut this case by producing evidence that its action was based on a reasonable nonage factor. Thus, the RFOA provision codifies a safe harbor analogous to the “legitimate, nondiscriminatory reason” (LNR) justification later recognized in Title VII suits. Ibid.; McDonnell Douglas Corp. v. Green, 411 U. S. 792, 802 (1973).
Assuming the McDonnell Douglas framework applies to ADEA suits, see O’Connor v. Consolidated Coin Caterers Corp., 517 U. S. 308, 311 (1996), this “rebuttal” function of the RFOA provision is arguably redundant with the judicially established LNR justification. See ante, at 9–10. But, at most, that merely demonstrates Congress’ abundance of caution in codifying an express statutory exemption from liability in the absence of discriminatory intent. See Fort Stewart Schools v. FLRA, 495 U. S. 641, 646 (1990) (provisions that, although “technically unnecessary,” are sometimes “inserted out of an abundance of caution—a drafting imprecision venerable enough to have left its mark on legal Latin (ex abundanti cautela) ”). It is noteworthy that even after McDonnell Douglas was decided, lower courts continued to rely on the RFOA exemption, in lieu of the LNR justification, as the basis for rebutting a prima facie case of age discrimination. See, e.g., Krieg v. Paul Revere Life Ins. Co., 718 F. 2d 998, 999 (CA11 1983) (per curiam); Schwager v. Sun Oil Co. of Pa., 591 F. 2d 58, 61 (CA10 1979); Bittar v. Air Canada, 512 F. 2d 582, 582–583 (CA5 1975) (per curiam).
In any event, the RFOA provision also plays a distinct (and clearly nonredundant) role in “mixed-motive” cases. In such cases, an adverse action taken in substantial part because of an employee’s age may be “otherwise prohibited” by §4(a). See Desert Palace, Inc. v. Costa, 539 U. S. 90, 93 (2003); Price Waterhouse v. Hopkins, 490 U. S. 228, 262–266 (1989) (O’Connor, J., concurring in judgment). The RFOA exemption makes clear that such conduct is nevertheless lawful so long as it is “based on” a reasonable factor other than age.
Finally, the RFOA provision’s reference to “reasonable” factors serves only to prevent the employer from gaining the benefit of the statutory safe harbor by offering an irrational justification. Reliance on an unreasonable nonage factor would indicate that the employer’s explanation is, in fact, no more than a pretext for intentional discrimination. See Reeves v. Sanderson Plumbing Products, Inc., 530 U. S. 133, 147 (2000); see also Hazen Paper, 507 U. S., at 613–614.
II
The legislative history of the ADEA confirms what its text plainly indicates—that Congress never intended the statute to authorize disparate impact claims. The drafters of the ADEA and the Congress that enacted it understood that age discrimination was qualitatively different from the kinds of discrimination addressed by Title VII, and that many legitimate employment practices would have a disparate impact on older workers. Accordingly, Congress determined that the disparate impact problem would best be addressed through noncoercive measures, and that the ADEA’s prohibitory provisions should be reserved for combating intentional age-based discrimination.
A
Although Congress rejected proposals to address age discrimination in the Civil Rights Act of 1964, §715 of that Act directed the Secretary of Labor to undertake a study of age discrimination in employment and to submit to Congress a report containing “such recommendations for legislation to prevent arbitrary discrimination in employment because of age as he determines advisable,” 78 Stat. 265. See General Dynamics Land Systems, Inc. v. Cline, 540 U. S. 581, 586–587 (2004); EEOC v. Wyoming, supra, at 229. In response, Secretary Willard Wirtz submitted the report that provided the blueprint for the ADEA. See Report of the Secretary of Labor, The Older American Worker: Age Discrimination in Employment (June 1965), reprinted in U. S. Equal Employment Opportunity Commission, Legislative History of the Age Discrimination in Employment Act 83 (1981) (hereinafter Wirtz Report or Report). Because the ADEA was modeled on the Wirtz Report’s findings and recommendations, the Report provides critical insights into the statute’s meaning. See generally Blumrosen, Interpreting the ADEA: Intent or Impact 14–20, in Age Discrimination in Employment Act: A Compliance Manual for Lawyers and Personnel Practitioners 83–89 (M. Lake ed. 1982); see also General Dynamics, supra, at 587–590 (relying on the Wirtz Report to interpret the ADEA); EEOC v. Wyoming, 460 U. S., at 230–231 (discussing the Report’s role in the drafting of the ADEA).
The Wirtz Report reached two conclusions of central relevance to the question presented by this case. First, the Report emphasized that age discrimination is qualitatively different from the types of discrimination prohibited by Title VII of the Civil Rights Act of 1964 ( i.e., race, color, religion, sex, and national origin discrimination). Most importantly—in stark contrast to the types of discrimination addressed by Title VII—the Report found no evidence that age discrimination resulted from intolerance or animus towards older workers. Rather, age discrimination was based primarily upon unfounded assumptions about the relationship between an individual’s age and her ability to perform a job. Wirtz Report 2. In addition, whereas ability is nearly always completely unrelated to the characteristics protected by Title VII, the Report found that, in some cases, “there is in fact a relationship between [an individual’s] age and his ability to perform the job.” Ibid. (emphasis deleted).
Second, the Wirtz Report drew a sharp distinction between “ ‘arbitrary discrimination’ ” (which the Report clearly equates with disparate treatment) and circumstances or practices having a disparate impact on older workers. See id., at 2, 21–22. The Report defined “arbitrary” discrimination as adverse treatment of older workers “because of assumptions about the effect of age on their ability to do a job when there is in fact no basis for these assumptions.” Id., at 2 (emphasis in original). While the “most obvious kind” of arbitrary discrimination is the setting of unjustified maximum age limits for employment, id., at 6, naturally the Report’s definition encompasses a broad range of disparate treatment.
The Report distinguished such “arbitrary” ( i.e., intentional and unfounded) discrimination from two other phenomena. One involves differentiation of employees based on a genuine relationship between age and ability to perform a job. See id., at 2. In this connection, the Report examined “circumstances which unquestionably affect older workers more strongly, as a group, than they do younger workers,” including questions of health, educational attainment, and technological change. Id., at 11–14.[ Footnote 1 ] In addition, the Report assessed “institutional arrangements”—such as seniority rules, workers’ compensation laws, and pension plans—which, though intended to benefit older workers, might actually make employers less likely to hire or retain them. Id., at 2, 15–17.
The Report specifically recommended legislative action to prohibit “arbitrary discrimination,” i.e., disparate treatment. Id., at 21–22. In sharp contrast, it recommended that the other two types of “discrimination”—both involving factors or practices having a disparate impact on older workers—be addressed through noncoercive measures: programs to increase the availability of employment; continuing education; and adjustment of pension systems, workers’ compensation, and other institutional arrangements. Id., at 22–25. These recommendations found direct expression in the ADEA, which was drafted at Congress’ command that the Secretary of Labor make “specific legislative recommendations for implementing the [Wirtz Report’s] conclusions,” Fair Labor Standards Amendments of 1966, §606, 80 Stat. 845. See also General Dynamics, 540 U. S., at 589 (“[T]he ADEA … begins with statements of purpose and findings that mirror the Wirtz Report”).
B
The ADEA’s structure confirms Congress’ determination to prohibit only “arbitrary” discrimination ( i.e., disparate treatment based on unfounded assumptions), while addressing practices with a disparate adverse impact on older workers through noncoercive measures. Section 2—which sets forth the findings and purposes of the statute—draws a clear distinction between “the setting of arbitrary age limits regardless of potential for job performance” and “certain otherwise desirable practices [that] may work to the disadvantage of older persons.” 29 U. S. C. §621(a)(2). In response to these problems, §2 identifies three purposes of the ADEA: “[1] to promote employment of older persons based on their ability rather than age; [2] to prohibit arbitrary age discrimination in employment; [and 3] to help employers and workers find ways of meeting problems arising from the impact of age on employment.” §621(b).
Each of these three purposes corresponds to one of the three substantive statutory sections that follow. Section 3 seeks to “promote employment of older persons” by directing the Secretary of Labor to undertake a program of research and education related to “the needs and abilities of older workers, and their potentials for continued employment and contribution to the economy.” §622(a). Section 4, which contains the ADEA’s core prohibitions, corresponds to the second purpose: to “prohibit arbitrary age discrimination in employment.” Finally, §5 addresses the third statutory purpose by requiring the Secretary of Labor to undertake a study of “institutional and other arrangements giving rise to involuntary retirement” and to submit any resulting findings and legislative recommendations to Congress. §624(a)(1).
Section 4—including §4(a)(2)—must be read in light of the express statutory purpose the provision was intended to effect: the prohibition of “arbitrary age discrimination in employment.” §621(b). As the legislative history makes plain, “arbitrary” age discrimination had a very specific meaning for the ADEA’s drafters. It meant disparate treatment of older workers, predominantly because of unfounded assumptions about the relationship between age and ability. See supra, at 8–10. Again, such intentional discrimination was clearly distinguished from circumstances and practices merely having a disparate impact on older workers, which—as ADEA §§2, 3, and 5 make clear—Congress intended to address through research, education, and possible future legislative action.
C
In addition to this affirmative evidence of congressional intent, I find it telling that the legislative history is devoid of any discussion of disparate impact claims or of the complicated issues such claims raise in the ADEA context. See Gold, Disparate Impact Under the Age Discrimination in Employment Act of 1967, 25 Berkeley J. Emp. & Lab. L. 1, 40 (2004). At the time the ADEA was enacted, the predominant focus of antidiscrimination law was on intentional discrimination; the concept of disparate impact liability, by contrast, was quite novel. See, e.g., Gold, Griggs ’ Folly: An Essay on the Theory, Problems, and Origin of the Adverse Impact Definition of Employment Discrimination and a Recommendation for Reform, 7 Indus. Rel. L. J. 429, 518–520 (1985); Blumrosen, Strangers in Paradise: Griggs v. Duke Power Co. and the Concept of Employment Discrimination, 71 Mich. L. Rev. 59, 69–71 (1972). Had Congress intended to inaugurate disparate impact liability in the ADEA, one would expect to find some indication of that intent in the text and the legislative history. There is none.
D
Congress’ decision not to authorize disparate impact claims is understandable in light of the questionable utility of such claims in the age-discrimination context. No one would argue that older workers have suffered disadvantages as a result of entrenched historical patterns of discrimination, like racial minorities have. See Massachusetts Bd. of Retirement v. Murgia, 427 U. S. 307, 313–314 (1976) (per curiam); see also Wirtz Report 5–6. Accordingly, disparate impact liability under the ADEA cannot be justified, and is not necessary, as a means of redressing the cumulative results of past discrimination. Cf. Griggs, 401 U. S., at 430 (reasoning that disparate impact liability is necessary under Title VII to prevent perpetuation of the results of past racial discrimination).
Moreover, the Wirtz Report correctly concluded that—unlike the classifications protected by Title VII—there often is a correlation between an individual’s age and her ability to perform a job. Wirtz Report 2, 11–15. That is to be expected, for “physical ability generally declines with age,” Murgia, supra, at 315, and in some cases, so does mental capacity, see Gregory v. Ashcroft, 501 U. S. 452, 472 (1991). Perhaps more importantly, advances in technology and increasing access to formal education often leave older workers at a competitive disadvantage vis-ŕ-vis younger workers. Wirtz Report 11–15. Beyond these performance-affecting factors, there is also the fact that many employment benefits, such as salary, vacation time, and so forth, increase as an employee gains experience and seniority. See, e.g., Finnegan v. Trans World Airlines, Inc., 967 F. 2d 1161, 1164 (CA7 1992) (“[V]irtually all elements of a standard compensation package are positively correlated with age”). Accordingly, many employer decisions that are intended to cut costs or respond to market forces will likely have a disproportionate effect on older workers. Given the myriad ways in which legitimate business practices can have a disparate impact on older workers, it is hardly surprising that Congress declined to subject employers to civil liability based solely on such effects.
III
The plurality and Justice Scalia offer two principal arguments in favor of their reading of the statute: that the relevant provision of the ADEA should be read in pari materia with the parallel provision of Title VII, and that we should give interpretive weight or deference to agency statements relating to disparate impact liability. I find neither argument persuasive.
A
The language of the ADEA’s prohibitory provisions was modeled on, and is nearly identical to, parallel provisions in Title VII. See McKennon v. Nashville Banner Publishing Co., 513 U. S. 352, 357 (1995); Lorillard v. Pons, 434 U. S. 575, 584 (1978). Because Griggs, supra, held that Title VII’s §703(a)(2) permits disparate impact claims, the plurality concludes that we should read §4(a)(2) of the ADEA similarly. Ante, at 4–9.
Obviously, this argument would be a great deal more convincing had Griggs been decided before the ADEA was enacted. In that case, we could safely assume that Congress had notice (and therefore intended) that the language at issue here would be read to authorize disparate impact claims. See, e.g., Department of Energy v. Ohio, 503 U. S. 607, 626 (1992); Holmes v. Securities Investor Protection Corporation, 503 U. S. 258, 268 (1992). But Griggs was decided four years after the ADEA’s enactment, and there is no reason to suppose that Congress in 1967 could have foreseen the interpretation of Title VII that was to come. See Fogerty v. Fantasy, Inc., 510 U. S. 517, 523, n. 9 (1994); see also supra, at 10–11 (discussing novelty of disparate impact theory at the time of the ADEA’s enactment).
To be sure, where two statutes use similar language we generally take this as “a strong indication that [they] should be interpreted pari passu.” Northcross v. Board of Ed. of Memphis City Schools, 412 U. S. 427, 428 (1973) (per curiam). But this is not a rigid or absolute rule, and it “ ‘readily yields’ ” to other indicia of congressional intent. General Dynamics, 540 U. S., at 595 (quoting Atlantic Cleaners & Dyers, Inc. v. United States, 286 U. S. 427, 433 (1932)). Indeed, “ ‘the meaning [of the same words] well may vary to meet the purposes of the law.’ ” United States v. Cleveland Indians Baseball Co., 532 U. S. 200, 213 (2001) (alteration in original) (quoting Atlantic Cleaners & Dyers, supra, at 433). Accordingly, we have not hesitated to give a different reading to the same language—whether appearing in separate statutes or in separate provisions of the same statute—if there is strong evidence that Congress did not intend the language to be used uniformly. See, e.g., General Dynamics, supra, at 595–597 (“age” has different meaning where used in different parts of the ADEA); Cleveland Indians, supra, at 213 (“wages paid” has different meanings in different provisions of Title 26 U. S. C.); Robinson v. Shell Oil Co., 519 U. S. 337, 343–344 (1997) (“employee” has different meanings in different parts of Title VII); Fogerty, supra, at 522–525 (Copyright Act’s attorney’s fees provision has different meaning than the analogous provision in Title VII, despite their “virtually identical language”). Such is the case here.
First, there are significant textual differences between Title VII and the ADEA that indicate differences in congressional intent. Most importantly, whereas the ADEA’s RFOA provision protects employers from liability for any actions not motivated by age, see supra, at 4–7, Title VII lacks any similar provision. In addition, the ADEA’s structure demonstrates Congress’ intent to combat intentional discrimination through §4’s prohibitions while addressing employment practices having a disparate impact on older workers through independent noncoercive mechanisms. See supra, at 8–11. There is no analogy in the structure of Title VII. Furthermore, as the Congresses that adopted both Title VII and the ADEA clearly recognized, the two statutes were intended to address qualitatively different kinds of discrimination. See supra, at 7–8. Disparate impact liability may have a legitimate role in combating the types of discrimination addressed by Title VII, but the nature of aging and of age discrimination makes such liability inappropriate for the ADEA. See supra, at 12–13.
Finally, nothing in the Court’s decision in Griggs itself provides any reason to extend its holding to the ADEA. As the plurality tacitly acknowledges, ante, at 6, the decision in Griggs was not based on any analysis of Title VII’s actual language. Rather, the ratio decidendi was the statute’s perceived purpose, i.e.,
“to achieve equality of employment opportunities and remove barriers that have operated in the past to favor an identifiable group of white employees over other employees. Under the Act, practices, procedures, or tests neutral on their face, and even neutral in terms of intent, cannot be maintained if they operate to ‘freeze’ the status quo of prior discriminatory employment practices.” 401 U. S., at 429–430.
In other words, the Court in Griggs reasoned that disparate impact liability was necessary to achieve Title VII’s ostensible goal of eliminating the cumulative effects of historical racial discrimination. However, that rationale finds no parallel in the ADEA context, see Murgia, 427 U. S., at 313–314, and it therefore should not control our decision here.
Even venerable canons of construction must bow, in an appropriate case, to compelling evidence of congressional intent. In my judgment, the significant differences between Title VII and the ADEA are more than sufficient to overcome the default presumption that similar language is to be read similarly. See Fogerty, supra, at 523–524 (concluding that the “normal indication” that similar language should be read similarly is “overborne” by differences between the legislative history and purposes of two statutes).
B
The plurality asserts that the agencies charged with the ADEA’s administration “have consistently interpreted the [statute] to authorize relief on a disparate-impact theory.” Ante, at 10. In support of this claim, the plurality describes a 1968 interpretive bulletin issued by the Department of Labor as “permitt[ing]” disparate impact claims. Ibid. (citing 29 CFR §860.103(f)(1)(i) (1970)). And the plurality cites, without comment, an Equal Employment Opportunities Commission (EEOC) policy statement construing the RFOA provision. Ante, at 11 (citing 29 CFR §1625.7 (2004)) . It is unclear what interpretive value the plurality means to assign to these agency statements. But Justice Scalia, at least, thinks that the EEOC statement is entitled to deference under Chevron U. S. A. Inc. v. Natural Resources Defense Council, Inc., 467 U. S. 837 (1984), and that “that is sufficient to resolve this case.” Ante, at 5 (opinion concurring in part and concurring in judgment). I disagree and, for the reasons that follow, would give no weight to the statements in question.
The 1968 Labor Department bulletin to which the plurality alludes was intended to “provide ‘a practical guide to employers and employees as to how the office representing the public interest in its enforcement will seek to apply it.’ ” 29 CFR §860.1 (1970) (quoting Skidmore v. Swift & Co., 323 U. S. 134, 138 (1944)). In discussing the RFOA provision, the bulletin states that “physical fitness requirements” and “[e]valuation factors such as quantity or quality of production, or educational level” can qualify as reasonable nonage factors, so long as they have a valid relationship to job qualifications and are uniformly applied. §§860.103(f)(1), (2). But the bulletin does not construe the ADEA’s prohibitory provisions, nor does it state or imply that §4(a) authorizes disparate impact claims. Rather, it establishes “a nonexclusive objective test for employers to use in determining whether they could be certain of qualifying for the” RFOA exemption. Public Employees Retirement System of Ohio v. Betts, 492 U. S. 158, 172 (1989) (discussing 1968 bulletin’s interpretation of the §4(f)(2) exemption). Moreover, the very same bulletin states unequivocally that “[t]he clear purpose [of the ADEA] is to insure that age, within the limits prescribed by the Act, is not a determining factor in making any decision regarding the hiring, dismissal, promotion or any other term condition or privilege of employment of an individual.” §860.103(c) (emphasis added). That language is all about discriminatory intent.
The EEOC statement cited by the plurality and relied upon by Justice Scalia is equally unhelpful. This “interpretative rule or policy statement,” promulgated in 1981, superseded the 1968 Labor Department bulletin after responsibility for enforcing the ADEA was transferred from Labor to the EEOC. See 46 Fed. Reg. 47724 (1981). It states, in relevant part:
“[W]hen an employment practice, including a test, is claimed as a basis for different treatment of employees or applicants for employment on the grounds that it is a ‘factor other than’ age, and such a practice has an adverse impact on individuals within the protected age group, it can only be justified as a business necessity.” 29 CFR §1625.7(d) (2004).
Like the 1968 bulletin it replaces, this statement merely spells out the agency’s view, for purposes of its enforcement policy, of what an employer must do to be certain of gaining the safety of the RFOA haven. It says nothing about whether disparate impact claims are authorized by the ADEA.
For Justice Scalia, “[t]his is an absolutely classic case for deference to agency interpretation.” Ante, at 1 (opinion concurring in part and concurring in judgment). I disagree. Under Chevron, we will defer to a reasonable agency interpretation of ambiguous statutory language, see 467 U. S., at 843–844, provided that the interpretation has the requisite “force of law,” Christensen v. Harris County, 529 U. S.576, 587 (2000). The rationale for such deference is that Congress has explicitly or implicitly delegated to the agency responsible for administering a statute the authority to choose among permissible constructions of ambiguous statutory text. See Chevron, supra, at 844. The question now before us is not what it takes to qualify for the RFOA exemption, but rather whether §4(a)(2) of the ADEA authorizes disparate impact claims. But the EEOC statement does not purport to interpret the language of §4(a) at all. Quite simply, the agency has not actually exercised its delegated authority to resolve any ambiguity in the relevant provision’s text, much less done so in a reasonable or persuasive manner. As to the specific question presented, therefore, the regulation is not entitled to any deference. See John Hancock Mut. Life Ins. Co. v. Harris Trust and Sav. Bank, 510 U. S. 86, 106–109, and n. 17 (1993); see also SEC v. Sloan, 436 U. S. 103, 117–118 (1978); Adamo Wrecking Co. v. United States, 434 U. S. 275, 287–289, and n. 5 (1978).[ Footnote 2 ]
Justice Scalia’s attempt to link the EEOC’s RFOA regulation to §4(a)(2) is premised on a dubious chain of inferences that, in my view, highlights the hazards of his approach. Because the RFOA provision is “relevant only as a response to employer actions ‘otherwise prohibited’ by the ADEA,” he reasons, the “unavoidable meaning” of the EEOC statement is that the agency “interprets the ADEA to prohibit employer actions that have an ‘adverse impact on individuals within the protected age group.’ ” Ante, at 4 (opinion concurring in part and concurring in judgment) (quoting 29 CFR §1625.7(d) (2004)). But, of course, disparate treatment clearly has an “adverse impact on individuals within the protected age group,” ibid., and Justice Scalia’s reading of the EEOC’s rule is hardly “unavoidable.” The regulation says only that if an employer wants to rely on a practice—say, a physical fitness test—as the basis for an exemption from liability, and that test adversely affects older workers, the employer can be sure of qualifying for the exemption only if the test is sufficiently job related. Such a limitation makes sense in disparate treatment cases. A test that harms older workers and is unrelated to the job may be a pretext for—or even a means of effectuating—intentional discrimination. See supra, at 6–7. Justice Scalia completes his analytical chain by inferring that the EEOC regulation must be read to interpret §4(a)(2) to allow disparate impact claims because that is the only provision of the ADEA that could “conceivably” be so interpreted. Ante, at 4 (opinion concurring in part and concurring judgment). But the support for that inference is doubtful, to say the least. The regulation specifically refers to employment practices claimed as a basis for “different treatment of employees or applicants for employment,” 29 CFR §1625.7(d) (2004) (emphasis added). Section 4(a)(2), of course, does not apply to “applicants for employment” at all—it is only §4(a)(1) that protects this group. See 29 U. S. C. §623(a). That suggests that the EEOC must have read the RFOA to provide a defense against claims under §4(a)(1)—which unquestionably permits only disparate treatment claims, see supra, at 2.
This discussion serves to illustrate why it makes little sense to attribute to the agency a construction of the relevant statutory text that the agency itself has not actually articulated so that we can then “defer” to that reading. Such an approach is particularly troubling where applied to a question as weighty as whether a statute does or does not subject employers to liability absent discriminatory intent. This is not, in my view, what Chevron contemplated.
As an interpretation of the RFOA provision, moreover, the EEOC regulation is both unreasonable on its face and directly at odds with the Court’s holding in today’s case. It says that the RFOA exemption is available only if the employer’s practice is justified by a “business necessity.” But the Court has rejected that reading of the RFOA provision, and rightly so: There may be many “reasonable” means by which an employer can advance its goals, and a given nonage factor can certainly be “reasonable” without being necessary. Ante, at 14; see also Western Air Lines, Inc. v. Criswell, 472 U. S. 400, 419 (1985) (distinguishing “ ‘reasonable necessity’ ” standard from “ ‘reasonableness’ ”). Of course, it is elementary that “no deference is due to agency interpretations at odds with the plain language of the statute itself.” Betts, 492 U. S., at 171. The agency clearly misread the RFOA provision it was attempting to construe. That error is not necessarily dispositive of the disparate impact question. But I think it highlights the improvidence of giving weight (let alone deferring) to the regulation’s purported assumption that an entirely different provision of the statute, which is not even the subject of the regulation, authorizes disparate impact claims. In my view, we should simply acknowledge that this regulation is of no help in answering the question presented.
IV
Although I would not read the ADEA to authorize disparate impact claims, I agree with the Court that, if such claims are allowed, they are strictly circumscribed by the RFOA exemption. See ante, at 13–14. That exemption requires only that the challenged employment practice be based on a “reasonable” nonage factor—that is, one that is rationally related to some legitimate business objective. I also agree with the Court, ante, at 11, that, if disparate impact claims are to be permitted under the ADEA, they are governed by the standards set forth in our decision in Wards Cove Packing Co. v. Atonio, 490 U. S. 642 (1989). That means, as the Court holds, ante, at 12, that “a plaintiff must demonstrate that it is the application of a specific or particular employment practice that has created the disparate impact under attack,” Wards Cove, supra, at 657 (emphasis added); see also Watson v. Fort Worth Bank & Trust, 487 U. S. 977, 994 (1988) (opinion of O’Connor, J.). It also means that once the employer has produced evidence that its action was based on a reasonable nonage factor, the plaintiff bears the burden of disproving this assertion. See Wards Cove, supra, at 659–660; see also Watson, supra, at 997 (opinion of O’Connor, J.). Even if petitioners’ disparate impact claim were cognizable under the ADEA, that claim clearly would fail in light of these requirements.
Notes
Footnote 1
It is in this connection that the Report refers to formal employment standards requiring a high school diploma. See Wirtz Report 3. The Wirtz Report did say that such a requirement would be “unfair” if an older worker’s years of experience had given him an equivalent education. Ibid. But the plurality is mistaken to find in this statement a congressional “goal” of eliminating job requirements with a disparate impact on older workers. See ante, at 6, n. 5. Rather, the Wirtz Report discussed the diploma requirement in the context of a broader discussion of the effects of “wholly impersonal forces—most of them part of what is properly, if sometimes too casually, called ‘progress.’ ” Wirtz Report 3. These forces included “the pace of changing technology, changing jobs, changing educational requirements, and changing personnel practices,” which “increase[d] the need for special efforts if older workers’ employment prospects are to improve significantly.” Ibid. (emphasis added); see also id., at 11–15 (discussing the educational attainments of older workers, together with health and technological change, in a section entitled “The Necessary Recognition of Forces of Circumstance”). The Report recommended that such forces be addressed through noncoercive instead of prohibitory measures, and it specifically focused on the need for educational opportunities for older workers. See id., at 23–25.
Footnote 2
Because the EEOC regulation does not actually interpret the text at issue, we need not address the degree of deference to which the regulation would otherwise be entitled. Cf. General Dynamics Land Systems, Inc. v. Cline, 540 U. S. 581, 600 (2004) (declining to address whether EEOC’s regulations interpreting the ADEA are entitled to Chevron deference).
Justice O’Connor delivered the opinion of the Court.
We confront here the problem of a “mixed” petition for habeas corpus relief in which a state prisoner presents a federal court with a single petition containing some claims that have been exhausted in the state courts and some that have not. More precisely, we consider whether a federal district court has discretion to stay the mixed petition to allow the petitioner to present his unexhausted claims to the state court in the first instance, and then to return to federal court for review of his perfected petition.
I
Petitioner Charles Russell Rhines was convicted in South Dakota state court of first-degree murder and third-degree burglary and sentenced to death. His conviction became final on December 2, 1996, when we denied his initial petition for certiorari. Rhines v. South Dakota, 519 U. S. 1013. On December 5, 1996, Rhines filed a petition for state habeas corpus. App. 32. The state court denied his petition, and the Supreme Court of South Dakota affirmed on February 9, 2000, Rhines v. Weber, 2002 SD 19, 608 N. W. 2d 303. Rhines filed his pro se petition for federal habeas corpus pursuant to 28 U. S. C. §2254 in the United States District Court for the District of South Dakota on February 22, 2000. App. 3. Because the 1-year statute of limitations imposed by the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA) was tolled while Rhines’ state habeas corpus petition was pending, see 28 U. S. C. §2244(d)(2), he still had more than 11 months left before the expiration of the limitations period.
With the assistance of court-appointed counsel, Rhines filed an amended petition for writ of habeas corpus and statement of exhaustion on November 20, 2000, asserting 35 claims of constitutional defects in his conviction and sentence. App. 39–60. The State challenged 12 of those claims as unexhausted. Id., at 72–79. On July 3, 2002, approximately 18 months after Rhines had filed his amended federal habeas corpus petition, the District Court held that 8 of the 35 claims had not been exhausted. At this time, the AEDPA 1-year statute of limitations had run. See Duncan v. Walker, 533 U. S. 167, 181–182 (2001) (holding that the statute of limitations is not tolled during the pendancy of a federal petition). As a result, if the District Court had dismissed Rhines’ mixed petition at that point, he would have been unable to refile in federal court after exhausting the unexhausted claims. Rhines therefore moved the District Court to hold his pending habeas petition in abeyance while he presented his unexhausted claims to the South Dakota courts. On July 3, 2002, the District Court granted the motion and issued a stay “conditioned upon petitioner commencing state court exhaustion proceedings within sixty days of this order and returning to this court within sixty days of completing such exhaustion.” App. 136. In compliance with that order, Rhines filed his second state habeas corpus petition on August 22, 2003.
The State appealed the District Court’s stay of Rhines’ mixed petition to the United States Court of Appeals for the Eighth Circuit. Relying on its decision in Akins v. Kenney, 341 F. 3d 681, 686 (2003) (holding that “a district court has no authority to hold a habeas petition containing unexhausted claims in abeyance absent truly exceptional circumstances” (internal quotation marks omitted)), the Court of Appeals vacated the stay and remanded the case to the District Court to determine whether Rhines could proceed by deleting unexhausted claims from his petition. 346 F. 3d 799 (2003). We granted certiorari to resolve a split in the Circuits regarding the propriety of the District Court’s “stay-and-abeyance” procedure. 542 U. S. ___ (2004). Compare, e.g., Crews v. Horn, 360 F. 3d 146, 152 (CA3 2004); and Zarvela v. Artuz, 254 F. 3d 374, 381 (CA2 2001), with 346 F. 3d 799 (2003) (case below).
II
Fourteen years before Congress enacted AEDPA, we held in Rose v. Lundy, 455 U. S. 509 (1982), that federal district courts may not adjudicate mixed petitions for habeas corpus, that is, petitions containing both exhausted and unexhausted claims. We reasoned that the interests of comity and federalism dictate that state courts must have the first opportunity to decide a petitioner’s claims. Id., at 518–519. We noted that “[b]ecause ‘it would be unseemly in our dual system of government for a federal district court to upset a state court conviction without an opportunity to the state courts to correct a constitutional violation,’ federal courts apply the doctrine of comity.” Id., at 518 (quoting Darr v. Burford, 339 U. S. 200, 204 (1950)). That doctrine “ ‘teaches that one court should defer action on causes properly within its jurisdiction until the courts of another sovereignty with concurrent powers, and already cognizant of the litigation, have had an opportunity to pass upon the matter.’ ” 455 U. S., at 518.
Accordingly, we imposed a requirement of “total exhaustion” and directed federal courts to effectuate that requirement by dismissing mixed petitions without prejudice and allowing petitioners to return to state court to present the unexhausted claims to that court in the first instance. Id., at 522. When we decided Lundy, there was no statute of limitations on the filing of federal habeas corpus petitions. As a result, petitioners who returned to state court to exhaust their previously unexhausted claims could come back to federal court to present their perfected petitions with relative ease. See Slack v. McDaniel, 529 U. S. 473, 486 (2000) (dismissal without prejudice under Lundy “contemplated that the prisoner could return to federal court after the requisite exhaustion”).
The enactment of AEDPA in 1996 dramatically altered the landscape for federal habeas corpus petitions. AEDPA preserved Lundy ’s total exhaustion requirement, see 28 U. S. C. §2254(b)(1)(A) (“An application for a writ of habeas corpus… shall not be granted unless it appears that… the applicant has exhausted the remedies available in the courts of the State”), but it also imposed a 1-year statute of limitations on the filing of federal petitions, §2244(d). Although the limitations period is tolled during the pendency of a “properly filed application for State post-conviction or other collateral review,” §2244(d)(2), the filing of a petition for habeas corpus in federal court does not toll the statute of limitations, Duncan, supra, at 181–182.
As a result of the interplay between AEDPA’s 1-year statute of limitations and Lundy ’s dismissal requirement, petitioners who come to federal court with “mixed” petitions run the risk of forever losing their opportunity for any federal review of their unexhausted claims. If a petitioner files a timely but mixed petition in federal district court, and the district court dismisses it under Lundy after the limitations period has expired, this will likely mean the termination of any federal review. For example, if the District Court in this case had dismissed the petition because it contained unexhausted claims, AEDPA’s 1-year statute of limitations would have barred Rhines from returning to federal court after exhausting the previously unexhausted claims in state court. Similarly, if a district court dismisses a mixed petition close to the end of the 1-year period, the petitioner’s chances of exhausting his claims in state court and refiling his petition in federal court before the limitations period runs are slim. The problem is not limited to petitioners who file close to the AEDPA deadline. Even a petitioner who files early will have no way of controlling when the district court will resolve the question of exhaustion. Thus, whether a petitioner ever receives federal review of his claims may turn on which district court happens to hear his case.
We recognize the gravity of this problem and the difficulty it has posed for petitioners and federal district courts alike. In an attempt to solve the problem, some district courts have adopted a version of the “stay-and-abeyance” procedure employed by the District Court below. Under this procedure, rather than dismiss the mixed petition pursuant to Lundy, a district court might stay the petition and hold it in abeyance while the petitioner returns to state court to exhaust his previously unexhausted claims. Once the petitioner exhausts his state remedies, the district court will lift the stay and allow the petitioner to proceed in federal court.
District courts do ordinarily have authority to issue stays, see Landis v. North American Co., 299 U. S. 248, 254 (1936), where such a stay would be a proper exercise of discretion, see Clinton v. Jones, 520 U. S. 681, 706 (1997). AEDPA does not deprive district courts of that authority, cf. 28 U. S. C. §2254(b)(1)(A) (“An application for a writ of habeas corpus… shall not be granted unless it appears that… the applicant has exhausted the remedies available in the courts of the State” (emphasis added)), but it does circumscribe their discretion. Any solution to this problem must therefore be compatible with AEDPA’s purposes.
One of the statute’s purposes is to “reduce delays in the execution of state and federal criminal sentences, particularly in capital cases.” Woodford v. Garceau, 538 U. S. 202, 206 (2003). See also Duncan, 533 U. S., at 179. AEDPA’s 1-year limitations period “quite plainly serves the well-recognized interest in the finality of state court judgments.” Ibid. It “reduces the potential for delay on the road to finality by restricting the time that a prospective federal habeas petitioner has in which to seek federal habeas review.” Ibid.
Moreover, Congress enacted AEDPA against the backdrop of Lundy ’s total exhaustion requirement. The tolling provision in §2244(d)(2) “balances the interests served by the exhaustion requirement and the limitation period,” “by protecting a state prisoner’s ability later to apply for federal habeas relief while state remedies are being pursued.” Duncan, supra, at 179. AEDPA thus encourages petitioners to seek relief from state courts in the first instance by tolling the 1-year limitations period while a “properly filed application for State post-conviction or other collateral review” is pending. 28 U. S. C. §2244(d)(2). This scheme reinforces the importance of Lundy ’s “simple and clear instruction to potential litigants: before you bring any claims to federal court, be sure that you first have taken each one to state court.” 455 U. S., at 520.
Stay and abeyance, if employed too frequently, has the potential to undermine these twin purposes. Staying a federal habeas petition frustrates AEDPA’s objective of encouraging finality by allowing a petitioner to delay the resolution of the federal proceedings. It also undermines AEDPA’s goal of streamlining federal habeas proceedings by decreasing a petitioner’s incentive to exhaust all his claims in state court prior to filing his federal petition. Cf. Duncan, supra, at 180 (“[D]iminution of statutory incentives to proceed first in state court would… increase the risk of the very piecemeal litigation that the exhaustion requirement is designed to reduce”).
For these reasons, stay and abeyance should be available only in limited circumstances. Because granting a stay effectively excuses a petitioner’s failure to present his claims first to the state courts, stay and abeyance is only appropriate when the district court determines there was good cause for the petitioner’s failure to exhaust his claims first in state court. Moreover, even if a petitioner had good cause for that failure, the district court would abuse its discretion if it were to grant him a stay when his unexhausted claims are plainly meritless. Cf. 28 U. S. C. §2254(b)(2) (“An application for a writ of habeas corpus may be denied on the merits, notwithstanding the failure of the applicant to exhaust the remedies available in the courts of the State”).
Even where stay and abeyance is appropriate, the district court’s discretion in structuring the stay is limited by the timeliness concerns reflected in AEDPA. A mixed petition should not be stayed indefinitely. Though, generally, a prisoner’s “principal interest… is in obtaining speedy federal relief on his claims,” Lundy, supra, at 520 (plurality opinion), not all petitioners have an incentive to obtain federal relief as quickly as possible. In particular, capital petitioners might deliberately engage in dilatory tactics to prolong their incarceration and avoid execution of the sentence of death. Without time limits, petitioners could frustrate AEDPA’s goal of finality by dragging out indefinitely their federal habeas review. Thus, district courts should place reasonable time limits on a petitioner’s trip to state court and back. See, e.g., Zarvela, 254 F. 3d, at 381 (“[District courts] should explicitly condition the stay on the prisoner’s pursuing state court remedies within a brief interval, normally 30 days, after the stay is entered and returning to federal court within a similarly brief interval, normally 30 days after state court exhaustion is completed”). And if a petitioner engages in abusive litigation tactics or intentional delay, the district court should not grant him a stay at all. See id., at 380–381.
On the other hand, it likely would be an abuse of discretion for a district court to deny a stay and to dismiss a mixed petition if the petitioner had good cause for his failure to exhaust, his unexhausted claims are potentially meritorious, and there is no indication that the petitioner engaged in intentionally dilatory litigation tactics. In such circumstances, the district court should stay, rather than dismiss, the mixed petition. See Lundy, 455 U. S., at 522 (the total exhaustion requirement was not intended to “unreasonably impair the prisoner’s right to relief”). In such a case, the petitioner’s interest in obtaining federal review of his claims outweighs the competing interests in finality and speedy resolution of federal petitions. For the same reason, if a petitioner presents a district court with a mixed petition and the court determines that stay and abeyance is inappropriate, the court should allow the petitioner to delete the unexhausted claims and to proceed with the exhausted claims if dismissal of the entire petition would unreasonably impair the petitioner’s right to obtain federal relief. See id., at 520 (plurality opinion) (“[A petitioner] can always amend the petition to delete the unexhausted claims, rather than returning to state court to exhaust all of his claims”).
The Court of Appeals erred to the extent it concluded that stay and abeyance is always impermissible. We therefore vacate the judgment of the Court of Appeals and remand the case for that court to determine, consistent with this opinion, whether the District Court’s grant of a stay in this case constituted an abuse of discretion.
It is so ordered.
Justice O’CONNOR, with whom Justice Stevens, Justice Souter, and Justice Breyer join, dissenting.
José Ernesto MedellÃn offered proof to the Court of Appeals that reasonable jurists would find debatable or wrong the District Courtâs disposition of his claim that Texas violated his rights under the Vienna Convention on Consular Relations and that he is thereby entitled to review and reconsideration of his conviction and sentence. Three specific issues deserve further consideration: (1) whether the International Court of Justiceâs judgment in MedellÃnâs favor, Case Concerning Avena and Other Mexican Nationals (Mex. v. U. S.), 2004 I. C. J. No. 128 (Judgment of Mar. 31), is binding on American courts; (2) whether Article 36(1)(b) of the Convention creates a judicially enforceable individual right; and (3) whether Article 36(2) of the Convention sometimes requires state procedural default rules to be set aside so that the treaty can be given âfull effect.â Accordingly, I would vacate the denial of a certificate of appealability and remand for resolution of these issues.
The Court dismisses the writ (and terminates federal proceedings) on the basis of speculation: MedellÃn might obtain relief in new state court proceedingsâbecause of the Presidentâs recent memorandum about whose constitutionality the Court remains rightfully agnostic, or he might be unable to secure ultimate relief in federal courtâbecause of questions about whose resolution the Court is likewise, rightfully, undecided.
Justice O’Connor, with whom Justice Breyer joins except as to Part III, concurring in part and concurring in the judgment.
I join the Court’s opinion except for Part II–A. Although I agree with most of the Court’s reasoning, I write separately to emphasize two points. First, I think respondents’ claim implicates important associational interests, and I see no reason to minimize those interests to dispose of this case. Second, I agree with the Court that only Oklahoma’s semiclosed primary law is properly before us, that standing alone it imposes only a modest, nondiscriminatory burden on respondents’ associational rights, and that this burden is justified by the State’s legitimate regulatory interests. I note, however, that there are some grounds for concern that other state laws may unreasonably restrict voters’ ability to change party registration so as to participate in the Libertarian Party of Oklahoma’s (LPO) primary. A realistic assessment of regulatory burdens on associational rights would, in an appropriate case, require examination of the cumulative effects of the State’s overall scheme governing primary elections; and any finding of a more severe burden would trigger more probing review of the justifications offered by the State.
I
Nearly every State in the Nation now mandates that political parties select their candidates for national or statewide office by means of primary elections. See Galderisi & Ezra, Congressional Primaries in Historical and Theoretical Context, in Congressional Primaries and the Politics of Representation 17, and n. 34 (P. Galderisi, M. Ezra, & M. Lyons eds. 2001). Primaries constitute both a “ ‘crucial juncture’ ” in the electoral process, California Democratic Party v. Jones, 530 U. S. 567, 575 (2000) (quoting Tashjian v. Republican Party of Conn., 479 U. S. 208, 216 (1986)), and a vital forum for expressive association among voters and political parties, see Kusper v. Pontikes, 414 U. S. 51, 58 (1973) (“[A] basic function of a political party is to select the candidates for public office to be offered to the voters at general elections[, and a] prime objective of most voters in associating themselves with a particular party must surely be to gain a voice in that selection process”). It is here that the parties invite voters to join in selecting their standard bearers. The outcome is pivotal, of course, for it dictates the range of choices available at—and often the presumptive winner of—the general election.
“No right is more precious in a free country than that of having a voice in the election of those who make the laws under which, as good citizens, we must live,” Wesberry v. Sanders, 376 U. S. 1, 17 (1964), and “[t]he right to associate with the political party of one’s choice is an integral part of this basic constitutional freedom,” Kusper, supra, at 57. The Court has repeatedly reaffirmed that the First and Fourteenth Amendments protect the rights of voters and parties to associate through primary elections. See, e.g., California Democratic Party, supra, at 574–575; Tashjian, supra, at 214; Kusper, supra, at 56–57. Indeed, constitutional protection of associational rights is especially important in this context because the aggregation of votes is, in some sense, the essence of the electoral process. To have a meaningful voice in this process, the individual voter must join together with likeminded others at the polls. And the choice of who will participate in selecting a party’s candidate obviously plays a critical role in determining both the party’s message and its prospects of success in the electoral contest. See California Democratic Party, supra, at 575; see also Democratic Party of United States v. Wisconsin ex rel. La Follette, 450 U. S. 107, 122 (1981) (“[T]he freedom to associate for the ‘common advancement of political beliefs’ necessarily presupposes the freedom to identify the people who constitute the association” (quoting Kusper, supra, at 56)).
The majority questions whether the LPO and voters registered with another party have any constitutionally cognizable interest in associating with one another through the LPO’s primary. See ante, at 5–6. Its doubts on this point appear to stem from two implicit premises: first, that a voter forms a cognizable association with a political party only by registering with that party; and second, that a voter can only form a cognizable association with one party at a time. Neither of these premises is sound, in my view. As to the first, registration with a political party surely may signify an important personal commitment, which may be accompanied by faithful voting and even activism beyond the polls. But for many voters, registration serves principally as a mandatory (and perhaps even ministerial) prerequisite to participation in the party’s primaries. The act of casting a ballot in a given primary may, for both the voter and the party, constitute a form of association that is at least as important as the act of registering. See La Follette, supra, at 130, n. 2 (Powell, J., dissenting) (“[T]he act of voting in the Democratic primary fairly can be described as an act of affiliation with the Democratic Party”). The fact that voting is episodic does not, in my judgment, undermine its associational significance; it simply reflects the special character of the electoral process, which allows citizens to join together at regular intervals to shape government through the choice of public officials.
As to the question of dual associations, I fail to see why registration with one party should negate a voter’s First Amendment interest in associating with a second party. We surely would not say, for instance, that a registered Republican or Democrat has no protected interest in associating with the Libertarian Party by attending meetings or making political contributions. The validity of voters’ and parties’ interests in dual associations seems particularly clear where minor parties are concerned. For example, a voter may have a longstanding affiliation with a major party that she wishes to maintain, but she may nevertheless have a substantial interest in associating with a minor party during particular election cycles or in elections for particular offices. The voter’s refusal to disaffiliate from the major party may reflect her abiding commitment to that party (which is not necessarily inconsistent with her desire to associate with a second party), the objective costs of disaffiliation, see, e.g., infra, at 9–10, or both. The minor party, for its part, may have a significant interest in augmenting its voice in the political process by associating with sympathetic members of the major parties.
None of this is to suggest that the State does not have a superseding interest in restricting certain forms of association. We have never questioned, for example, the States’ authority to restrict voters’ public registration to a single party or to limit each voter to participating in a single party’s primary. But the fact that a State’s regulatory authority may ultimately trump voters’ or parties’ associational interests in a particular context is no reason to dismiss the validity of those interests. As a more general matter, I question whether judicial inquiry into the genuineness, intensity, or duration of a given voter’s association with a given party is a fruitful way to approach constitutional challenges to regulations like the one at issue here. Primary voting is an episodic and sometimes isolated act of association, but it is a vitally important one and should be entitled to some level of constitutional protection. Accordingly, where a party invites a voter to participate in its primary and the voter seeks to do so, we should begin with the premise that there are significant associational interests at stake. From this starting point, we can then ask to what extent and in what manner the State may justifiably restrict those interests.
II
As to the remainder of the constitutional analysis, I am substantially in accord with the Court’s reasoning. Our constitutional system assigns the States broad authority to regulate the electoral process, and we have recognized that, “as a practical matter, there must be substantial regulation of elections if they are to be fair and honest and if some sort of order, rather than chaos, is to accompany the democratic processes,” Storer v. Brown, 415 U. S. 724, 730 (1974). We have sought to balance the associational interests of parties and voters against the States’ regulatory interests through the flexible standard of review reaffirmed by the Court today. See ante, at 3. Under that standard, “the rigorousness of our inquiry into the propriety of a state election law depends upon the extent to which a challenged regulation burdens First and Fourteenth Amendment rights.” Burdick v. Takushi, 504 U. S. 428, 434 (1992). Regulations imposing severe burdens on associational rights must be narrowly tailored to advance a compelling government interest. Timmons v. Twin Cities Area New Party, 520 U. S. 351, 358 (1997) . Regulations imposing lesser burdens are subject to less intensive scrutiny, and reasonable, nondiscriminatory restrictions ordinarily will be sustained if they serve important regulatory interests. Ibid.
This regime reflects the limited but important role of courts in reviewing electoral regulation. Although the State has a legitimate—and indeed critical—role to play in regulating elections, it must be recognized that it is not a wholly independent or neutral arbiter. Rather, the State is itself controlled by the political party or parties in power, which presumably have an incentive to shape the rules of the electoral game to their own benefit. Recognition of that basic reality need not render suspect most electoral regulations. Where the State imposes only reasonable and genuinely neutral restrictions on associational rights, there is no threat to the integrity of the electoral process and no apparent reason for judicial intervention. As such restrictions become more severe, however, and particularly where they have discriminatory effects, there is increasing cause for concern that those in power may be using electoral rules to erect barriers to electoral competition. In such cases, applying heightened scrutiny helps to ensure that such limitations are truly justified and that the State’s asserted interests are not merely a pretext for exclusionary or anticompetitive restrictions.
Throughout the proceedings in the lower courts, respondents framed their suit as a facial challenge to Oklahoma’s semiclosed primary law. The sum of their argument was that, by requiring voters to register either as Libertarians or Independents in order to participate in the LPO’s primary, state law imposes a severe and unjustified burden on the LPO’s and Oklahoma voters’ associational rights. For the reasons explained by the Court, ante, at 14–15, that is the only claim properly before us. Assuming (as I believe we must under the circumstances) that Oklahoma provides reasonable avenues for voters to reregister as Independents or Libertarians, I agree with the Court that the semiclosed primary law imposes only a modest and politically neutral burden on associational rights. The burden is not altogether trivial: A voter with a significant commitment to a major party (for example) must forfeit registration with that party in order to participate in the LPO primary in any given election cycle, and the LPO cannot define the bounds of the association as broadly as it would like. Seepost, at 3, and n. 1 (Stevens, J., dissenting); see alsosupra, at 4 (discussing the interest in dual associations). But neither is it severe or discriminatory.
Oklahoma’s semiclosed primary law simply requires that voters wishing to participate in the LPO’s primary do what they would have to do in order to participate in any other party’s primary. By providing a reasonably fixed party-related electoral base from the close of registration until the date of the vote, this requirement facilitates campaign planning. And assuming the availability of reasonable reregistration procedures, a party’s inability to persuade a voter to disaffiliate from a rival party would suggest not the presence of anticompetitive regulatory restrictions, but rather the party’s failure to win the voter’s allegiance. The semiclosed primary law, standing alone, does not impose a significant obstacle to participation in the LPO’s primary, nor does it indicate partisan self dealing or a lockup of the political process that would warrant heightened judicial scrutiny.
For essentially the reasons explained by the Court, see ante, at 10–14, I agree that Oklahoma has a legitimate interest in requiring voters to disaffiliate from one party before participating in another party’s primary. On the record before us, I also agree that the State’s regulatory interests are adequate to justify the limited burden the semiclosed primary law imposes on respondents’ freedom of association. And finally, I agree that this case is distinguishable from Tashjian. See ante, at 8–10. I joined the dissent in that case, and I think the Court’s application of strict scrutiny there is difficult to square with the flexible standard of review articulated in our more recent cases, see supra, at 5–6. But Tashjian is entitled to respect under principles of stare decisis, and it can be fairly distinguished on the grounds that the closed primary law in that case imposed a greater burden on associational interests than does Oklahoma’s semiclosed primary law, see ante, at 9, while the State’s regulatory interests in Tashjian were weaker than they are here, compare ante, at 10–14, with Tashjian, 479 U. S., at 217–225.
III
In briefing and oral argument before this Court, respondents raise for the first time the claim that Oklahoma’s semiclosed primary law severely burdens their associational rights not through the law’s own operation, but rather because other state laws make it quite difficult for voters to reregister as Independents or Libertarians so as to participate in the LPO primary. See Brief for Respondents 12–24. Respondents characterize Oklahoma’s regulatory scheme as follows.
Partisan primaries in Oklahoma are held on the last Tuesday in July of each even-numbered year. Okla. Stat. Ann., Tit. 26, §1–102 (West Supp. 2005). To field a party candidate in an election, the LPO must obtain “recognized” party status. See ibid.; see also §§1–107, 5–104 (West 1997 and Supp. 2005). This requires it to submit, no later than May 1 of any even-numbered year ( i.e., any election year), a petition with the signatures of registered voters equal to at least five percent of the total votes cast in the most recent gubernatorial or Presidential election. §1–108 (West Supp. 2005). The State Election Board then has 30 days to determine whether the petition is sufficient. §1–108(3) . The LPO has attained recognized party status in this fashion in every Presidential election year since 1980. However, unless the party’s candidate receives at least 10 percent of the total votes cast for Governor or President in the general election (which no minor party has been able to do in any State in recent history), it loses recognized party status. §1–109. To regain party status, the group must go through the petition process again. Ibid.When a party loses its recognized status, as the LPO has after every general election in which it has participated, the affiliation of any voter registered with the party is changed to Independent. §1–110. As the District Court noted, “it is highly likely that the ranks of independents, and, indeed, of registered Republicans and Democrats, contain numerous voters who sympathize with the LPO but who simply do not wish to go through the motions of re-registering every time they are purged from the rolls.” Memorandum Opinion, Case No. CIV–00–1071–F (WD Okla., Jan. 24, 2003), App. to Pet. for Cert. A–48. And the Republican and Democratic parties in Oklahoma, as it turns out, do not permit voters registered as Independents to participate in their primaries.
Most importantly, according to respondents, the deadline for changing party affiliation makes it quite difficult for the LPO to invite voters to reregister in order to participate in its primary. Assuming the LPO submits its petition for recognized party status on the May 1 deadline, the State has until May 31 to determine whether party status will be conferred. See Okla. Stat. Ann., Tit. 26, §1–108 (West Supp. 2005). But in order to participate in the LPO primary, a voter registered with another party must change her party affiliation to Independent or Libertarian no later than June 1. See §4–119. Moreover, no candidate for office is permitted officially to declare her candidacy with the State Election Board until the period between the first Monday in June and the next succeeding Wednesday. §5–110.
If this characterization of state law is accurate, a registered Democrat or Republican sympathetic to the LPO or to an LPO candidate in a given election year would seem to face a genuine dilemma. On the one hand, she may stick with her major party registration and forfeit the opportunity to participate in the LPO primary. Alternatively, she may reregister as a Libertarian or Independent, thus forfeiting her opportunity to participate in the major party primary, though no candidate will have officially declared yet and the voter may not yet know whether the LPO will even be permitted to conduct a primary. Moreover, she must make this choice roughly eight weeks before the primaries, at a time when most voters have not yet even tuned in to the election, much less decided upon a candidate. See California Democratic Party, 530 U. S., at 586 (Kennedy, J., concurring). That might pose a special difficulty for voters attracted to minor party candidates, for whom support may not coalesce until comparatively late in the election cycle. See Anderson v. Celebrezze, 460 U. S. 780, 791–792 (1983) (discussing emergence of independent candidacies late in the election cycle).
Throughout the proceedings in the lower courts, which included a full bench trial before the District Court, respondents made no attempt to challenge these other electoral requirements or to argue that they were relevant to respondents’ challenge to the semiclosed primary law. The lower courts, accordingly, gave little or no consideration to how these various regulations interrelate or operate in practice, nor did the State seek to justify them. Given this posture, I agree with the Court that it would be neither proper nor prudent for us to rule on the reformulated claim that respondents now urge. See ante, at 14–15.
Nevertheless, respondents’ allegations are troubling, and, if they had been properly raised, the Court would want to examine the cumulative burdens imposed by the overall scheme of electoral regulations upon the rights of voters and parties to associate through primary elections. A panoply of regulations, each apparently defensible when considered alone, may nevertheless have the combined effect of severely restricting participation and competition. Even if each part of a regulatory regime might be upheld if challenged separately, one or another of these parts might have to fall if the overall scheme unreasonably curtails associational freedoms. Oklahoma’s requirement that a voter register as an Independent or a Libertarian in order to participate in the LPO’s primary is not itself unduly onerous; but that is true only to the extent that the State provides reasonable avenues through which a voter can change her registration status. The State’s regulations governing changes in party affiliation are not properly before us now. But if it were shown, in an appropriate case, that such regulations imposed a weighty or discriminatory restriction on voters’ ability to participate in the LPO’s or some other party’s primary, then more probing scrutiny of the State’s justifications would be required.
Justice O’Connor delivered the opinion of the Court.
On occasion, a would-be doctrinal rule or test finds its way into our case law through simple repetition of a phrase—however fortuitously coined. A quarter century ago, in Agins v. City of Tiburon, 447 U. S. 255 (1980), the Court declared that government regulation of private property “effects a taking if [such regulation] does not substantially advance legitimate state interests ….” Id., at 260. Through reiteration in a half dozen or so decisions since Agins, this language has been ensconced in our Fifth Amendment takings jurisprudence. See Monterey v. Del Monte Dunes at Monterey, Ltd., 526 U. S. 687, 704 (1999) (citing cases).
In the case before us, the lower courts applied Agins ’ “substantially advances” formula to strike down a Hawaii statute that limits the rent that oil companies may charge to dealers who lease service stations owned by the companies. The lower courts held that the rent cap effects an uncompensated taking of private property in violation of the Fifth and Fourteenth Amendments because it does not substantially advance Hawaii’s asserted interest in controlling retail gasoline prices. This case requires us to decide whether the “substantially advances” formula announced in Agins is an appropriate test for determining whether a regulation effects a Fifth Amendment taking. We conclude that it is not.
I
The State of Hawaii, whose territory comprises an archipelago of 132 islands clustered in the midst of the Pacific Ocean, is located over 1,600 miles from the U. S. mainland and ranks among the least populous of the 50 States. Because of Hawaii’s small size and geographic isolation, its wholesale market for oil products is highly concentrated. When this lawsuit began in 1997, only two refineries and six gasoline wholesalers were doing business in the State. As of that time, respondent Chevron U. S. A. Inc. was the largest refiner and marketer of gasoline in Hawaii: It controlled 60 percent of the market for gasoline produced or refined in-state and 30 percent of the wholesale market on the State’s most populous island, Oahu.
Gasoline is sold at retail in Hawaii from about 300 different service stations. About half of these stations are leased from oil companies by independent lessee-dealers, another 75 or so are owned and operated by “open” dealers, and the remainder are owned and operated by the oil companies. Chevron sells most of its product through 64 independent lessee-dealer stations. In a typical lessee-dealer arrangement, Chevron buys or leases land from a third party, builds a service station, and then leases the station to a dealer on a turnkey basis. Chevron charges the lessee-dealer a monthly rent, defined as a percentage of the dealer’s margin on retail sales of gasoline and other goods. In addition, Chevron requires the lessee-dealer to enter into a supply contract, under which the dealer agrees to purchase from Chevron whatever is necessary to satisfy demand at the station for Chevron’s product. Chevron unilaterally sets the wholesale price of its product.
The Hawaii Legislature enacted Act 257 in June 1997, apparently in response to concerns about the effects of market concentration on retail gasoline prices. See 1997 Haw. Sess. Laws no. 257, §1. The statute seeks to protect independent dealers by imposing certain restrictions on the ownership and leasing of service stations by oil companies. It prohibits oil companies from converting existing lessee-dealer stations to company-operated stations and from locating new company-operated stations in close proximity to existing dealer-operated stations. Haw. Rev. Stat. §§486H–10.4(a), (b) (1998 Cum. Supp.). More importantly for present purposes, Act 257 limits the amount of rent that an oil company may charge a lessee-dealer to 15 percent of the dealer’s gross profits from gasoline sales plus 15 percent of gross sales of products other than gasoline. §486H–10.4(c).
Thirty days after Act 257’s enactment, Chevron sued the Governor and Attorney General of Hawaii in their official capacities (collectively Hawaii) in the United States District Court for the District of Hawaii, raising several federal constitutional challenges to the statute. As pertinent here, Chevron claimed that the statute’s rent cap provision, on its face, effected a taking of Chevron’s property in violation of the Fifth and Fourteenth Amendments. Chevron sought a declaration to this effect as well as an injunction against the application of the rent cap to its stations. Chevron swiftly moved for summary judgment on its takings claim, arguing that the rent cap does not substantially advance any legitimate government interest. Hawaii filed a cross-motion for summary judgment on all of Chevron’s claims.
To facilitate resolution of the summary judgment motions, the parties jointly stipulated to certain relevant facts. They agreed that Act 257 reduces by about $207,000 per year the aggregate rent that Chevron would otherwise charge on 11 of its 64 lessee-dealer stations. On the other hand, the statute allows Chevron to collect more rent than it would otherwise charge at its remaining 53 lessee-dealer stations, such that Chevron could increase its overall rental income from all 64 stations by nearly $1.1 million per year. The parties further stipulated that, over the past 20 years, Chevron has not fully recovered the costs of maintaining lessee-dealer stations in any State through rent alone. Rather, the company recoups its expenses through a combination of rent and product sales. Finally, the joint stipulation states that Chevron has earned in the past, and anticipates that it will continue to earn under Act 257, a return on its investment in lessee-dealer stations in Hawaii that satisfies any constitutional standard.
The District Court granted summary judgment to Chevron, holding that “Act 257 fails to substantially advance a legitimate state interest, and as such, effects an unconstitutional taking in violation of the Fifth and Fourteenth Amendments.” Chevron U. S. A. Inc. v. Cayetano, 57 F. Supp. 2d 1003, 1014 (1998). The District Court accepted Hawaii’s argument that the rent cap was intended to prevent concentration of the retail gasoline market—and, more importantly, resultant high prices for consumers—by maintaining the viability of independent lessee-dealers. Id., at 1009–1010. The court concluded that the statute would not substantially advance this interest, however, because it would not actually reduce lessee-dealers’ costs or retail prices. It found that the rent cap would allow incumbent lessee-dealers, upon transferring occupancy rights to a new lessee, to charge the incoming lessee a premium reflecting the value of the rent reduction. Accordingly, the District Court reasoned, the incoming lessee’s overall expenses would be the same as in the absence of the rent cap, so there would be no savings to pass along to consumers. Id., at 1010–1012. Nor would incumbent lessees benefit from the rent cap, the court found, because the oil company lessors would unilaterally raise wholesale fuel prices in order to offset the reduction in their rental income. Id., at 1012–1014.
On appeal, a divided panel of the Court of Appeals for the Ninth Circuit held that the District Court had applied the correct legal standard to Chevron’s takings claim. Chevron U. S. A. Inc. v. Cayetano, 224 F. 3d 1030, 1033–1037 (2000). The Court of Appeals vacated the grant of summary judgment, however, on the ground that a genuine issue of material fact remained as to whether the Act would benefit consumers. Id., at 1037–1042. Judge William Fletcher concurred in the judgment, maintaining that the “reasonableness” standard applicable to “ordinary rent and price control laws” should instead govern Chevron’s claim. Id., at 1048.
On remand, the District Court entered judgment for Chevron after a 1-day bench trial in which Chevron and Hawaii called competing expert witnesses (both economists) to testify. 198 F. Supp. 2d 1182 (2002). Finding Chevron’s expert witness to be “more persuasive” than the State’s expert, the District Court once again concluded that oil companies would raise wholesale gasoline prices to offset any rent reduction required by Act 257, and that the result would be an increase in retail gasoline prices. Id., at 1187–1189. Even if the rent cap did reduce lessee-dealers’ costs, the court found, they would not pass on any savings to consumers. Id., at 1189. The court went on to reiterate its determination that Act 257 would enable incumbent lessee-dealers to sell their leaseholds at a premium, such that incoming lessees would not obtain any of the benefits of the rent cap. Id., at 1189–1190. And while it acknowledged that the rent cap could preclude oil companies from constructively evicting dealers through excessive rents, the court found no evidence that Chevron or any other oil company would attempt to charge such rents in the absence of the cap. Id., at 1191. Finally, the court concluded that Act 257 would in fact decrease the number of lessee-dealer stations because the rent cap would discourage oil companies from building such stations. Id., at 1191–1192. Based on these findings, the District Court held that “Act 257 effect[ed] an unconstitutional regulatory taking given its failure to substantially advance any legitimate state interest.” Id., at 1193.
The Ninth Circuit affirmed, holding that its decision in the prior appeal barred Hawaii from challenging the application of the “substantially advances” test to Chevron’s takings claim or from arguing for a more deferential standard of review. 363 F. 3d 846, 849–855 (2004). The panel majority went on to reject Hawaii’s challenge to the application of the standard to the facts of the case. Id., at 855–858. Judge Fletcher dissented, renewing his contention that Act 257 should not be reviewed under the “substantially advances” standard. Id., at 859–861. We granted certiorari, 543 U. S. ___ (2004), and now reverse.
II
A
The Takings Clause of the Fifth Amendment, made applicable to the States through the Fourteenth, see Chicago, B. & Q. R. Co. v. Chicago, 166 U. S. 226 (1897), provides that private property shall not “be taken for public use, without just compensation.” As its text makes plain, the Takings Clause “does not prohibit the taking of private property, but instead places a condition on the exercise of that power.” First English Evangelical Lutheran Church of Glendale v. County of Los Angeles, 482 U. S. 304, 314 (1987). In other words, it “is designed not to limit the governmental interference with property rights per se, but rather to secure compensation in the event of otherwise proper interference amounting to a taking.” Id., at 315 (emphasis in original). While scholars have offered various justifications for this regime, we have emphasized its role in “bar[ring] Government from forcing some people alone to bear public burdens which, in all fairness and justice, should be borne by the public as a whole.” Armstrong v. United States, 364 U. S. 40, 49 (1960); see also Monongahela Nav. Co. v. United States, 148 U. S. 312, 325 (1893).
The paradigmatic taking requiring just compensation is a direct government appropriation or physical invasion of private property. See, e.g., United States v. Pewee Coal Co., 341 U. S. 114 (1951) (Government’s seizure and operation of a coal mine to prevent a national strike of coal miners effected a taking); United States v. General Motors Corp., 323 U. S. 373 (1945) (Government’s occupation of private warehouse effected a taking). Indeed, until the Court’s watershed decision in Pennsylvania Coal Co. v. Mahon, 260 U. S. 393 (1922), “it was generally thought that the Takings Clause reached only a ‘direct appropriation’ of property, or the functional equivalent of a ‘practical ouster of [the owner’s] possession.’ ” Lucas v. South Carolina Coastal Council, 505 U. S. 1003, 1014 (1992) (citations omitted and emphasis added; brackets in original); see also id., at 1028, n. 15 (“[E]arly constitutional theorists did not believe the Takings Clause embraced regulations of property at all”).
Beginning with Mahon, however, the Court recognized that government regulation of private property may, in some instances, be so onerous that its effect is tantamount to a direct appropriation or ouster—and that such “regulatory takings” may be compensable under the Fifth Amendment. In Justice Holmes’ storied but cryptic formulation, “while property may be regulated to a certain extent, if regulation goes too far it will be recognized as a taking.” 260 U. S., at 415. The rub, of course, has been—and remains—how to discern how far is “too far.” In answering that question, we must remain cognizant that “government regulation—by definition—involves the adjustment of rights for the public good,” Andrus v. Allard, 444 U. S. 51, 65 (1979), and that “Government hardly could go on if to some extent values incident to property could not be diminished without paying for every such change in the general law,” Mahon, supra, at 413.
Our precedents stake out two categories of regulatory action that generally will be deemed per se takings for Fifth Amendment purposes. First, where government requires an owner to suffer a permanent physical invasion of her property—however minor—it must provide just compensation. See Loretto v. Teleprompter Manhattan CATV Corp., 458 U. S. 419 (1982) (state law requiring landlords to permit cable companies to install cable facilities in apartment buildings effected a taking). A second categorical rule applies to regulations that completely deprive an owner of “ all economically beneficial us[e]” of her property. Lucas, 505 U. S., at 1019 (emphasis in original). We held in Lucas that the government must pay just compensation for such “total regulatory takings,” except to the extent that “background principles of nuisance and property law” independently restrict the owner’s intended use of the property. Id., at 1026–1032.
Outside these two relatively narrow categories (and the special context of land-use exactions discussed below, see infra, at 16–18), regulatory takings challenges are governed by the standards set forth in Penn Central Transp. Co. v. New York City, 438 U. S. 104 (1978). The Court in Penn Central acknowledged that it had hitherto been “unable to develop any ‘set formula’ ” for evaluating regulatory takings claims, but identified “several factors that have particular significance.” Id., at 124. Primary among those factors are “[t]he economic impact of the regulation on the claimant and, particularly, the extent to which the regulation has interfered with distinct investment-backed expectations.” Ibid. In addition, the “character of the governmental action”—for instance whether it amounts to a physical invasion or instead merely affects property interests through “some public program adjusting the benefits and burdens of economic life to promote the common good”—may be relevant in discerning whether a taking has occurred. Ibid. The Penn Central factors—though each has given rise to vexing subsidiary questions—have served as the principal guidelines for resolving regulatory takings claims that do not fall within the physical takings or Lucas rules. See, e.g., Palazzolo v. Rhode Island, 533 U. S. 606, 617–618 (2001); id., at 632–634 (O’Connor, J., concurring).
Although our regulatory takings jurisprudence cannot be characterized as unified, these three inquiries (reflected in Loretto, Lucas, and Penn Central ) share a common touchstone. Each aims to identify regulatory actions that are functionally equivalent to the classic taking in which government directly appropriates private property or ousts the owner from his domain. Accordingly, each of these tests focuses directly upon the severity of the burden that government imposes upon private property rights. The Court has held that physical takings require compensation because of the unique burden they impose: A permanent physical invasion, however minimal the economic cost it entails, eviscerates the owner’s right to exclude others from entering and using her property—perhaps the most fundamental of all property interests. See Dolan v. City of Tigard, 512 U. S. 374, 384 (1994); Nollan v. California Coastal Comm’n, 483 U. S. 825, 831–832 (1987); Loretto, supra, at 433; Kaiser Aetna v. United States, 444 U. S. 164, 176 (1979). In the Lucas context, of course, the complete elimination of a property’s value is the determinative factor. See Lucas, supra, at 1017 (positing that “total deprivation of beneficial use is, from the landowner’s point of view, the equivalent of a physical appropriation”). And the Penn Central inquiry turns in large part, albeit not exclusively, upon the magnitude of a regulation’s economic impact and the degree to which it interferes with legitimate property interests.
B
In Agins v. City of Tiburon, a case involving a facial takings challenge to certain municipal zoning ordinances, the Court declared that “[t]he application of a general zoning law to particular property effects a taking if the ordinance does not substantially advance legitimate state interests, see Nectow v. Cambridge, 277 U. S. 183, 188 (1928), or denies an owner economically viable use of his land, see Penn Central Transp. Co. v. New York City, 438 U. S. 104, 138, n. 36 (1978).” 447 U. S., at 260. Because this statement is phrased in the disjunctive, Agins ’ “substantially advances” language has been read to announce a stand-alone regulatory takings test that is wholly independent of Penn Central or any other test. Indeed, the lower courts in this case struck down Hawaii’s rent control statute as an “unconstitutional regulatory taking,” 198 F. Supp. 2d, at 1193, based solely upon a finding that it does not substantially advance the State’s asserted interest in controlling retail gasoline prices. See supra, at 6–7. Although a number of our takings precedents have recited the “substantially advances” formula minted in Agins, this is our first opportunity to consider its validity as a freestanding takings test. We conclude that this formula prescribes an inquiry in the nature of a due process, not a takings, test, and that it has no proper place in our takings jurisprudence.
There is no question that the “substantially advances” formula was derived from due process, not takings, precedents. In support of this new language, Agins cited Nectow v. Cambridge, 277 U. S. 183, a 1928 case in which the plaintiff claimed that a city zoning ordinance “deprived him of his property without due process of law in contravention of the Fourteenth Amendment,” id., at 185. Agins then went on to discuss Village of Euclid v. Ambler Realty Co., 272 U. S. 365 (1926), a historic decision holding that a municipal zoning ordinance would survive a substantive due process challenge so long as it was not “clearly arbitrary and unreasonable, having no substantial relation to the public health, safety, morals, or general welfare.” Id., at 395 (emphasis added); see also Nectow, supra, at 188 (quoting the same “substantial relation” language from Euclid ).
When viewed in historical context, the Court’s reliance on Nectow and Euclid is understandable. Agins was the Court’s first case involving a challenge to zoning regulations in many decades, so it was natural to turn to these seminal zoning precedents for guidance. See Brief for United States as Amicus Curiae in Agins v. City of Tiburon, O. T. 1979, No. 602, pp. 12–13 (arguing that Euclid “set out the principles applicable to a determination of the facial validity of a zoning ordinance attacked as a violation of the Takings Clause of the Fifth Amendment”). Moreover, Agins ’ apparent commingling of due process and takings inquiries had some precedent in the Court’s then-recent decision in Penn Central. See 438 U. S., at 127 (stating in dicta that “[i]t is … implicit in Goldblatt [v. Hempstead, 369 U. S. 590 (1962),] that a use restriction on real property may constitute a ‘taking’ if not reasonably necessary to the effectuation of a substantial public purpose, see Nectow v. Cambridge, supra ”). But see Goldblatt, supra, at 594–595 (quoting “ ‘reasonably necessary’ ” language from Lawton v. Steele, 152 U. S. 133, 137 (1894), a due process case, and applying a deferential “ ‘reasonableness’ ” standard to determine whether a challenged regulation was a “valid exercise of the … police power” under the Due Process Clause). Finally, when Agins was decided, there had been some history of referring to deprivations of property without due process of law as “takings,” see, e.g., Rowan v. Post Office Dept., 397 U. S. 728, 740 (1970), and the Court had yet to clarify whether “regulatory takings” claims were properly cognizable under the Takings Clause or the Due Process Clause, see Williamson County Regional Planning Comm’n v. Hamilton Bank of Jefferson City, 473 U. S. 172, 197–199 (1985).
Although Agins ’ reliance on due process precedents is understandable, the language the Court selected was regrettably imprecise. The “substantially advances” formula suggests a means-ends test: It asks, in essence, whether a regulation of private property is effective in achieving some legitimate public purpose. An inquiry of this nature has some logic in the context of a due process challenge, for a regulation that fails to serve any legitimate governmental objective may be so arbitrary or irrational that it runs afoul of the Due Process Clause. See, e.g., County of Sacramento v. Lewis, 523 U. S. 833, 846 (1998) (stating that the Due Process Clause is intended, in part, to protect the individual against “the exercise of power without any reasonable justification in the service of a legitimate governmental objective”). But such a test is not a valid method of discerning whether private property has been “taken” for purposes of the Fifth Amendment.
In stark contrast to the three regulatory takings tests discussed above, the “substantially advances” inquiry reveals nothing about the magnitude or character of the burden a particular regulation imposes upon private property rights. Nor does it provide any information about how any regulatory burden is distributed among property owners. In consequence, this test does not help to identify those regulations whose effects are functionally comparable to government appropriation or invasion of private property; it is tethered neither to the text of the Takings Clause nor to the basic justification for allowing regulatory actions to be challenged under the Clause.
Chevron appeals to the general principle that the Takings Clause is meant “ ‘to bar Government from forcing some people alone to bear public burdens which, in all fairness and justice, should be borne by the public as a whole.’ ” Brief for Respondent 17–21 (quoting Armstrong, 364 U. S., at 49). But that appeal is clearly misplaced, for the reasons just indicated. A test that tells us nothing about the actual burden imposed on property rights, or how that burden is allocated cannot tell us when justice might require that the burden be spread among taxpayers through the payment of compensation. The owner of a property subject to a regulation that effectively serves a legitimate state interest may be just as singled out and just as burdened as the owner of a property subject to an ineffective regulation. It would make little sense to say that the second owner has suffered a taking while the first has not. Likewise, an ineffective regulation may not significantly burden property rights at all, and it may distribute any burden broadly and evenly among property owners. The notion that such a regulation nevertheless “takes” private property for public use merely by virtue of its ineffectiveness or foolishness is untenable.
Instead of addressing a challenged regulation’s effect on private property, the “substantially advances” inquiry probes the regulation’s underlying validity. But such an inquiry is logically prior to and distinct from the question whether a regulation effects a taking, for the Takings Clause presupposes that the government has acted in pursuit of a valid public purpose. The Clause expressly requires compensation where government takes private property “ for public use.” It does not bar government from interfering with property rights, but rather requires compensation “in the event of otherwise proper interference amounting to a taking.” First English Evangelical Lutheran Church, 482 U. S., at 315 (emphasis added). Conversely, if a government action is found to be impermissible—for instance because it fails to meet the “public use” requirement or is so arbitrary as to violate due process—that is the end of the inquiry. No amount of compensation can authorize such action.
Chevron’s challenge to the Hawaii statute in this case illustrates the flaws in the “substantially advances” theory. To begin with, it is unclear how significantly Hawaii’s rent cap actually burdens Chevron’s property rights. The parties stipulated below that the cap would reduce Chevron’s aggregate rental income on 11 of its 64 lessee-dealer stations by about $207,000 per year, but that Chevron nevertheless expects to receive a return on its investment in these stations that satisfies any constitutional standard. See supra, at 4. Moreover, Chevron asserted below, and the District Court found, that Chevron would recoup any reductions in its rental income by raising wholesale gasoline prices. See supra, at 5. In short, Chevron has not clearly argued—let alone established—that it has been singled out to bear any particularly severe regulatory burden. Rather, the gravamen of Chevron’s claim is simply that Hawaii’s rent cap will not actually serve the State’s legitimate interest in protecting consumers against high gasoline prices. Whatever the merits of that claim, it does not sound under the Takings Clause. Chevron plainly does not seek compensation for a taking of its property for a legitimate public use, but rather an injunction against the enforcement of a regulation that it alleges to be fundamentally arbitrary and irrational.
Finally, the “substantially advances” formula is not only doctrinally untenable as a takings test—its application as such would also present serious practical difficulties. The Agins formula can be read to demand heightened means-ends review of virtually any regulation of private property. If so interpreted, it would require courts to scrutinize the efficacy of a vast array of state and federal regulations—a task for which courts are not well suited. Moreover, it would empower—and might often require—courts to substitute their predictive judgments for those of elected legislatures and expert agencies.
Although the instant case is only the tip of the proverbial iceberg, it foreshadows the hazards of placing courts in this role. To resolve Chevron’s takings claim, the District Court was required to choose between the views of two opposing economists as to whether Hawaii’s rent control statute would help to prevent concentration and supracompetitive prices in the State’s retail gasoline market. Finding one expert to be “more persuasive” than the other, the court concluded that the Hawaii Legislature’s chosen regulatory strategy would not actually achieve its objectives. See 198 F. Supp. 2d, at 1187–1193. Along the way, the court determined that the State was not entitled to enact a prophylactic rent cap without actual evidence that oil companies had charged, or would charge, excessive rents. See id., at 1191 . Based on these findings, the District Court enjoined further enforcement of Act 257’s rent cap provision against Chevron. We find the proceedings below remarkable, to say the least, given that we have long eschewed such heightened scrutiny when addressing substantive due process challenges to government regulation. See, e.g., Exxon Corp. v. Governor of Maryland, 437 U. S. 117, 124–125 (1978); Ferguson v. Skrupa, 372 U. S. 726, 730–732 (1963). The reasons for deference to legislative judgments about the need for, and likely effectiveness of, regulatory actions are by now well established, and we think they are no less applicable here.
For the foregoing reasons, we conclude that the “substantially advances” formula announced in Agins is not a valid method of identifying regulatory takings for which the Fifth Amendment requires just compensation. Since Chevron argued only a “substantially advances” theory in support of its takings claim, it was not entitled to summary judgment on that claim.
III
We emphasize that our holding today—that the “substantially advances” formula is not a valid takings test—does not require us to disturb any of our prior holdings. To be sure, we applied a “substantially advances” inquiry in Agins itself, see 447 U. S., at 261–262 (finding that the challenged zoning ordinances “substantially advance[d] legitimate governmental goals”), and arguably also in Keystone Bituminous Coal Assn. v. DeBenedictis, 480 U. S. 470, 485–492 (1987) (quoting “ ‘substantially advance[s]’ ” language and then finding that the challenged statute was intended to further a substantial public interest). But in no case have we found a compensable taking based on such an inquiry. Indeed, in most of the cases reciting the “substantially advances” formula, the Court has merely assumed its validity when referring to it in dicta. See Tahoe-Sierra Preservation Council, Inc. v. Tahoe Regional Planning Agency, 535 U. S. 302, 334 (2002); Del Monte Dunes, 526 U. S., at 704; Lucas, 505 U. S., at 1016; Yee v. Escondido, 503 U. S. 519, 534 (1992); United States v. Riverside Bayview Homes, Inc., 474 U. S. 121, 126 (1985).
It might be argued that this formula played a role in our decisions in Nollan v. California Coastal Comm’n, 483 U. S. 825 (1987), and Dolan v. City of Tigard, 512 U. S. 374 (1994). See Brief for Respondent 21–23. But while the Court drew upon the language of Agins in these cases, it did not apply the “substantially advances” test that is the subject of today’s decision. Both Nollan and Dolan involved Fifth Amendment takings challenges to adjudicative land-use exactions—specifically, government demands that a landowner dedicate an easement allowing public access to her property as a condition of obtaining a development permit. See Dolan, supra, at 379–380 (permit to expand a store and parking lot conditioned on the dedication of a portion of the relevant property for a “greenway,” including a bike/pedestrian path); Nollan, supra, at 828 (permit to build a larger residence on beachfront property conditioned on dedication of an easement allowing the public to traverse a strip of the property between the owner’s seawall and the mean high-tide line).
In each case, the Court began with the premise that, had the government simply appropriated the easement in question, this would have been a per se physical taking. Dolan, supra, at 384; Nollan, supra, at 831–832. The question was whether the government could, without paying the compensation that would otherwise be required upon effecting such a taking, demand the easement as a condition for granting a development permit the government was entitled to deny. The Court in Nolan answered in the affirmative, provided that the exaction would substantially advance the same government interest that would furnish a valid ground for denial of the permit. 483 U. S., at 834–837. The Court further refined this requirement in Dolan, holding that an adjudicative exaction requiring dedication of private property must also be “ ‘rough[ly] proportiona[l]’ … both in nature and extent to the impact of the proposed development.” 512 U. S., at 391; see also Del Monte Dunes, supra, at 702 (emphasizing that we have not extended this standard “beyond the special context of [such] exactions”).
Although Nollan and Dolan quoted Agins ’ language, see Dolan, supra, at 385; Nollan, supra, at 834, the rule those decisions established is entirely distinct from the “substantially advances” test we address today. Whereas the “substantially advances” inquiry before us now is unconcerned with the degree or type of burden a regulation places upon property, Nollan and Dolan both involved dedications of property so onerous that, outside the exactions context, they would be deemed per se physical takings. In neither case did the Court question whether the exaction would substantially advance some legitimate state interest. See Dolan, supra, at 387–388; Nollan, supra, at 841. Rather, the issue was whether the exactions substantially advanced the same interests that land-use authorities asserted would allow them to deny the permit altogether. As the Court explained in Dolan, these cases involve a special application of the “doctrine of ‘unconstitutional conditions,’ ” which provides that “the government may not require a person to give up a constitutional right—here the right to receive just compensation when property is taken for a public use—in exchange for a discretionary benefit conferred by the government where the benefit has little or no relationship to the property.” 512 U. S., at 385. That is worlds apart from a rule that says a regulation affecting property constitutes a taking on its face solely because it does not substantially advance a legitimate government interest. In short, Nollan and Dolan cannot be characterized as applying the “substantially advances” test we address today, and our decision should not be read to disturb these precedents.
***
Twenty-five years ago, the Court posited that a regulation of private property “effects a taking if [it] does not substantially advance [a] legitimate state interes[t].” Agins, supra, at 260. The lower courts in this case took that statement to its logical conclusion, and in so doing, revealed its imprecision. Today we correct course. We hold that the “substantially advances” formula is not a valid takings test, and indeed conclude that it has no proper place in our takings jurisprudence. In so doing, we reaffirm that a plaintiff seeking to challenge a government regulation as an uncompensated taking of private property may proceed under one of the other theories discussed above—by alleging a “physical” taking, a Lucas -type “total regulatory taking,” a Penn Central taking, or a land-use exaction violating the standards set forth in Nollan and Dolan. Because Chevron argued only a “substantially advances” theory in support of its takings claim, it was not entitled to summary judgment on that claim. Accordingly, we reverse the judgment of the Ninth Circuit and remand the case for further proceedings consistent with this opinion.
It is so ordered.
Justice O’Connor, with whom The Chief Justice and Justice Thomas join as to all but Part III, dissenting.
We enforce the “outer limits” of Congress’ Commerce Clause authority not for their own sake, but to protect historic spheres of state sovereignty from excessive federal encroachment and thereby to maintain the distribution of power fundamental to our federalist system of government. United States v. Lopez, 514 U. S. 549, 557 (1995); NLRB v. Jones & Laughlin Steel Corp., 301 U. S. 1, 37 (1937). One of federalism’s chief virtues, of course, is that it promotes innovation by allowing for the possibility that “a single courageous State may, if its citizens choose, serve as a laboratory; and try novel social and economic experiments without risk to the rest of the country.” New State Ice Co. v. Liebmann, 285 U. S. 262, 311 (1932) (Brandeis, J., dissenting).
This case exemplifies the role of States as laboratories. The States’ core police powers have always included authority to define criminal law and to protect the health, safety, and welfare of their citizens. Brecht v. Abrahamson, 507 U. S. 619, 635 (1993); Whalen v. Roe, 429 U. S. 589, 603, n. 30 (1977). Exercising those powers, California (by ballot initiative and then by legislative codification) has come to its own conclusion about the difficult and sensitive question of whether marijuana should be available to relieve severe pain and suffering. Today the Court sanctions an application of the federal Controlled Substances Act that extinguishes that experiment, without any proof that the personal cultivation, possession, and use of marijuana for medicinal purposes, if economic activity in the first place, has a substantial effect on interstate commerce and is therefore an appropriate subject of federal regulation. In so doing, the Court announces a rule that gives Congress a perverse incentive to legislate broadly pursuant to the Commerce Clause—nestling questionable assertions of its authority into comprehensive regulatory schemes—rather than with precision. That rule and the result it produces in this case are irreconcilable with our decisions in Lopez, supra, and United States v. Morrison, 529 U. S. 598 (2000). Accordingly I dissent.
I
In Lopez, we considered the constitutionality of the Gun-Free School Zones Act of 1990, which made it a federal offense “for any individual knowingly to possess a firearm … at a place the individual knows, or has reasonable cause to believe, is a school zone,” 18 U. S. C. §922(q)(2)(A). We explained that “Congress’ commerce authority includes the power to regulate those activities having a substantial relation to interstate commerce …, i.e., those activities that substantially affect interstate commerce.” 514 U. S., at 558–559 (citation omitted). This power derives from the conjunction of the Commerce Clause and the Necessary and Proper Clause. Garcia v. San Antonio Metropolitan Transit Authority, 469 U. S. 528, 585–586 (1985) (O’Connor, J., dissenting) (explaining that United States v. Darby, 312 U. S. 100 (1941), United States v. Wrightwood Dairy Co., 315 U. S. 110 (1942), and Wickard v. Filburn, 317 U. S. 111 (1942), based their expansion of the commerce power on the Necessary and Proper Clause, and that “the reasoning of these cases underlies every recent decision concerning the reach of Congress to activities affecting interstate commerce”); ante, at 2 (Scalia, J., concurring in judgment). We held in Lopez that the Gun-Free School Zones Act could not be sustained as an exercise of that power.
Our decision about whether gun possession in school zones substantially affected interstate commerce turned on four considerations. Lopez, supra, at 559–567; see also Morrison, supra, at 609–613. First, we observed that our “substantial effects” cases generally have upheld federal regulation of economic activity that affected interstate commerce, but that §922(q) was a criminal statute having “nothing to do with ‘commerce’ or any sort of economic enterprise.” Lopez, 514 U. S., at 561. In this regard, we also noted that “[s]ection 922(q) is not an essential part of a larger regulation of economic activity, in which the regulatory scheme could be undercut unless the intrastate activity were regulated. It cannot, therefore, be sustained under our cases upholding regulations of activities that arise out of or are connected with a commercial transaction, which viewed in the aggregate, substantially affects interstate commerce.” Ibid. Second, we noted that the statute contained no express jurisdictional requirement establishing its connection to interstate commerce. Ibid.
Third, we found telling the absence of legislative findings about the regulated conduct’s impact on interstate commerce. We explained that while express legislative findings are neither required nor, when provided, dispositive, findings “enable us to evaluate the legislative judgment that the activity in question substantially affect[s] interstate commerce, even though no such substantial effect [is] visible to the naked eye.” Id., at 563. Finally, we rejected as too attenuated the Government’s argument that firearm possession in school zones could result in violent crime which in turn could adversely affect the national economy. Id., at 563–567. The Constitution, we said, does not tolerate reasoning that would “convert congressional authority under the Commerce Clause to a general police power of the sort retained by the States.” Id., at 567. Later in Morrison, supra, we relied on the same four considerations to hold that §40302 of the Violence Against Women Act of 1994, 42 U. S. C. §13981, exceeded Congress’ authority under the Commerce Clause.
In my view, the case before us is materially indistinguishable from Lopez and Morrison when the same considerations are taken into account.
II
A
What is the relevant conduct subject to Commerce Clause analysis in this case? The Court takes its cues from Congress, applying the above considerations to the activity regulated by the Controlled Substances Act (CSA) in general. The Court’s decision rests on two facts about the CSA: (1) Congress chose to enact a single statute providing a comprehensive prohibition on the production, distribution, and possession of all controlled substances, and (2) Congress did not distinguish between various forms of intrastate noncommercial cultivation, possession, and use of marijuana. See 21 U. S. C. §§841(a)(1), 844(a). Today’s decision suggests that the federal regulation of local activity is immune to Commerce Clause challenge because Congress chose to act with an ambitious, all-encompassing statute, rather than piecemeal. In my view, allowing Congress to set the terms of the constitutional debate in this way, i.e., by packaging regulation of local activity in broader schemes, is tantamount to removing meaningful limits on the Commerce Clause.
The Court’s principal means of distinguishing Lopez from this case is to observe that the Gun-Free School Zones Act of 1990 was a “brief, single-subject statute,” ante, at 20, see also ante, at 19, whereas the CSA is “a lengthy and detailed statute creating a comprehensive framework for regulating the production, distribution, and possession of five classes of ‘controlled substances,’ ” ibid. Thus, according to the Court, it was possible in Lopez to evaluate in isolation the constitutionality of criminalizing local activity (there gun possession in school zones), whereas the local activity that the CSA targets (in this case cultivation and possession of marijuana for personal medicinal use) cannot be separated from the general drug control scheme of which it is a part.
Today’s decision allows Congress to regulate intrastate activity without check, so long as there is some implication by legislative design that regulating intrastate activity is essential (and the Court appears to equate “essential” with “necessary”) to the interstate regulatory scheme. Seizing upon our language in Lopez that the statute prohibiting gun possession in school zones was “not an essential part of a larger regulation of economic activity, in which the regulatory scheme could be undercut unless the intrastate activity were regulated,” 514 U. S., at 561, the Court appears to reason that the placement of local activity in a comprehensive scheme confirms that it is essential to that scheme. Ante, at 21–22. If the Court is right, then Lopez stands for nothing more than a drafting guide: Congress should have described the relevant crime as “transfer or possession of a firearm anywhere in the nation”—thus including commercial and noncommercial activity, and clearly encompassing some activity with assuredly substantial effect on interstate commerce. Had it done so, the majority hints, we would have sustained its authority to regulate possession of firearms in school zones. Furthermore, today’s decision suggests we would readily sustain a congressional decision to attach the regulation of intrastate activity to a pre-existing comprehensive (or even not-so-comprehensive) scheme. If so, the Court invites increased federal regulation of local activity even if, as it suggests, Congress would not enact a new interstate scheme exclusively for the sake of reaching intrastate activity, see ante, at 22, n. 33; ante, at 6 (Scalia, J., concurring in judgment).
I cannot agree that our decision in Lopez contemplated such evasive or overbroad legislative strategies with approval. Until today, such arguments have been made only in dissent. See Morrison, 529 U. S., at 657 (Breyer, J., dissenting) (given that Congress can regulate “ ‘an essential part of a larger regulation of economic activity,’ ” “can Congress save the present law by including it, or much of it, in a broader ‘Safe Transport’ or ‘Worker Safety’ act?”). Lopez and Morrison did not indicate that the constitutionality of federal regulation depends on superficial and formalistic distinctions. Likewise I did not understand our discussion of the role of courts in enforcing outer limits of the Commerce Clause for the sake of maintaining the federalist balance our Constitution requires, see Lopez, 514 U. S., at 557; id., at 578 (Kennedy, J., concurring), as a signal to Congress to enact legislation that is more extensive and more intrusive into the domain of state power. If the Court always defers to Congress as it does today, little may be left to the notion of enumerated powers.
The hard work for courts, then, is to identify objective markers for confining the analysis in Commerce Clause cases. Here, respondents challenge the constitutionality of the CSA as applied to them and those similarly situated. I agree with the Court that we must look beyond respondents’ own activities. Otherwise, individual litigants could always exempt themselves from Commerce Clause regulation merely by pointing to the obvious—that their personal activities do not have a substantial effect on interstate commerce. See Maryland v. Wirtz, 392 U. S. 183, 193 (1968); Wickard, 317 U. S., at 127–128. The task is to identify a mode of analysis that allows Congress to regulate more than nothing (by declining to reduce each case to its litigants) and less than everything (by declining to let Congress set the terms of analysis). The analysis may not be the same in every case, for it depends on the regulatory scheme at issue and the federalism concerns implicated. See generally Lopez, 514 U. S., at 567; id., at 579 (Kennedy, J., concurring).
A number of objective markers are available to confine the scope of constitutional review here. Both federal and state legislation—including the CSA itself, the California Compassionate Use Act, and other state medical marijuana legislation—recognize that medical and nonmedical ( i.e., recreational) uses of drugs are realistically distinct and can be segregated, and regulate them differently. See 21 U. S. C. §812; Cal. Health & Safety Code Ann. §11362.5 (West Supp. 2005); ante, at 1 (opinion of the Court). Respondents challenge only the application of the CSA to medicinal use of marijuana. Cf. United States v. Raines, 362 U. S. 17, 20–22 (1960) (describing our preference for as-applied rather than facial challenges). Moreover, because fundamental structural concerns about dual sovereignty animate our Commerce Clause cases, it is relevant that this case involves the interplay of federal and state regulation in areas of criminal law and social policy, where “States lay claim by right of history and expertise.” Lopez, supra, at 583 (Kennedy, J., concurring); see also Morrison, supra, at 617–619; Lopez, supra, at 580 (Kennedy, J., concurring) (“The statute before us upsets the federal balance to a degree that renders it an unconstitutional assertion of the commerce power, and our intervention is required”); cf. Garcia, 469 U. S., at 586 (O’Connor, J., dissenting) (“[S]tate autonomy is a relevant factor in assessing the means by which Congress exercises its powers” under the Commerce Clause). California, like other States, has drawn on its reserved powers to distinguish the regulation of medicinal marijuana. To ascertain whether Congress’ encroachment is constitutionally justified in this case, then, I would focus here on the personal cultivation, possession, and use of marijuana for medicinal purposes.
B
Having thus defined the relevant conduct, we must determine whether, under our precedents, the conduct is economic and, in the aggregate, substantially affects interstate commerce. Even if intrastate cultivation and possession of marijuana for one’s own medicinal use can properly be characterized as economic, and I question whether it can, it has not been shown that such activity substantially affects interstate commerce. Similarly, it is neither self-evident nor demonstrated that regulating such activity is necessary to the interstate drug control scheme.
The Court’s definition of economic activity is breathtaking. It defines as economic any activity involving the production, distribution, and consumption of commodities. And it appears to reason that when an interstate market for a commodity exists, regulating the intrastate manufacture or possession of that commodity is constitutional either because that intrastate activity is itself economic, or because regulating it is a rational part of regulating its market. Putting to one side the problem endemic to the Court’s opinion—the shift in focus from the activity at issue in this case to the entirety of what the CSA regulates, see Lopez, supra, at 565 (“depending on the level of generality, any activity can be looked upon as commercial”)—the Court’s definition of economic activity for purposes of Commerce Clause jurisprudence threatens to sweep all of productive human activity into federal regulatory reach.
The Court uses a dictionary definition of economics to skirt the real problem of drawing a meaningful line between “what is national and what is local,” Jones & Laughlin Steel, 301 U. S., at 37. It will not do to say that Congress may regulate noncommercial activity simply because it may have an effect on the demand for commercial goods, or because the noncommercial endeavor can, in some sense, substitute for commercial activity. Most commercial goods or services have some sort of privately producible analogue. Home care substitutes for daycare. Charades games substitute for movie tickets. Backyard or windowsill gardening substitutes for going to the supermarket. To draw the line wherever private activity affects the demand for market goods is to draw no line at all, and to declare everything economic. We have already rejected the result that would follow—a federal police power. Lopez, supra, at 564.
In Lopez and Morrison, we suggested that economic activity usually relates directly to commercial activity. See Morrison, 529 U. S., at 611, n. 4 (intrastate activities that have been within Congress’ power to regulate have been “of an apparent commercial character”); Lopez, 514 U. S., at 561 (distinguishing the Gun-Free School Zones Act of 1990 from “activities that arise out of or are connected with a commercial transaction”). The homegrown cultivation and personal possession and use of marijuana for medicinal purposes has no apparent commercial character. Everyone agrees that the marijuana at issue in this case was never in the stream of commerce, and neither were the supplies for growing it. (Marijuana is highly unusual among the substances subject to the CSA in that it can be cultivated without any materials that have traveled in interstate commerce.) Lopez makes clear that possession is not itself commercial activity. Ibid. And respondents have not come into possession by means of any commercial transaction; they have simply grown, in their own homes, marijuana for their own use, without acquiring, buying, selling, or bartering a thing of value. Cf. id., at 583 (Kennedy, J., concurring) (“The statute now before us forecloses the States from experimenting … and it does so by regulating an activity beyond the realm of commerce in the ordinary and usual sense of that term”).
The Court suggests that Wickard, which we have identified as “perhaps the most far reaching example of Commerce Clause authority over intrastate activity,” Lopez, supra, at 560, established federal regulatory power over any home consumption of a commodity for which a national market exists. I disagree. Wickard involved a challenge to the Agricultural Adjustment Act of 1938 (AAA), which directed the Secretary of Agriculture to set national quotas on wheat production, and penalties for excess production. 317 U. S., at 115–116. The AAA itself confirmed that Congress made an explicit choice not to reach—and thus the Court could not possibly have approved of federal control over—small-scale, noncommercial wheat farming. In contrast to the CSA’s limitless assertion of power, Congress provided an exemption within the AAA for small producers. When Filburn planted the wheat at issue in Wickard, the statute exempted plantings less than 200 bushels (about six tons), and when he harvested his wheat it exempted plantings less than six acres. Id., at 130, n. 30. Wickard, then, did not extend Commerce Clause authority to something as modest as the home cook’s herb garden. This is not to say that Congress may never regulate small quantities of commodities possessed or produced for personal use, or to deny that it sometimes needs to enact a zero tolerance regime for such commodities. It is merely to say that Wickard did not hold or imply that small-scale production of commodities is always economic, and automatically within Congress’ reach.
Even assuming that economic activity is at issue in this case, the Government has made no showing in fact that the possession and use of homegrown marijuana for medical purposes, in California or elsewhere, has a substantial effect on interstate commerce. Similarly, the Government has not shown that regulating such activity is necessary to an interstate regulatory scheme. Whatever the specific theory of “substantial effects” at issue ( i.e., whether the activity substantially affects interstate commerce, whether its regulation is necessary to an interstate regulatory scheme, or both), a concern for dual sovereignty requires that Congress’ excursion into the traditional domain of States be justified.
That is why characterizing this as a case about the Necessary and Proper Clause does not change the analysis significantly. Congress must exercise its authority under the Necessary and Proper Clause in a manner consistent with basic constitutional principles. Garcia, 469 U. S., at 585 (O’Connor, J., dissenting) (“It is not enough that the ‘end be legitimate’; the means to that end chosen by Congress must not contravene the spirit of the Constitution”). As Justice Scalia recognizes, see ante, at 7 (opinion concurring in judgment), Congress cannot use its authority under the Clause to contravene the principle of state sovereignty embodied in the Tenth Amendment. Ibid. Likewise, that authority must be used in a manner consistent with the notion of enumerated powers—a structural principle that is as much part of the Constitution as the Tenth Amendment’s explicit textual command. Accordingly, something more than mere assertion is required when Congress purports to have power over local activity whose connection to an intrastate market is not self-evident. Otherwise, the Necessary and Proper Clause will always be a back door for unconstitutional federal regulation. Cf. Printz v. United States, 521 U. S. 898, 923 (1997) (the Necessary and Proper Clause is “the last, best hope of those who defend ultra vires congressional action”). Indeed, if it were enough in “substantial effects” cases for the Court to supply conceivable justifications for intrastate regulation related to an interstate market, then we could have surmised in Lopez that guns in school zones are “never more than an instant from the interstate market” in guns already subject to extensive federal regulation, ante, at 8 (Scalia, J., concurring in judgment), recast Lopez as a Necessary and Proper Clause case, and thereby upheld the Gun-Free School Zones Act of 1990. (According to the Court’s and the concurrence’s logic, for example, the Lopez court should have reasoned that the prohibition on gun possession in school zones could be an appropriate means of effectuating a related prohibition on “sell[ing]” or “deliver[ing]” firearms or ammunition to “any individual who the licensee knows or has reasonable cause to believe is less than eighteen years of age.” 18 U. S. C. §922(b)(1) (1988 ed., Supp. II).)
There is simply no evidence that homegrown medicinal marijuana users constitute, in the aggregate, a sizable enough class to have a discernable, let alone substantial, impact on the national illicit drug market—or otherwise to threaten the CSA regime. Explicit evidence is helpful when substantial effect is not “visible to the naked eye.” See Lopez, 514 U. S., at 563. And here, in part because common sense suggests that medical marijuana users may be limited in number and that California’s Compassionate Use Act and similar state legislation may well isolate activities relating to medicinal marijuana from the illicit market, the effect of those activities on interstate drug traffic is not self-evidently substantial.
In this regard, again, this case is readily distinguishable from Wickard. To decide whether the Secretary could regulate local wheat farming, the Court looked to “the actual effects of the activity in question upon interstate commerce.” 317 U. S., at 120. Critically, the Court was able to consider “actual effects” because the parties had “stipulated a summary of the economics of the wheat industry.” Id., at 125. After reviewing in detail the picture of the industry provided in that summary, the Court explained that consumption of homegrown wheat was the most variable factor in the size of the national wheat crop, and that on-site consumption could have the effect of varying the amount of wheat sent to market by as much as 20 percent. Id., at 127. With real numbers at hand, the Wickard Court could easily conclude that “a factor of such volume and variability as home-consumed wheat would have a substantial influence on price and market conditions” nationwide. Id., at 128; see also id., at 128–129 (“This record leaves us in no doubt” about substantial effects).
The Court recognizes that “the record in the Wickard case itself established the causal connection between the production for local use and the national market” and argues that “we have before us findings by Congress to the same effect.” Ante, at 17 (emphasis added). The Court refers to a series of declarations in the introduction to the CSA saying that (1) local distribution and possession of controlled substances causes “swelling” in interstate traffic; (2) local production and distribution cannot be distinguished from interstate production and distribution; (3) federal control over intrastate incidents “is essential to effective control” over interstate drug trafficking. 21 U. S. C. §§801(1)–(6). These bare declarations cannot be compared to the record before the Court in Wickard.
They amount to nothing more than a legislative insistence that the regulation of controlled substances must be absolute. They are asserted without any supporting evidence—descriptive, statistical, or otherwise. “[S]imply because Congress may conclude a particular activity substantially affects interstate commerce does not necessarily make it so.” Hodel v. Virginia Surface Mining & Reclamation Assn., Inc., 452 U. S. 264, 311 (1981) (Rehnquist, J., concurring in judgment). Indeed, if declarations like these suffice to justify federal regulation, and if the Court today is right about what passes rationality review before us, then our decision in Morrison should have come out the other way. In that case, Congress had supplied numerous findings regarding the impact gender-motivated violence had on the national economy. 529 U. S., at 614; id., at 628–636 (Souter, J., dissenting) (chronicling findings). But, recognizing that “ ‘ “[w]hether particular operations affect interstate commerce sufficiently to come under the constitutional power of Congress to regulate them is ultimately a judicial rather than a legislative question,” ’ ” we found Congress’ detailed findings inadequate. Id., at 614 (quoting Lopez, supra, at 557, n. 2, in turn quoting Heart of Atlanta Motel, Inc. v. United States, 379 U. S. 241, 273 (1964) (Black, J., concurring)). If, as the Court claims, today’s decision does not break with precedent, how can it be that voluminous findings, documenting extensive hearings about the specific topic of violence against women, did not pass constitutional muster in Morrison, while the CSA’s abstract, unsubstantiated, generalized findings about controlled substances do?
In particular, the CSA’s introductory declarations are too vague and unspecific to demonstrate that the federal statutory scheme will be undermined if Congress cannot exert power over individuals like respondents. The declarations are not even specific to marijuana. (Facts about substantial effects may be developed in litigation to compensate for the inadequacy of Congress’ findings; in part because this case comes to us from the grant of a preliminary injunction, there has been no such development.) Because here California, like other States, has carved out a limited class of activity for distinct regulation, the inadequacy of the CSA’s findings is especially glaring. The California Compassionate Use Act exempts from other state drug laws patients and their caregivers “who posses[s] or cultivat[e] marijuana for the personal medical purposes of the patient upon the written or oral recommendation of a physician” to treat a list of serious medical conditions. Cal. Health & Safety Code Ann. §§11362.5(d), 11362.7(h) (West Supp. 2005) (emphasis added). Compare ibid. with, e.g., §11357(b) (West 1991) (criminalizing marijuana possession in excess of 28.5 grams); §11358 (criminalizing marijuana cultivation). The Act specifies that it should not be construed to supersede legislation prohibiting persons from engaging in acts dangerous to others, or to condone the diversion of marijuana for nonmedical purposes. §11362.5(b)(2) (West Supp. 2005). To promote the Act’s operation and to facilitate law enforcement, California recently enacted an identification card system for qualified patients. §§11362.7–11362.83. We generally assume States enforce their laws, see Riley v. National Federation of Blind of N. C., Inc., 487 U. S. 781, 795 (1988), and have no reason to think otherwise here.
The Government has not overcome empirical doubt that the number of Californians engaged in personal cultivation, possession, and use of medical marijuana, or the amount of marijuana they produce, is enough to threaten the federal regime. Nor has it shown that Compassionate Use Act marijuana users have been or are realistically likely to be responsible for the drug’s seeping into the market in a significant way. The Government does cite one estimate that there were over 100,000 Compassionate Use Act users in California in 2004, Reply Brief for Petitioners 16, but does not explain, in terms of proportions, what their presence means for the national illicit drug market. See generally Wirtz, 392 U. S., at 196, n. 27 (Congress cannot use “a relatively trivial impact on commerce as an excuse for broad general regulation of state or private activities”); cf. General Accounting Office, Marijuana: Early Experience with Four States’ Laws That Allow Use for Medical Purposes 21–23 (Rep. No. 03–189, Nov. 2002), http://www.gao.gov/new.items/d03189.pdf (as visited June 3, 2005 and available in Clerk of Court’s case file) (in four California counties before the identification card system was enacted, voluntarily registered medical marijuana patients were less than 0.5 percent of the population; in Alaska, Hawaii, and Oregon, statewide medical marijuana registrants represented less than 0.05 percent of the States’ populations). It also provides anecdotal evidence about the CSA’s enforcement. See Reply Brief for Petitioners 17–18. The Court also offers some arguments about the effect of the Compassionate Use Act on the national market. It says that the California statute might be vulnerable to exploitation by unscrupulous physicians, that Compassionate Use Act patients may overproduce, and that the history of the narcotics trade shows the difficulty of cordoning off any drug use from the rest of the market. These arguments are plausible; if borne out in fact they could justify prosecuting Compassionate Use Act patients under the federal CSA. But, without substantiation, they add little to the CSA’s conclusory statements about diversion, essentiality, and market effect. Piling assertion upon assertion does not, in my view, satisfy the substantiality test of Lopez and Morrison.
III
We would do well to recall how James Madison, the father of the Constitution, described our system of joint sovereignty to the people of New York: “The powers delegated by the proposed constitution to the federal government are few and defined. Those which are to remain in the State governments are numerous and indefinite…. The powers reserved to the several States will extend to all the objects which, in the ordinary course of affairs, concern the lives, liberties, and properties of the people, and the internal order, improvement, and prosperity of the State.” The Federalist No. 45, pp. 292–293 (C. Rossiter ed. 1961).
Relying on Congress’ abstract assertions, the Court has endorsed making it a federal crime to grow small amounts of marijuana in one’s own home for one’s own medicinal use. This overreaching stifles an express choice by some States, concerned for the lives and liberties of their people, to regulate medical marijuana differently. If I were a California citizen, I would not have voted for the medical marijuana ballot initiative; if I were a California legislator I would not have supported the Compassionate Use Act. But whatever the wisdom of California’s experiment with medical marijuana, the federalism principles that have driven our Commerce Clause cases require that room for experiment be protected in this case. For these reasons I dissent.
Justice O’Connor delivered the opinion of the Court.
This case concerns respondent John David Stumpf ’s conviction and death sentence for the murder of Mary Jane Stout. In adjudicating Stumpf ’s petition for a writ of habeas corpus, the United States Court of Appeals for the Sixth Circuit granted him relief on two grounds: that his guilty plea was not knowing, voluntary, and intelligent, and that his conviction and sentence could not stand because the State, in a later trial of Stumpf ’s accomplice, pursued a theory of the case inconsistent with the theory it had advanced in Stumpf ’s case. We granted certiorari to review both holdings. 543 U. S. ___ (2005).
I
On May 14, 1984, Stumpf and two other men, Clyde Daniel Wesley and Norman Leroy Edmonds, were traveling in Edmonds’ car along Interstate 70 through Guernsey County, Ohio. Needing money for gas, the men stopped the car along the highway. While Edmonds waited in the car, Stumpf and Wesley walked to the home of Norman and Mary Jane Stout, about 100 yards away. Stumpf and Wesley, each concealing a gun, talked their way into the home by telling the Stouts they needed to use the phone. Their real object, however, was robbery: Once inside, Stumpf held the Stouts at gunpoint, while Wesley ransacked the house. When Mr. Stout moved toward Stumpf, Stumpf shot him twice in the head, causing Mr. Stout to black out. After he regained consciousness, Mr. Stout heard two male voices coming from another room, and then four gunshots—the shots that killed his wife. Edmonds was arrested shortly afterward, and his statements led the police to issue arrest warrants for Stumpf and Wesley. Stumpf, who surrendered to the police, at first denied any knowledge of the crimes. After he was told that Mr. Stout had survived, however, Stumpf admitted to participating in the robbery and to shooting Mr. Stumpf. But he claimed not to have shot Mrs. Stout, and he has maintained that position ever since.
The proceedings against Stumpf occurred while Wesley, who had been arrested in Texas, was still resisting extradition to Ohio. Stumpf was indicted for aggravated murder, attempted aggravated murder, aggravated robbery, and two counts of grand theft. With respect to the aggravated murder charge, the indictment listed four statutory “specifications”—three of them aggravating circumstances making Stumpf eligible for the death penalty. See App. 117–118; Ohio Rev. Code Ann. §2929.03 (Anderson 1982).* The case was assigned to a three-judge panel in the Court of Common Pleas.
Rather than proceed to trial, however, Stumpf and the State worked out a plea agreement: Stumpf would plead guilty to aggravated murder and attempted aggravated murder, and the State would drop most of the other charges; with respect to the aggravated murder charge, Stumpf would plead guilty to one of the three capital specifications, with the State dropping the other two. The plea was accepted after a colloquy with the presiding judge, and after a hearing in which the panel satisfied itself as to the factual basis for the plea.
Because the capital specification to which Stumpf pleaded guilty left him eligible for the death penalty, a contested penalty hearing was held before the same three-judge panel. Stumpf ’s mitigation case was based in part on his difficult childhood, limited education, dependable work history, youth, and lack of prior serious offenses. Stumpf ’s principal argument, however, was that he had participated in the plot only at the urging and under the influence of Wesley, that it was Wesley who had fired the fatal shots at Mrs. Stout, and that Stumpf ’s assertedly minor role in the murder counseled against the death sentence. See §2929.04(B)(6) (directing the sentencer to consider as a potential mitigating circumstance, “[i]f the offender was a participant in the offense but not the principal offender, the degree of the offender’s participation in the offense”). The State, on the other hand, argued that Stumpf had indeed shot Mrs. Stout. Still, while the prosecutor claimed Stumpf ’s allegedly primary role in the shooting as a special reason to reject Stumpf ’s mitigation argument, the prosecutor also noted that Ohio law did not restrict the death penalty to those who commit murder by their own hands—an accomplice to murder could also receive the death penalty, so long as he acted with the specific intent to cause death. As a result, the State argued, Stumpf deserved death even if he had not personally shot Mrs. Stout, because the circumstances of the robbery provided a basis from which to infer Stumpf ’s intent to cause death. The three-judge panel, agreeing with the State’s first contention, specifically found that Stumpf “was the principal offender” in the aggravated murder of Mrs. Stout. App. 196. Determining that the aggravating factors in Stumpf ’s case outweighed any mitigating factors, the panel sentenced Stumpf to death.
Afterward, Wesley was successfully extradited to Ohio to stand trial. His case was tried to a jury, before the same judge who had presided over the panel overseeing Stumpf ’s proceedings, and with the same prosecutor. This time, however, the prosecutor had new evidence: James Eastman, Wesley’s cellmate after his extradition, testified that Wesley had admitted to firing the shots that killed Mrs. Stout. The prosecutor introduced Eastman’s testimony in Wesley’s trial, and in his closing argument he argued for Wesley’s credibility and lack of motive to lie. The prosecutor claimed that Eastman’s testimony, combined with certain circumstantial evidence and with the implausibility of Wesley’s own account of events, proved that Wesley was the principal offender in Mrs. Stout’s murder—and that Wesley therefore deserved to be put to death. One way Wesley countered this argument was by noting that the prosecutor had taken a contrary position in Stumpf ’s trial, and that Stumpf had already been sentenced to death for the crime. Wesley also took the stand in his own defense, and testified that Stumpf had shot Mrs. Stout. In the end, the jury sentenced Wesley to life imprisonment with the possibility of parole after 20 years.
After the Wesley trial, Stumpf, whose direct appeal was still pending in the Ohio Court of Appeals, returned to the Court of Common Pleas with a motion to withdraw his guilty plea or vacate his death sentence. Stumpf argued that Eastman’s testimony, and the prosecution’s endorsement of that testimony in Wesley’s trial, cast doubt upon Stumpf ’s conviction and sentence. The State (represented again by the same prosecutor who had tried both Wesley’s case and Stumpf ’s original case) disagreed. According to the prosecutor, the court’s first task was to decide whether the Eastman testimony was sufficient to alter the court’s prior determination that Stumpf had been the shooter. Id., at 210. Contrary to the argument he had presented in the Wesley trial, however, the prosecutor now noted that Eastman’s testimony was belied by certain other evidence (ballistics evidence and Wesley’s testimony in his own defense) confirming Stumpf to have been the primary shooter. In the alternative, the State noted as it had before that an aider-and-abettor theory might allow the death sentence to be imposed against Stumpf even if he had not shot Mrs. Stout.
Although one judge speculated during oral argument that the court’s earlier conclusion about Stumpf ’s principal role in the killing “may very well have had an effect upon” the prior sentencing determination, ibid., the Court of Common Pleas denied Stumpf ’s motion in a brief summary order without explanation. That order was appealed together with the original judgment in Stumpf ’s case, and the Ohio Court of Appeals affirmed, as did the Ohio Supreme Court. State v. Stumpf, 32 Ohio St. 3d 95, 512 N. E. 2d 598 (1987), cert. denied, 484 U. S. 1079 (1988).
After a subsequent request for state postconviction relief was denied by the state courts, Stumpf filed this federal habeas petition in the United States District Court for the Southern District of Ohio in November 1995. The District Court denied Stumpf relief, but granted permission to appeal on four claims, including the two at issue here. The United States Court of Appeals for the Sixth Circuit reversed, concluding that habeas relief was warranted on “either or both” of “two alternative grounds.” Stumpf v. Mitchell, 367 F. 3d 594, 596 (2004). First, the court determined that Stumpf ’s guilty plea was invalid because it had not been entered knowingly and intelligently. More precisely, the court concluded that Stumpf had pleaded guilty to aggravated murder without understanding that specific intent to cause death was a necessary element of the charge under Ohio law. See Ohio Rev. Code Ann. §§2903.01(B) and (D). Noting that Stumpf had all along denied shooting Mrs. Stout, and considering those denials inconsistent with an informed choice to plead guilty to aggravated murder, the Court of Appeals concluded that Stumpf must have entered his plea out of ignorance. Second, the court concluded that “Stumpf ’s due process rights were violated by the state’s deliberate action in securing convictions of both Stumpf and Wesley for the same crime, using inconsistent theories.” 367 F. 3d, at 596. This violation, the court held, required setting aside “both Stumpf ’s plea and his sentence.” Id., at 616. One member of the panel dissented.
II
Because Stumpf filed his habeas petition before enactment of the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA), we review his claims under the standards of the pre-AEDPA habeas statute. See Lindh v. Murphy, 521 U. S. 320 (1997). Moreover, because petitioner has not argued that Stumpf ’s habeas claims were barred as requiring announcement of a new rule, we do not apply the rule of Teague v. Lane, 489 U. S. 288 (1989), to this case. See Schiro v. Farley, 510 U. S. 222, 229 (1994); Godinez v. Moran, 509 U. S. 389, 397, n. 8 (1993).
A
The Court of Appeals concluded that Stumpf ’s plea of guilty to aggravated murder was invalid because he was not aware of the specific intent element of the charge—a determination we find unsupportable.
Stumpf ’s guilty plea would indeed be invalid if he had not been aware of the nature of the charges against him, including the elements of the aggravated murder charge to which he pleaded guilty. A guilty plea operates as a waiver of important rights, and is valid only if done voluntarily, knowingly, and intelligently, “with sufficient awareness of the relevant circumstances and likely consequences.” Brady v. United States, 397 U. S. 742, 748 (1970). Where a defendant pleads guilty to a crime without having been informed of the crime’s elements, this standard is not met and the plea is invalid. Henderson v. Morgan, 426 U. S. 637 (1976).
But the Court of Appeals erred in finding that Stumpf had not been properly informed before pleading guilty. In Stumpf ’s plea hearing, his attorneys represented on the record that they had explained to their client the elements of the aggravated murder charge; Stumpf himself then confirmed that this representation was true. See App. 135, 137–138. While the court taking a defendant’s plea is responsible for ensuring “a record adequate for any review that may be later sought,” Boykin v. Alabama, 395 U. S. 238, 244 (1969), we have never held that the judge must himself explain the elements of each charge to the defendant on the record. Rather, the constitutional prerequisites of a valid plea may be satisfied where the record accurately reflects that the nature of the charge and the elements of the crime were explained to the defendant by his own, competent counsel. Cf. Henderson, supra, at 647 (granting relief to a defendant unaware of the elements of his crime, but distinguishing that case from others where “the record contains either an explanation of the charge by the trial judge, or at least a representation by defense counsel that the nature of the offense has been explained to the accused”). Where a defendant is represented by competent counsel, the court usually may rely on that counsel’s assurance that the defendant has been properly informed of the nature and elements of the charge to which he is pleading guilty.
Seeking to counter this natural inference, Stumpf argues, in essence, that his choice to plead guilty to the aggravated murder charge was so inconsistent with his denial of having shot the victim that he could only have pleaded guilty out of ignorance of the charge’s specific intent requirement. But Stumpf ’s asserted inconsistency is illusory. The aggravated murder charge’s intent element did not require any showing that Stumpf had himself shot Mrs. Stout. Rather, Ohio law considers aiders and abettors equally in violation of the aggravated murder statute, so long as the aiding and abetting is done with the specific intent to cause death. See In re Washington, 81 Ohio St. 3d 337, 691 N. E. 2d 285 (1998); State v. Scott, 61 Ohio St. 2d 155, 165, 400 N. E. 2d 375, 382 (1980). As a result, Stumpf ’s steadfast assertion that he had not shot Mrs. Stout would not necessarily have precluded him from admitting his specific intent under the statute.
That is particularly so given the other evidence in this case. Stumpf and Wesley had gone to the Stouts’ home together, carrying guns and intending to commit armed robbery. Stumpf, by his own admission, shot Mr. Stout in the head at close range. Taken together, these facts could show that Wesley and Stumpf had together agreed to kill both of the Stouts in order to leave no witnesses to the crime. And that, in turn, could make both men guilty of aggravated murder regardless of who actually killed Mrs. Stout. See ibid., at 165, 400 N. E. 2d, at 382.
Stumpf also points to aspects of the plea hearing transcript which he says show that both he and his attorneys were confused about the relevance and timing of defenses Stumpf and his attorneys had planned to make. First, at one point during the hearing, the presiding judge stated that by pleading guilty Stumpf would waive his trial rights and his right to testify in his own behalf. Stumpf ’s attorney answered that Stumpf “was going to respond but we have informed him that there is, after the plea, a hearing or trial relative to the underlying facts so that he is of the belief that there will be a presentation of evidence.” App. 140. The presiding judge responded that “[o]f course in the sentencing portion of this trial you do have those rights to speak in your own behalf [and] to present evidence and testimony on your own behalf.” Ibid. A few moments later, there was another exchange along similar lines, after the judge asked Stumpf whether he was “in fact guilty of” the aggravated murder charge and its capital specification:
“[DEFENSE COUNSEL]: … Your Honor, the defendant has asked me to explain his answer. His answer is yes. He will recite that with obviously his understanding of his right to present evidence at a later time relative to his conduct, but he’ll respond to that.
“JUDGE HENDERSON: At no time am I implying that the defendant will not have the right to present evidence in [the] mitigation hearing …. And I’m going to ask that the defendant, himself, respond to the question that I asked with that understanding that he has the right to present evidence in mitigation. I’m going to ask the defendant if he is in fact guilty of the charge set forth in Count one, including specification one … ?
“THE DEFENDANT: Yes, sir.” Id., at 142.
Reviewing this exchange, the Court of Appeals concluded that Stumpf “obviously … was reiterating his desire to challenge the [S]tate’s account of his actions”—that is, to show that he did not intend to kill Mrs. Stout. 367 F. 3d, at 607. But the desire to contest the State’s version of events would not necessarily entail the desire to contest the aggravated murder charge or any of its elements. Rather, Stumpf ’s desire to put on evidence “relative to the underlying facts” and “relative to his conduct” could equally have meant that Stumpf was eager to make his mitigation case—an interpretation bolstered by the attorney’s and Stumpf ’s approving answers after the presiding judge confirmed that the defense could put on evidence “in mitigation” and in “the sentencing” phase. While Stumpf ’s mitigation case was premised on the argument that Stumpf had not shot Mrs. Stout, that was fully consistent with his plea of guilty to aggravated murder. See supra, at 7–8.
Finally, Stumpf, like the Court of Appeals, relies on the perception that he obtained a bad bargain by his plea—that the State’s dropping several non-murder charges and two of the three capital murder specifications was a bad tradeoff for Stumpf ’s guilty plea. But a plea’s validity may not be collaterally attacked merely because the defendant made what turned out, in retrospect, to be a poor deal. See Brady, 397 U. S., at 757; Mabry v. Johnson, 467 U. S. 504, 508 (1984). Rather, the shortcomings of the deal Stumpf obtained cast doubt on the validity of his plea only if they show either that he made the unfavorable plea on the constitutionally defective advice of counsel, see Tollett v. Henderson, 411 U. S. 258, 267 (1973), or that he could not have understood the terms of the bargain he and Ohio agreed to. Though Stumpf did bring an independent claim asserting ineffective assistance of counsel, that claim is not before us in this case. And in evaluating the validity of Stumpf ’s plea, we are reluctant to accord much weight to his post hoc reevaluation of the wisdom of the bargain. Stumpf pleaded guilty knowing that the State had copious evidence against him, including the testimony of Mr. Stout; the plea eliminated two of the three capital specifications the State could rely on in seeking the death penalty; and the plea allowed Stumpf to assert his acceptance of responsibility as an argument in mitigation. Under these circumstances, the plea may well have been a knowing, voluntary, and intelligent reaction to a litigation situation that was difficult, to say the least. The Court of Appeals erred in concluding that Stumpf was uninformed about the nature of the charge he pleaded guilty to, and we reverse that portion of the judgment below.
B
The Court of Appeals was also wrong to hold that prosecutorial inconsistencies between the Stumpf and Wesley cases required voiding Stumpf ’s guilty plea. Stumpf ’s assertions of inconsistency relate entirely to the prosecutor’s arguments about which of the two men, Wesley or Stumpf, shot Mrs. Stout. For the reasons given above, see supra, at 7–8, the precise identity of the triggerman was immaterial to Stumpf ’s conviction for aggravated murder. Moreover, Stumpf has never provided an explanation of how the prosecution’s postplea use of inconsistent arguments could have affected the knowing, voluntary, and intelligent nature of his plea.
The prosecutor’s use of allegedly inconsistent theories may have a more direct effect on Stumpf ’s sentence, however, for it is at least arguable that the sentencing panel’s conclusion about Stumpf ’s principal role in the offense was material to its sentencing determination. The opinion below leaves some ambiguity as to the overlap between how the lower court resolved Stumpf ’s due process challenge to his conviction, and how it resolved Stumpf ’s challenge to his sentence. It is not clear whether the Court of Appeals would have concluded that Stumpf was entitled to resentencing had the court not also considered the conviction invalid. Likewise, the parties’ briefing to this Court, and the question on which we granted certiorari, largely focused on the lower court’s determination about Stumpf ’s conviction. See, e.g., Pet. for Cert. ii (requesting review of Stumpf ’s conviction, not sentence); Reply Brief for Petitioner 3 (challenge to Court of Appeals’ decision is focused on issue of conviction); Brief for Respondent 15, n. 3 (“arguments regarding Stumpf ’s death sentence are not before this Court”). In these circumstances, it would be premature for this Court to resolve the merits of Stumpf ’s sentencing claim, and we therefore express no opinion on whether the prosecutor’s actions amounted to a due process violation, or whether any such violation would have been prejudicial. The Court of Appeals should have the opportunity to consider, in the first instance, the question of how Eastman’s testimony and the prosecutor’s conduct in the Stumpf and Wesley cases relate to Stumpf ’s death sentence in particular. Accordingly, we vacate the portion of the judgment below relating to Stumpf ’s prosecutorial inconsistency claim, and we remand the case for further proceedings consistent with this opinion.
It is so ordered.
Notes
* Unless otherwise noted, all citations to Ohio statutes refer to the versions of those statutes in effect in 1984, at the time of the crime and trial.
Justice O’Connor delivered the opinion of the Court.
Title 28 U. S. C. §2255 establishes a “1-year period of limitation” within which a federal prisoner may file a motion to vacate, set aside, or correct his sentence under that section. That period runs from “the latest” of a number of events, which are enumerated in subparagraphs (1) through (4) of ¶6 of that section. This case involves subparagraph (3), which provides that the limitation period begins to run on “the date on which the right asserted was initially recognized by the Supreme Court, if that right has been newly recognized by the Supreme Court and made retroactively applicable to cases on collateral review.” We must decide whether the date from which the limitation period begins to run under ¶6(3) is the date on which this Court “initially recognized” the right asserted in an applicant’s §2255 motion, or whether, instead, it is the date on which the right is “made retroactiv[e].”
I
Petitioner Michael Donald Dodd was indicted on June 25, 1993, for knowingly and intentionally engaging in a continuing criminal enterprise in violation of 21 U. S. C. §§841 and 846, conspiring to possess with intent to distribute marijuana in violation of §841(a)(1), conspiring to possess with intent to distribute cocaine in violation of §841(a)(1), and 16 counts of using and possessing a passport obtained by false statement in violation of 18 U. S. C. §1546(a). He was convicted of all counts except the cocaine charge, and was sentenced to 360 months’ imprisonment followed by five years of supervised release. The Court of Appeals for the Eleventh Circuit affirmed on May 7, 1997. 111 F. 3d 867 (per curiam). Because Dodd did not file a petition for certiorari, his conviction became final on August 6, 1997. See Clay v. United States, 537 U. S. 522, 525 (2003).
On April 4, 2001, more than three years after his conviction became final, Dodd filed a pro se motion under 28 U. S. C. §2255 seeking to set aside his conviction for knowingly and intentionally engaging in a continuing criminal enterprise, based on our decision in Richardson v. United States, 526 U. S. 813 (1999). Richardson held that a jury must agree unanimously that a defendant is guilty of each of the specific violations that together constitute the continuing criminal enterprise. Id., at 815. Dodd argued, among other things, that he was entitled to relief because his jury had not been instructed that they had to agree unanimously on each predicate violation. App. 9. The District Court dismissed Dodd’s §2255 motion as time barred. Id., at 11–15. Because Richardson had been decided more than one year before Dodd filed his motion, the court held that the motion was untimely; it also rejected Dodd’s request for equitable tolling. App. 13–15.
Dodd appealed, arguing that the limitation period in §2255, ¶6(3), did not begin to run until April 19, 2002, when the Court of Appeals for the Eleventh Circuit held in Ross v. United States, 289 F. 3d 677, that the right recognized in Richardson applies retroactively to cases on collateral review. The Eleventh Circuit held that the limitation period began to run on “the date the Supreme Court initially recognizes the right”—the date Richardson was decided—and accordingly affirmed the dismissal of Dodd’s motion as time barred. 365 F. 3d 1273, 1283 (2004).
We granted certiorari, 543 U. S. __ (2004), to resolve a conflict in the Courts of Appeals over when the limitation period in ¶6(3) begins to run. Compare, e.g., 365 F. 3d, at 1283 (case below) (period runs from date of Supreme Court decision initially recognizing right asserted); and United States v. Lopez, 248 F. 3d 427, 432–433 (CA5 2001) (same), with Pryor v. United States, 278 F. 3d 612, 616 (CA6 2002) (period does not begin to run until right has been held retroactively applicable to cases on collateral review); and United States v. Valdez, 195 F. 3d 544, 547–548 (CA9 1999) (same).
II
Section 2255, ¶6, provides:
“A 1-year period of limitation shall apply to a motion under this section. The limitation period shall run from the latest of—
“(1) the date on which the judgment of conviction becomes final;
“(2) the date on which the impediment to making a motion created by governmental action in violation of the Constitution or laws of the United States is removed, if the movant was prevented from making a motion by such governmental action;
“(3) the date on which the right asserted was initially recognized by the Supreme Court, if that right has been newly recognized by the Supreme Court and made retroactively applicable to cases on collateral review; or
“(4) the date on which the facts supporting the claim or claims presented could have been discovered through the exercise of due diligence.”
In most cases, the operative date from which the limitation period is measured will be the one identified in ¶6(1): “the date on which the judgment of conviction becomes final.” Ibid. ; see also Clay, supra, at 524. But later filings are permitted where subparagraphs (2)–(4) apply. This case involves ¶6(3), which gives §2255 applicants one year from “the date on which the right asserted was initially recognized by the Supreme Court, if that right has been newly recognized by the Supreme Court and made retroactively applicable to cases on collateral review.” Dodd contends that under subparagraph (3), the limitation period runs from the date on which the right asserted was made retroactively applicable. The United States, on the other hand, argues that it runs from the date on which this Court initially recognized the right asserted.
We believe that the text of ¶6(3) settles this dispute. It unequivocally identifies one, and only one, date from which the 1-year limitation period is measured: “the date on which the right asserted was initially recognized by the Supreme Court.” We “must presume that [the] legislature says in a statute what it means and means in a statute what it says there.” Connecticut Nat. Bank v. Germain, 503 U. S. 249, 253–254 (1992). What Congress has said in ¶6(3) is clear: an applicant has one year from the date on which the right he asserts was initially recognized by this Court.
Dodd urges us to adopt a different interpretation. He contends that the second clause in ¶6(3) affects the applicable date under that provision. He reads ¶6(3) as containing “three distinct prerequisites” that “must be satisfied before the limitation period begins.” Brief for Petitioner 8. Those three prerequisites are: (1) the right asserted by the applicant “was initially recognized” by this Court; (2) this Court “newly recognized” the right; and (3) a court must have “made” the right “retroactively applicable to cases on collateral review.” Id., at 13–14 (internal quotation marks omitted). Because the Court of Appeals for the Eleventh Circuit did not hold the right recognized in Richardson v. United States, 526 U. S. 813 (1999), retroactively applicable until April 19, 2002, when it decided Ross, 289 F. 3d 677, Dodd contends that he had until April 19, 2003—one year from the date when all three prerequisites were satisfied—to file his §2255 motion.
Dodd’s interpretation does not square with the only natural reading of the text. Paragraph 6(3) identifies one date and one date only as the date from which the 1-year limitation period runs: “the date on which the right asserted was initially recognized by the Supreme Court.” Dodd’s reliance on the second clause to identify the operative date is misplaced. That clause—“if that right has been newly recognized by the Supreme Court and made retroactively applicable to cases on collateral review”—imposes a condition on the applicability of this subsection. See Webster’s Third New International Dictionary 1124 (1993) (the definition of “if ” is “in the event that” or “on condition that”). It therefore limits ¶6(3)’s application to cases in which applicants are seeking to assert rights “newly recognized by the Supreme Court and made retroactively applicable to cases on collateral review.” §2255, ¶6(3). That means that ¶6(3)’s date—“the date on which the right asserted was initially recognized by the Supreme Court”—does not apply at all if the conditions in the second clause—the right “has been newly recognized by the Supreme Court and made retroactively applicable to cases on collateral review”—have not been satisfied. As long as the conditions in the second clause are satisfied so that ¶6(3) applies in the first place, that clause has no impact whatsoever on the date from which the 1-year limitation period in ¶6(3) begins to run. Thus, if this Court decides a case recognizing a new right, a federal prisoner seeking to assert that right will have one year from this Court’s decision within which to file his §2255 motion. He may take advantage of the date in the first clause of ¶6(3) only if the conditions in the second clause are met.
We recognize that the statute of limitations in ¶6(3) makes it difficult for applicants filing second or successive §2255 motions to obtain relief. The limitation period in ¶6(3) applies to “all motions” under §2255, initial motions as well as second or successive ones. Section 2255, ¶8(2), narrowly restricts an applicant’s ability to file a second or successive motion. An applicant may file a second or successive motion only in limited circumstances, such as where he seeks to take advantage of “a new rule of constitutional law, made retroactive to cases on collateral review by the Supreme Court, that was previously unavailable.” §2255, ¶8(2). Dodd points out that this Court rarely decides that a new rule is retroactively applicable within one year of initially recognizing that right. Thus, because of the interplay between ¶¶8(2) and 6(3), an applicant who files a second or successive motion seeking to take advantage of a new rule of constitutional law will be time barred except in the rare case in which this Court announces a new rule of constitutional law and makes it retroactive within one year.
Although we recognize the potential for harsh results in some cases, we are not free to rewrite the statute that Congress has enacted. “[W]hen the statute’s language is plain, the sole function of the courts—at least where the disposition required by the text is not absurd—is to enforce it according to its terms.” Hartford Underwriters Ins. Co. v. Union Planters Bank, N. A., 530 U. S. 1, 6 (2000) (internal quotation marks omitted). See also Tyler v. Cain, 533 U. S. 656, 663, n. 5 (2001) (“[E]ven if we disagreed with the legislative decision to establish stringent procedural requirements for retroactive application of new rules, we do not have license to question the decision on policy grounds”). The disposition required by the text here, though strict, is not absurd. It is for Congress, not this Court, to amend the statute if it believes that the interplay of ¶¶8(2) and 6(3) of §2255 unduly restricts federal prisoners’ ability to file second or successive motions.
Justice Stevens would hold, contrary to the plain text, that the limitation period in ¶6(3) begins to run when the right asserted is made retroactive, see post, at 9–10 (dissenting opinion), because he assumes that “the most natural reading of the statutory text would make it possible for the limitations period to expire before the cause of action accrues,” post, at 1. Justice Stevens analogizes this case to Graham County Soil & Water Conservation Dist. v. United States, ex rel. Wilson, post, p. __, see post, at 1 (dissenting opinion), but Graham Count y is distinguishable. The text of the statute at issue in Graham County is ambiguous, justifying the Court’s partial reliance on “the ‘standard rule that the limitations period commences when the plaintiff has a complete and present cause of action.’ ” See Graham County, post, at 5–8, and n. 2. Here, there is no such ambiguity; ¶6(3) clearly specifies the date on which the limitation period begins to run.
III
Dodd’s §2255 motion sought to benefit from our holding in Richardson, supra, which was decided on June 1, 1999. Thus, he had one year from that date within which to file his motion. Because he did not file his motion until April 4, 2001, the motion was untimely. We therefore affirm the judgment of the Court of Appeals for the Eleventh Circuit.
It is so ordered.
Justice O’Connor, concurring.
I write separately to put to rest one concern. The dissent worries that the Court’s opinion “imposes on defense counsel a rigid requirement to review all documents in what it calls the ‘case file’ of any prior conviction that the prosecution might rely on at trial.” Post, at 1 (opinion of Kennedy, J.). But the Court’s opinion imposes no such rule. See ante, at 14. Rather, today’s decision simply applies our longstanding case-by-case approach to determining whether an attorney’s performance was unconstitutionally deficient under Strickland v. Washington, 466 U. S. 668 (1984). Trial counsel’s performance in Rompilla’s case falls short under that standard, because the attorneys’ behavior was not “reasonable considering all the circumstances.” Id., at 688. In particular, there were three circumstances which made the attorneys’ failure to examine Rompilla’s prior conviction file unreasonable.
First, Rompilla’s attorneys knew that their client’s prior conviction would be at the very heart of the prosecution’s case. The prior conviction went not to a collateral matter, but rather to one of the aggravating circumstances making Rompilla eligible for the death penalty. The prosecutors intended not merely to mention the fact of prior conviction, but to read testimony about the details of the crime. That crime, besides being quite violent in its own right, was very similar to the murder for which Rompilla was on trial, and Rompilla had committed the murder at issue a mere three months after his release from prison on the earlier conviction. In other words, the prosecutor clearly planned to use details of the prior crime as powerful evidence that Rompilla was a dangerous man for whom the death penalty would be both appropriate punishment and a necessary means of incapacitation. Cf. App. 165–166 (prosecutor’s penalty-phase argument). This was evidence the defense should have been prepared to meet: A reasonable defense lawyer would have attached a high importance to obtaining the record of the prior trial, in order to anticipate and find ways of deflecting the prosecutor’s aggravation argument.
Second, the prosecutor’s planned use of the prior conviction threatened to eviscerate one of the defense’s primary mitigation arguments. Rompilla was convicted on the basis of strong circumstantial evidence. His lawyers structured the entire mitigation argument around the hope of convincing the jury that residual doubt about Rompilla’s guilt made it inappropriate to impose the death penalty. In announcing an intention to introduce testimony about Rompilla’s similar prior offense, the prosecutor put Rompilla’s attorneys on notice that the prospective defense on mitigation likely would be ineffective and counterproductive. The similarities between the two crimes, combined with the timing and the already strong circumstantial evidence, raised a strong likelihood that the jury would reject Rompilla’s residual doubt argument. Rompilla’s attorneys’ reliance on this transparently weak argument risked damaging their credibility. Such a scenario called for further investigation, to determine whether circumstances of the prior case gave any hope of saving the residual doubt argument, or whether the best strategy instead would be to jettison that argument so as to focus on other, more promising issues. Cf. Yarborough v. Gentry, 540 U. S. 1, 7 (2003) (per curiam); Bell v. Cone, 535 U. S. 685, 700 (2002) (noting that sound tactical judgment may sometimes call for omitting certain defense evidence or arguments).
Third, the attorneys’ decision not to obtain Rompilla’s prior conviction file was not the result of an informed tactical decision about how the lawyers’ time would best be spent. Although Rompilla’s attorneys had ample warning that the details of Rompilla’s prior conviction would be critical to their case, their failure to obtain that file would not necessarily have been deficient if it had resulted from the lawyers’ careful exercise of judgment about how best to marshal their time and serve their client. But Rompilla’s attorneys did not ignore the prior case file in order to spend their time on other crucial leads. They did not determine that the file was so inaccessible or so large that examining it would necessarily divert them from other trial-preparation tasks they thought more promising. They did not learn at the 11th hour about the prosecution’s intent to use the prior conviction, when it was too late for them to change plans. Rather, their failure to obtain the crucial file “was the result of inattention, not reasoned strategic judgment.” Wiggins v. Smith, 539 U. S. 510, 534 (2003). As a result, their conduct fell below constitutionally required standards. See id., at 533 (“ ‘[S]trategic choices made after less than complete investigation are reasonable’ only to the extent that ‘reasonable professional judgments support the limitations on investigation’ ” (quoting Strickland, 466 U. S., at 690–691)).
In the particular circumstances of this case, the attorneys’ failure to obtain and review the case file from their client’s prior conviction did not meet standards of “reasonable professional judgmen[t].” Id., at 691. Because the Court’s opinion is consistent with the “ ‘case-by-case examination of the evidence’ ” called for under our cases, Williams v. Taylor, 529 U. S. 362, 391 (2000), I join the opinion.
Justice O’Connor, with whom The Chief Justice, Justice Scalia, and Justice Thomas join, dissenting.
Over two centuries ago, just after the Bill of Rights was ratified, Justice Chase wrote:
“An act of the Legislature (for I cannot call it a law) contrary to the great first principles of the social compact, cannot be considered a rightful exercise of legislative authority …. A few instances will suffice to explain what I mean…. [A] law that takes property from A. and gives it to B: It is against all reason and justice, for a people to entrust a Legislature with such powers; and, therefore, it cannot be presumed that they have done it.” Calder v. Bull, 3 Dall. 386, 388 (1798) (emphasis deleted).
Today the Court abandons this long-held, basic limitation on government power. Under the banner of economic development, all private property is now vulnerable to being taken and transferred to another private owner, so long as it might be upgraded— i.e., given to an owner who will use it in a way that the legislature deems more beneficial to the public—in the process. To reason, as the Court does, that the incidental public benefits resulting from the subsequent ordinary use of private property render economic development takings “for public use” is to wash out any distinction between private and public use of property—and thereby effectively to delete the words “for public use” from the Takings Clause of the Fifth Amendment. Accordingly I respectfully dissent.
I
Petitioners are nine resident or investment owners of 15 homes in the Fort Trumbull neighborhood of New London, Connecticut. Petitioner Wilhelmina Dery, for example, lives in a house on Walbach Street that has been in her family for over 100 years. She was born in the house in 1918; her husband, petitioner Charles Dery, moved into the house when they married in 1946. Their son lives next door with his family in the house he received as a wedding gift, and joins his parents in this suit. Two petitioners keep rental properties in the neighborhood.
In February 1998, Pfizer Inc., the pharmaceuticals manufacturer, announced that it would build a global research facility near the Fort Trumbull neighborhood. Two months later, New London’s city council gave initial approval for the New London Development Corporation (NLDC) to prepare the development plan at issue here. The NLDC is a private, nonprofit corporation whose mission is to assist the city council in economic development planning. It is not elected by popular vote, and its directors and employees are privately appointed. Consistent with its mandate, the NLDC generated an ambitious plan for redeveloping 90 acres of Fort Trumbull in order to “complement the facility that Pfizer was planning to build, create jobs, increase tax and other revenues, encourage public access to and use of the city’s waterfront, and eventually ‘build momentum’ for the revitalization of the rest of the city.” App. to Pet. for Cert. 5.
Petitioners own properties in two of the plan’s seven parcels—Parcel 3 and Parcel 4A. Under the plan, Parcel 3 is slated for the construction of research and office space as a market develops for such space. It will also retain the existing Italian Dramatic Club (a private cultural organization) though the homes of three plaintiffs in that parcel are to be demolished. Parcel 4A is slated, mysteriously, for “ ‘park support.’ ” Id., at 345–346. At oral argument, counsel for respondents conceded the vagueness of this proposed use, and offered that the parcel might eventually be used for parking. Tr. of Oral Arg. 36.
To save their homes, petitioners sued New London and the NLDC, to whom New London has delegated eminent domain power. Petitioners maintain that the Fifth Amendment prohibits the NLDC from condemning their properties for the sake of an economic development plan. Petitioners are not hold-outs; they do not seek increased compensation, and none is opposed to new development in the area. Theirs is an objection in principle: They claim that the NLDC’s proposed use for their confiscated property is not a “public” one for purposes of the Fifth Amendment. While the government may take their homes to build a road or a railroad or to eliminate a property use that harms the public, say petitioners, it cannot take their property for the private use of other owners simply because the new owners may make more productive use of the property.
II
The Fifth Amendment to the Constitution, made applicable to the States by the Fourteenth Amendment, provides that “private property [shall not] be taken for public use, without just compensation.” When interpreting the Constitution, we begin with the unremarkable presumption that every word in the document has independent meaning, “that no word was unnecessarily used, or needlessly added.” Wright v. United States, 302 U. S. 583, 588 (1938). In keeping with that presumption, we have read the Fifth Amendment’s language to impose two distinct conditions on the exercise of eminent domain: “the taking must be for a ‘public use’ and ‘just compensation’ must be paid to the owner.” Brown v. Legal Foundation of Wash., 538 U. S. 216, 231–232 (2003).
These two limitations serve to protect “the security of Property,” which Alexander Hamilton described to the Philadelphia Convention as one of the “great obj[ects] of Gov[ernment].” 1 Records of the Federal Convention of 1787, p. 302 (M. Farrand ed. 1934). Together they ensure stable property ownership by providing safeguards against excessive, unpredictable, or unfair use of the government’s eminent domain power—particularly against those owners who, for whatever reasons, may be unable to protect themselves in the political process against the majority’s will.
While the Takings Clause presupposes that government can take private property without the owner’s consent, the just compensation requirement spreads the cost of condemnations and thus “prevents the public from loading upon one individual more than his just share of the burdens of government.” Monongahela Nav. Co. v. United States, 148 U. S. 312, 325 (1893); see also Armstrong v. United States, 364 U. S. 40, 49 (1960). The public use requirement, in turn, imposes a more basic limitation, circumscribing the very scope of the eminent domain power: Government may compel an individual to forfeit her property for the public’s use, but not for the benefit of another private person. This requirement promotes fairness as well as security. Cf. Tahoe-Sierra Preservation Council, Inc. v. Tahoe Regional Planning Agency, 535 U. S. 302, 336 (2002) (“The concepts of ‘fairness and justice’ … underlie the Takings Clause”).
Where is the line between “public” and “private” property use? We give considerable deference to legislatures’ determinations about what governmental activities will advantage the public. But were the political branches the sole arbiters of the public-private distinction, the Public Use Clause would amount to little more than hortatory fluff. An external, judicial check on how the public use requirement is interpreted, however limited, is necessary if this constraint on government power is to retain any meaning. See Cincinnati v. Vester, 281 U. S. 439, 446 (1930) (“It is well established that … the question [of] what is a public use is a judicial one”).
Our cases have generally identified three categories of takings that comply with the public use requirement, though it is in the nature of things that the boundaries between these categories are not always firm. Two are relatively straightforward and uncontroversial. First, the sovereign may transfer private property to public ownership—such as for a road, a hospital, or a military base. See, e.g., Old Dominion Land Co. v. United States, 269 U. S. 55 (1925); Rindge Co. v. County of Los Angeles, 262 U. S. 700 (1923). Second, the sovereign may transfer private property to private parties, often common carriers, who make the property available for the public’s use—such as with a railroad, a public utility, or a stadium. See, e.g., National Railroad Passenger Corporation v. Boston & Maine Corp., 503 U. S. 407 (1992); Mt. Vernon-Woodberry Cotton Duck Co. v. Alabama Interstate Power Co., 240 U. S. 30 (1916). But “public ownership” and “use-by-the-public” are sometimes too constricting and impractical ways to define the scope of the Public Use Clause. Thus we have allowed that, in certain circumstances and to meet certain exigencies, takings that serve a public purpose also satisfy the Constitution even if the property is destined for subsequent private use. See, e.g., Berman v. Parker, 348 U. S. 26 (1954); Hawaii Housing Authority v. Midkiff, 467 U. S. 229 (1984).
This case returns us for the first time in over 20 years to the hard question of when a purportedly “public purpose” taking meets the public use requirement. It presents an issue of first impression: Are economic development takings constitutional? I would hold that they are not. We are guided by two precedents about the taking of real property by eminent domain. In Berman, we upheld takings within a blighted neighborhood of Washington, D. C. The neighborhood had so deteriorated that, for example, 64.3% of its dwellings were beyond repair. 348 U. S., at 30. It had become burdened with “overcrowding of dwellings,” “lack of adequate streets and alleys,” and “lack of light and air.” Id., at 34. Congress had determined that the neighborhood had become “injurious to the public health, safety, morals, and welfare” and that it was necessary to “eliminat[e] all such injurious conditions by employing all means necessary and appropriate for the purpose,” including eminent domain. Id., at 28. Mr. Berman’s department store was not itself blighted. Having approved of Congress’ decision to eliminate the harm to the public emanating from the blighted neighborhood, however, we did not second-guess its decision to treat the neighborhood as a whole rather than lot-by-lot. Id., at 34–35; see also Midkiff, 467 U. S., at 244 (“it is only the taking’s purpose, and not its mechanics, that must pass scrutiny”).
In Midkiff, we upheld a land condemnation scheme in Hawaii whereby title in real property was taken from lessors and transferred to lessees. At that time, the State and Federal Governments owned nearly 49% of the State’s land, and another 47% was in the hands of only 72 private landowners. Concentration of land ownership was so dramatic that on the State’s most urbanized island, Oahu, 22 landowners owned 72.5% of the fee simple titles. Id., at 232. The Hawaii Legislature had concluded that the oligopoly in land ownership was “skewing the State’s residential fee simple market, inflating land prices, and injuring the public tranquility and welfare,” and therefore enacted a condemnation scheme for redistributing title. Ibid.
In those decisions, we emphasized the importance of deferring to legislative judgments about public purpose. Because courts are ill-equipped to evaluate the efficacy of proposed legislative initiatives, we rejected as unworkable the idea of courts’ “ ‘deciding on what is and is not a governmental function and … invalidating legislation on the basis of their view on that question at the moment of decision, a practice which has proved impracticable in other fields.’ ” Id., at 240–241 (quoting United States ex rel. TVA v. Welch, 327 U. S. 546, 552 (1946)); see Berman, supra, at 32 (“[T]he legislature, not the judiciary, is the main guardian of the public needs to be served by social legislation”); see also Lingle v. Chevron U. S. A., Inc., 544 U. S. __ (2005). Likewise, we recognized our inability to evaluate whether, in a given case, eminent domain is a necessary means by which to pursue the legislature’s ends. Midkiff, supra, at 242; Berman, supra, at 103.
Yet for all the emphasis on deference, Berman and Midkiff hewed to a bedrock principle without which our public use jurisprudence would collapse: “A purely private taking could not withstand the scrutiny of the public use requirement; it would serve no legitimate purpose of government and would thus be void.” Midkiff, 467 U. S., at 245; id., at 241 (“[T]he Court’s cases have repeatedly stated that ‘one person’s property may not be taken for the benefit of another private person without a justifying public purpose, even though compensation be paid’ ” (quoting Thompson v. Consolidated Gas Util. Corp., 300 U. S. 55, 80 (1937))); see also Missouri Pacific R. Co. v. Nebraska, 164 U. S. 403, 417 (1896). To protect that principle, those decisions reserved “a role for courts to play in reviewing a legislature’s judgment of what constitutes a public use … [though] the Court in Berman made clear that it is ‘an extremely narrow’ one.” Midkiff, supra, at 240 (quoting Berman, supra, at 32).
The Court’s holdings in Berman and Midkiff were true to the principle underlying the Public Use Clause. In both those cases, the extraordinary, precondemnation use of the targeted property inflicted affirmative harm on society—in Berman through blight resulting from extreme poverty and in Midkiff through oligopoly resulting from extreme wealth. And in both cases, the relevant legislative body had found that eliminating the existing property use was necessary to remedy the harm. Berman, supra, at 28–29; Midkiff, supra, at 232. Thus a public purpose was realized when the harmful use was eliminated. Because each taking directly achieved a public benefit, it did not matter that the property was turned over to private use. Here, in contrast, New London does not claim that Susette Kelo’s and Wilhelmina Dery’s well-maintained homes are the source of any social harm. Indeed, it could not so claim without adopting the absurd argument that any single-family home that might be razed to make way for an apartment building, or any church that might be replaced with a retail store, or any small business that might be more lucrative if it were instead part of a national franchise, is inherently harmful to society and thus within the government’s power to condemn.
In moving away from our decisions sanctioning the condemnation of harmful property use, the Court today significantly expands the meaning of public use. It holds that the sovereign may take private property currently put to ordinary private use, and give it over for new, ordinary private use, so long as the new use is predicted to generate some secondary benefit for the public—such as increased tax revenue, more jobs, maybe even aesthetic pleasure. But nearly any lawful use of real private property can be said to generate some incidental benefit to the public. Thus, if predicted (or even guaranteed) positive side-effects are enough to render transfer from one private party to another constitutional, then the words “for public use” do not realistically exclude any takings, and thus do not exert any constraint on the eminent domain power.
There is a sense in which this troubling result follows from errant language in Berman and Midkiff. In discussing whether takings within a blighted neighborhood were for a public use, Berman began by observing: “We deal, in other words, with what traditionally has been known as the police power.” 348 U. S., at 32. From there it declared that “[o]nce the object is within the authority of Congress, the right to realize it through the exercise of eminent domain is clear.” Id., at 33. Following up, we said in Midkiff that “[t]he ‘public use’ requirement is coterminous with the scope of a sovereign’s police powers.” 467 U. S., at 240. This language was unnecessary to the specific holdings of those decisions. Berman and Midkiff simply did not put such language to the constitutional test, because the takings in those cases were within the police power but also for “public use” for the reasons I have described. The case before us now demonstrates why, when deciding if a taking’s purpose is constitutional, the police power and “public use” cannot always be equated. The Court protests that it does not sanction the bare transfer from A to B for B’s benefit. It suggests two limitations on what can be taken after today’s decision. First, it maintains a role for courts in ferreting out takings whose sole purpose is to bestow a benefit on the private transferee—without detailing how courts are to conduct that complicated inquiry. Ante, at 7. For his part, Justice Kennedy suggests that courts may divine illicit purpose by a careful review of the record and the process by which a legislature arrived at the decision to take—without specifying what courts should look for in a case with different facts, how they will know if they have found it, and what to do if they do not. Ante, at 2–3 (concurring opinion). Whatever the details of Justice Kennedy’s as-yet-undisclosed test, it is difficult to envision anyone but the “stupid staff[er]” failing it. See Lucas v. South Carolina Coastal Council, 505 U. S. 1003, 1025–1026, n. 12 (1992). The trouble with economic development takings is that private benefit and incidental public benefit are, by definition, merged and mutually reinforcing. In this case, for example, any boon for Pfizer or the plan’s developer is difficult to disaggregate from the promised public gains in taxes and jobs. See App. to Pet. for Cert. 275–277.
Even if there were a practical way to isolate the motives behind a given taking, the gesture toward a purpose test is theoretically flawed. If it is true that incidental public benefits from new private use are enough to ensure the “public purpose” in a taking, why should it matter, as far as the Fifth Amendment is concerned, what inspired the taking in the first place? How much the government does or does not desire to benefit a favored private party has no bearing on whether an economic development taking will or will not generate secondary benefit for the public. And whatever the reason for a given condemnation, the effect is the same from the constitutional perspective—private property is forcibly relinquished to new private ownership.
A second proposed limitation is implicit in the Court’s opinion. The logic of today’s decision is that eminent domain may only be used to upgrade—not downgrade—property. At best this makes the Public Use Clause redundant with the Due Process Clause, which already prohibits irrational government action. See Lingle, 544 U. S. __. The Court rightfully admits, however, that the judiciary cannot get bogged down in predictive judgments about whether the public will actually be better off after a property transfer. In any event, this constraint has no realistic import. For who among us can say she already makes the most productive or attractive possible use of her property? The specter of condemnation hangs over all property. Nothing is to prevent the State from replacing any Motel 6 with a Ritz-Carlton, any home with a shopping mall, or any farm with a factory. Cf. Bugryn v. Bristol, 63 Conn. App. 98, 774 A. 2d 1042 (2001) (taking the homes and farm of four owners in their 70’s and 80’s and giving it to an “industrial park”); 99 Cents Only Stores v. Lancaster Redevelopment Authority, 237 F. Supp. 2d 1123 (CD Cal. 2001) (attempted taking of 99 Cents store to replace with a Costco); Poletown Neighborhood Council v. Detroit, 410 Mich. 616, 304 N. W. 2d 455 (1981) (taking a working-class, immigrant community in Detroit and giving it to a General Motors assembly plant), overruled by County of Wayne v. Hathcock, 471 Mich. 415, 684 N. W. 2d 765 (2004); Brief for the Becket Fund for Religious Liberty as Amicus Curiae 4–11 (describing takings of religious institutions’ properties); Institute for Justice, D. Berliner, Public Power, Private Gain: A Five-Year, State-by-State Report Examining the Abuse of Eminent Domain (2003) (collecting accounts of economic development takings).
The Court also puts special emphasis on facts peculiar to this case: The NLDC’s plan is the product of a relatively careful deliberative process; it proposes to use eminent domain for a multipart, integrated plan rather than for isolated property transfer; it promises an array of incidental benefits (even aesthetic ones), not just increased tax revenue; it comes on the heels of a legislative determination that New London is a depressed municipality. See, e.g., ante, at 16 (“[A] one-to-one transfer of property, executed outside the confines of an integrated development plan, is not presented in this case”). Justice Kennedy, too, takes great comfort in these facts. Ante, at 4 (concurring opinion). But none has legal significance to blunt the force of today’s holding. If legislative prognostications about the secondary public benefits of a new use can legitimate a taking, there is nothing in the Court’s rule or in Justice Kennedy’s gloss on that rule to prohibit property transfers generated with less care, that are less comprehensive, that happen to result from less elaborate process, whose only projected advantage is the incidence of higher taxes, or that hope to transform an already prosperous city into an even more prosperous one.
Finally, in a coda, the Court suggests that property owners should turn to the States, who may or may not choose to impose appropriate limits on economic development takings. Ante, at 19. This is an abdication of our responsibility. States play many important functions in our system of dual sovereignty, but compensating for our refusal to enforce properly the Federal Constitution (and a provision meant to curtail state action, no less) is not among them.
***
It was possible after Berman and Midkiff to imagine unconstitutional transfers from A to B. Those decisions endorsed government intervention when private property use had veered to such an extreme that the public was suffering as a consequence. Today nearly all real property is susceptible to condemnation on the Court’s theory. In the prescient words of a dissenter from the infamous decision in Poletown, “[n]ow that we have authorized local legislative bodies to decide that a different commercial or industrial use of property will produce greater public benefits than its present use, no homeowner’s, merchant’s or manufacturer’s property, however productive or valuable to its owner, is immune from condemnation for the benefit of other private interests that will put it to a ‘higher’ use.” 410 Mich., at 644–645, 304 N. W. 2d, at 464 (opinion of Fitzgerald, J.). This is why economic development takings “seriously jeopardiz[e] the security of all private property ownership.” Id., at 645, 304 N. W. 2d, at 465 (Ryan, J., dissenting).
Any property may now be taken for the benefit of another private party, but the fallout from this decision will not be random. The beneficiaries are likely to be those citizens with disproportionate influence and power in the political process, including large corporations and development firms. As for the victims, the government now has license to transfer property from those with fewer resources to those with more. The Founders cannot have intended this perverse result. “[T]hat alone is a just government,” wrote James Madison, “which impartially secures to every man, whatever is his own.” For the National Gazette, Property, (Mar. 29, 1792), reprinted in 14 Papers of James Madison 266 (R. Rutland et al. eds. 1983).
I would hold that the takings in both Parcel 3 and Parcel 4A are unconstitutional, reverse the judgment of the Supreme Court of Connecticut, and remand for further proceedings.
Justice O’Connor, dissenting.
For essentially the reasons given by Justice Souter, post this page (dissenting opinion), as well as the reasons given in my concurrence in McCreary County v. American Civil Liberties Union of Ky., post, p. 881, I respectfully dissent.
Justice O’Connor, concurring.
I join in the Court’s opinion. The First Amendment expresses our Nation’s fundamental commitment to religious liberty by means of two provisions—one protecting the free exercise of religion, the other barring establishment of religion. They were written by the descendents of people who had come to this land precisely so that they could practice their religion freely. Together with the other First Amendment guarantees—of free speech, a free press, and the rights to assemble and petition—the Religion Clauses were designed to safeguard the freedom of conscience and belief that those immigrants had sought. They embody an idea that was once considered radical: Free people are entitled to free and diverse thoughts, which government ought neither to constrain nor to direct.
Reasonable minds can disagree about how to apply the Religion Clauses in a given case. But the goal of the Clauses is clear: to carry out the Founders’ plan of preserving religious liberty to the fullest extent possible in a pluralistic society. By enforcing the Clauses, we have kept religion a matter for the individual conscience, not for the prosecutor or bureaucrat. At a time when we see around the world the violent consequences of the assumption of religious authority by government, Americans may count themselves fortunate: Our regard for constitutional boundaries has protected us from similar travails, while allowing private religious exercise to flourish. The well-known statement that “[w]e are a religious people,” Zorach v. Clauson, 343 U. S. 306, 313 (1952), has proved true. Americans attend their places of worship more often than do citizens of other developed nations, R. Fowler, A. Hertzke, & L. Olson, Religion and Politics in America 28–29 (2d ed. 1999), and describe religion as playing an especially important role in their lives, Pew Global Attitudes Project, Among Wealthy Nations U. S. Stands Alone in its Embrace of Religion (Dec. 19, 2002). Those who would renegotiate the boundaries between church and state must therefore answer a difficult question: Why would we trade a system that has served us so well for one that has served others so poorly?
Our guiding principle has been James Madison’s—that “[t]he Religion … of every man must be left to the conviction and conscience of every man.” Memorial and Remonstrance Against Religious Assessments, 2 Writings of James Madison 183, 184 (G. Hunt ed. 1901) (hereinafter Memorial). To that end, we have held that the guarantees of religious freedom protect citizens from religious incursions by the States as well as by the Federal Government. Everson v. Board of Ed. of Ewing, 330 U. S. 1, 16 (1947); Cantwell v. Connecticut, 310 U. S. 296 (1940). Government may not coerce a person into worshiping against her will, nor prohibit her from worshiping according to it. It may not prefer one religion over another or promote religion over nonbelief. Everson, supra, at 15–16. It may not entangle itself with religion. Walz v. Tax Comm’n of City of New York, 397 U. S. 664, 674 (1970). And government may not, by “endorsing religion or a religious practice,” “mak[e] adherence to religion relevant to a person’s standing in the political community.” Wallace v. Jaffree, 472 U. S. 38, 69 (1985) (O’Connor, J., concurring in judgment).
When we enforce these restrictions, we do so for the same reason that guided the Framers—respect for religion’s special role in society. Our Founders conceived of a Republic receptive to voluntary religious expression, and provided for the possibility of judicial intervention when government action threatens or impedes such expression. Voluntary religious belief and expression may be as threatened when government takes the mantle of religion upon itself as when government directly interferes with private religious practices. When the government associates one set of religious beliefs with the state and identifies nonadherents as outsiders, it encroaches upon the individual’s decision about whether and how to worship. In the marketplace of ideas, the government has vast resources and special status. Government religious expression therefore risks crowding out private observance and distorting the natural interplay between competing beliefs. Allowing government to be a potential mouthpiece for competing religious ideas risks the sort of division that might easily spill over into suppression of rival beliefs. Tying secular and religious authority together poses risks to both.
Given the history of this particular display of the Ten Commandments, the Court correctly finds an Establishment Clause violation. See ante, at 19–25. The purpose behind the counties’ display is relevant because it conveys an unmistakable message of endorsement to the reasonable observer. See Lynch v. Donnelly, 465 U. S. 668, 690 (1984) (O’Connor, J., concurring).
It is true that many Americans find the Commandments in accord with their personal beliefs. But we do not count heads before enforcing the First Amendment. See West Virginia Bd. of Ed. v. Barnette, 319 U. S. 624, 638 (1943) (“The very purpose of a Bill of Rights was to withdraw certain subjects from the vicissitudes of political controversy, to place them beyond the reach of majorities and officials and to establish them as legal principles to be applied by the courts”). Nor can we accept the theory that Americans who do not accept the Commandments’ validity are outside the First Amendment’s protections. There is no list of approved and disapproved beliefs appended to the First Amendment—and the Amendment’s broad terms (“free exercise,” “establishment,” “religion”) do not admit of such a cramped reading. It is true that the Framers lived at a time when our national religious diversity was neither as robust nor as well recognized as it is now. They may not have foreseen the variety of religions for which this Nation would eventually provide a home. They surely could not have predicted new religions, some of them born in this country. But they did know that line-drawing between religions is an enterprise that, once begun, has no logical stopping point. They worried that “the same authority which can establish Christianity, in exclusion of all other Religions, may establish with the same ease any particular sect of Christians, in exclusion of all other Sects.” Memorial 186. The Religion Clauses, as a result, protect adherents of all religions, as well as those who believe in no religion at all.
***
We owe our First Amendment to a generation with a profound commitment to religion and a profound commitment to religious liberty—visionaries who held their faith “with enough confidence to believe that what should be rendered to God does not need to be decided and collected by Caesar.” Zorach, supra, at 324–325 (Jackson, J., dissenting). In my opinion, the display at issue was an establishment of religion in violation of our Constitution. For the reasons given above, I join in the Court’s opinion.
Justice O’Connor delivered the opinion of the Court.
The Individuals with Disabilities Education Act (IDEA or Act), 84 Stat. 175, as amended, 20 U. S. C. A. §1400 et seq. (main ed. and Supp. 2005), is a Spending Clause statute that seeks to ensure that “all children with disabilities have available to them a free appropriate public education,” §1400(d)(1)(A). Under IDEA, school districts must create an “individualized education program” (IEP) for each disabled child. §1414(d). If parents believe their child’s IEP is inappropriate, they may request an “impartial due process hearing.” §1415(f). The Act is silent, however, as to which party bears the burden of persuasion at such a hearing. We hold that the burden lies, as it typically does, on the party seeking relief.
I
A
Congress first passed IDEA as part of the Education of the Handicapped Act in 1970, 84 Stat. 175, and amended it substantially in the Education for All Handicapped Children Act of 1975, 89 Stat. 773. At the time the majority of disabled children in America were “either totally excluded from schools or sitting idly in regular classrooms awaiting the time when they were old enough to ‘drop out,’ ” H. R. Rep. No. 94–332, p. 2 (1975). IDEA was intended to reverse this history of neglect. As of 2003, the Act governed the provision of special education services to nearly 7 million children across the country. See Dept. of Education, Office of Special Education Programs, Data Analysis System, http://www.ideadata.org/tables27th/ar_ aa9.htm (as visited Nov. 9, 2005, and available in Clerk of Court’s case file).
IDEA is “frequently described as a model of ‘cooperative federalism.’ ” Little Rock School Dist. v. Mauney, 183 F. 3d 816, 830 (CA8 1999). It “leaves to the States the primary responsibility for developing and executing educational programs for handicapped children, [but] imposes significant requirements to be followed in the discharge of that responsibility.” Board of Ed. of Hendrick Hudson Central School Dist., Westchester Cty. v. Rowley, 458 U. S. 176, 183 (1982). For example, the Act mandates cooperation and reporting between state and federal educational authorities. Participating States must certify to the Secretary of Education that they have “policies and procedures” that will effectively meet the Act’s conditions. 20 U. S. C. §1412(a). (Unless otherwise noted, all citations to the Act are to the pre-2004 version of the statute because this is the version that was in effect during the proceedings below. We note, however, that nothing in the recent 2004 amendments, 118 Stat. 2674, appears to materially affect the rule announced here.) State educational agencies, in turn, must ensure that local schools and teachers are meeting the State’s educational standards. 20 U. S. C. §§1412(a)(11), 1412(a)(15)(A). Local educational agencies (school boards or other administrative bodies) can receive IDEA funds only if they certify to a state educational agency that they are acting in accordance with the State’s policies and procedures. §1413(a)(1).
The core of the statute, however, is the cooperative process that it establishes between parents and schools. Rowley, supra, at 205–206 (“Congress placed every bit as much emphasis upon compliance with procedures giving parents and guardians a large measure of participation at every stage of the administrative process,… as it did upon the measurement of the resulting IEP against a substantive standard”). The central vehicle for this collaboration is the IEP process. State educational authorities must identify and evaluate disabled children, §§1414(a)–(c), develop an IEP for each one, §1414(d)(2), and review every IEP at least once a year, §1414(d)(4). Each IEP must include an assessment of the child’s current educational performance, must articulate measurable educational goals, and must specify the nature of the special services that the school will provide. §1414(d)(1)(A).
Parents and guardians play a significant role in the IEP process. They must be informed about and consent to evaluations of their child under the Act. §1414(c)(3). Parents are included as members of “IEP teams.” §1414(d)(1)(B). They have the right to examine any records relating to their child, and to obtain an “independent educational evaluation of the[ir] child.” §1415(b)(1). They must be given written prior notice of any changes in an IEP, §1415(b)(3), and be notified in writing of the procedural safeguards available to them under the Act, §1415(d)(1). If parents believe that an IEP is not appropriate, they may seek an administrative “impartial due process hearing.” §1415(f). School districts may also seek such hearings, as Congress clarified in the 2004 amendments. See S. Rep. No. 108–185, p. 37 (2003). They may do so, for example, if they wish to change an existing IEP but the parents do not consent, or if parents refuse to allow their child to be evaluated. As a practical matter, it appears that most hearing requests come from parents rather than schools. Brief for Petitioners 7.
Although state authorities have limited discretion to determine who conducts the hearings, §1415(f)(1)), and responsibility generally for establishing fair hearing procedures, §1415(a), Congress has chosen to legislate the central components of due process hearings. It has imposed minimal pleading standards, requiring parties to file complaints setting forth “a description of the nature of the problem,” §1415(b)(7)(B)(ii), and “a proposed resolution of the problem to the extent known and available… at the time,” §1415(b)(7)(B)(iii). At the hearing, all parties may be accompanied by counsel, and may “present evidence and confront, cross-examine, and compel the attendance of witnesses.” §§1415(h)(1)–(2). After the hearing, any aggrieved party may bring a civil action in state or federal court. §1415(i)(2). Prevailing parents may also recover attorney’s fees. §1415(i)(3)(B). Congress has never explicitly stated, however, which party should bear the burden of proof at IDEA hearings.
B
This case concerns the educational services that were due, under IDEA, to petitioner Brian Schaffer. Brian suffers from learning disabilities and speech-language impairments. From prekindergarten through seventh grade he attended a private school and struggled academically. In 1997, school officials informed Brian’s mother that he needed a school that could better accommodate his needs. Brian’s parents contacted respondent Montgomery County Public Schools System (MCPS) seeking a placement for him for the following school year.
MCPS evaluated Brian and convened an IEP team. The committee generated an initial IEP offering Brian a place in either of two MCPS middle schools. Brian’s parents were not satisfied with the arrangement, believing that Brian needed smaller classes and more intensive services. The Schaffers thus enrolled Brian in another private school, and initiated a due process hearing challenging the IEP and seeking compensation for the cost of Brian’s subsequent private education.
In Maryland, IEP hearings are conducted by administrative law judges (ALJs). See Md. Educ. Code Ann. §8–413(c) (Lexis 2004). After a 3-day hearing, the ALJ deemed the evidence close, held that the parents bore the burden of persuasion, and ruled in favor of the school district. The parents brought a civil action challenging the result. The United States District Court for the District of Maryland reversed and remanded, after concluding that the burden of persuasion is on the school district. Brian S. v. Vance, 86 F. Supp. 2d 538 (2000). Around the same time, MCPS offered Brian a placement in a high school with a special learning center. Brian’s parents accepted, and Brian was educated in that program until he graduated from high school. The suit remained alive, however, because the parents sought compensation for the private school tuition and related expenses.
Respondents appealed to the United States Court of Appeals for the Fourth Circuit. While the appeal was pending, the ALJ reconsidered the case, deemed the evidence truly in “equipoise,” and ruled in favor of the parents. The Fourth Circuit vacated and remanded the appeal so that it could consider the burden of proof issue along with the merits on a later appeal. The District Court reaffirmed its ruling that the school district has the burden of proof. 240 F. Supp. 2d 396 (Md. 2002). On appeal, a divided panel of the Fourth Circuit reversed. Judge Michael, writing for the majority, concluded that petitioners offered no persuasive reason to “depart from the normal rule of allocating the burden to the party seeking relief.” 377 F. 3d 449, 453 (2004). We granted certiorari, 543 U. S. 1145 (2005), to resolve the following question: At an administrative hearing assessing the appropriateness of an IEP, which party bears the burden of persuasion?
II
A
The term “burden of proof” is one of the “slipperiest member[s] of the family of legal terms.” 2 J. Strong, McCormick on Evidence §342, p. 433 (5th ed. 1999) (hereinafter McCormick). Part of the confusion surrounding the term arises from the fact that historically, the concept encompassed two distinct burdens: the “burden of persuasion,” i.e., which party loses if the evidence is closely balanced, and the “burden of production,” i.e., which party bears the obligation to come forward with the evidence at different points in the proceeding. Director, Office of Workers’ Compensation Programs v. Greenwich Collieries, 512 U. S. 267, 272 (1994). We note at the outset that this case concerns only the burden of persuasion, as the parties agree, Brief for Respondents 14; Reply Brief for Petitioners 15, and when we speak of burden of proof in this opinion, it is this to which we refer.
When we are determining the burden of proof under a statutory cause of action, the touchstone of our inquiry is, of course, the statute. The plain text of IDEA is silent on the allocation of the burden of persuasion. We therefore begin with the ordinary default rule that plaintiffs bear the risk of failing to prove their claims. McCormick §337, at 412 (“The burdens of pleading and proof with regard to most facts have and should be assigned to the plaintiff who generally seeks to change the present state of affairs and who therefore naturally should be expected to bear the risk of failure or proof or persuasion”); C. Mueller & L. Kirkpatrick, Evidence §3.1, p. 104 (3d ed. 2003) (“Perhaps the broadest and most accepted idea is that the person who seeks court action should justify the request, which means that the plaintiffs bear the burdens on the elements in their claims”).
Thus, we have usually assumed without comment that plaintiffs bear the burden of persuasion regarding the essential aspects of their claims. For example, Title VII of the Civil Rights Act of 1964, 42 U. S. C. §2000e–2 et seq., does not directly state that plaintiffs bear the “ultimate” burden of persuasion, but we have so concluded. St. Mary’s Honor Center v. Hicks, 509 U. S. 502, 511 (1993); id., at 531 (Souter, J., dissenting). In numerous other areas, we have presumed or held that the default rule applies. See, e.g., Lujan v. Defenders of Wildlife, 504 U. S. 555, 561 (1992) (standing); Cleveland v. Policy Management Systems Corp., 526 U. S. 795, 806 (1999) (Americans with Disabilities Act); Hunt v. Cromartie, 526 U. S. 541, 553 (1999) (equal protection); Wharf (Holdings) Ltd. v. United Int’l Holdings, Inc., 532 U. S. 588, 593 (2001) (securities fraud); Doran v. Salem Inn, Inc., 422 U. S. 922, 931 (1975) (preliminary injunctions); Mt. Healthy City Bd. of Ed. v. Doyle, 429 U. S. 274, 287 (1977) (First Amendment). Congress also expressed its approval of the general rule when it chose to apply it to administrative proceedings under the Administrative Procedure Act, 5 U. S. C. §556(d); see also Greenwich Collieries, supra, at 271.
The ordinary default rule, of course, admits of exceptions. See McCormick §337, at 412–415. For example, the burden of persuasion as to certain elements of a plantiff’s claim may be shifted to defendants, when such elements can fairly be characterized as affirmative defenses or exemptions. See, e.g., FTC v. Morton Salt Co., 334 U. S. 37, 44–45 (1948). Under some circumstances this Court has even placed the burden of persuasion over an entire claim on the defendant. See Alaska Dept. of Environmental Conservation v. EPA, 540 U. S. 461, 494 (2004). But while the normal default rule does not solve all cases, it certainly solves most of them. Decisions that place the entire burden of persuasion on the opposing party at the outset of a proceeding—as petitioners urge us to do here—are extremely rare. Absent some reason to believe that Congress intended otherwise, therefore, we will conclude that the burden of persuasion lies where it usually falls, upon the party seeking relief.
B
Petitioners contend first that a close reading of IDEA’s text compels a conclusion in their favor. They urge that we should interpret the statutory words “due process” in light of their constitutional meaning, and apply the balancing test established by Mathews v. Eldridge, 424 U. S. 319 (1976). Even assuming that the Act incorporates constitutional due process doctrine, Eldridge is no help to petitioners, because “[o]utside the criminal law area, where special concerns attend, the locus of the burden of persuasion is normally not an issue of federal constitutional moment.” Lavine v. Milne, 424 U. S. 577, 585 (1976).
Petitioners next contend that we should take instruction from the lower court opinions of Mills v. Board of Education, 348 F. Supp. 866 (D. C. 1972), and Pennsylvania Association for Retarded Children v. Commonwealth, 334 F. Supp. 1257 (ED Pa. 1971) (hereinafter PARC ). IDEA’s drafters were admittedly guided “to a significant extent” by these two landmark cases. Rowley, 458 U. S., at 194. As the court below noted, however, the fact that Congress “took a number of the procedural safeguards from PARC and Mills and wrote them directly into the Act” does not allow us to “conclude… that Congress intended to adopt the ideas that it failed to write into the text of the statute.” 377 F. 3d, at 455.
Petitioners also urge that putting the burden of persuasion on school districts will further IDEA’s purposes because it will help ensure that children receive a free appropriate public education. In truth, however, very few cases will be in evidentiary equipoise. Assigning the burden of persuasion to school districts might encourage schools to put more resources into preparing IEPs and presenting their evidence. But IDEA is silent about whether marginal dollars should be allocated to litigation and administrative expenditures or to educational services. Moreover, there is reason to believe that a great deal is already spent on the administration of the Act. Litigating a due process complaint is an expensive affair, costing schools approximately $8,000-to-$12,000 per hearing. See Department of Education, J. Chambers, J. Harr, & A. Dhanani, What Are We Spending on Procedural Safeguards in Special Education 1999–2000, p. 8 (May 2003) (prepared under contract by American Institute for Research, Special Education Expenditure Project). Congress has also repeatedly amended the Act in order to reduce its administrative and litigation-related costs. For example, in 1997 Congress mandated that States offer mediation for IDEA disputes. Individuals with Disabilities Education Act Amendments of 1997, Pub. L. 105–17, §615(e), 111 Stat. 90, 20 U. S. C. §1415(e). In 2004, Congress added a mandatory “resolution session” prior to any due process hearing. Individuals with Disabilities Education Improvement Act of 2004, Pub. L. 108–446, §615(7)(f)(1)(B), 118 Stat. 2720, 20 U. S. C. A. §1415(f)(1)(B) (Supp. 2005). It also made new findings that “[p]arents and schools should be given expanded opportunities to resolve their disagreements in positive and constructive ways,” and that “[t]eachers, schools, local educational agencies, and States should be relieved of irrelevant and unnecessary paperwork burdens that do not lead to improved educational outcomes.” §§1400(c)(8)–(9).
Petitioners in effect ask this Court to assume that every IEP is invalid until the school district demonstrates that it is not. The Act does not support this conclusion. IDEA relies heavily upon the expertise of school districts to meet its goals. It also includes a so-called “stay-put” provision, which requires a child to remain in his or her “then-current educational placement” during the pendency of an IDEA hearing. §1415(j). Congress could have required that a child be given the educational placement that a parent requested during a dispute, but it did no such thing. Congress appears to have presumed instead that, if the Act’s procedural requirements are respected, parents will prevail when they have legitimate grievances. See Rowley, supra, at 206 (noting the “legislative conviction that adequate compliance with the procedures prescribed would in most cases assure much if not all of what Congress wished in the way of substantive content in an IEP”).
Petitioners’ most plausible argument is that “[t]he ordinary rule, based on considerations of fairness, does not place the burden upon a litigant of establishing facts peculiarly within the knowledge of his adversary.” United States v. New York, N. H. & H. R. Co., 355 U. S. 253, 256, n. 5 (1957); see also Concrete Pipe & Products of Cal., Inc. v. Construction Laborers Pension Trust for Southern Cal., 508 U. S. 602, 626 (1993). But this “rule is far from being universal, and has many qualifications upon its application.” Greenleaf’s Lessee v. Birth, 6 Pet. 302, 312 (1832); see also McCormick §337, at 413 (“Very often one must plead and prove matters as to which his adversary has superior access to the proof”). School districts have a “natural advantage” in information and expertise, but Congress addressed this when it obliged schools to safeguard the procedural rights of parents and to share information with them. See School Comm. of Burlington v. Department of Ed. of Mass., 471 U. S. 359, 368 (1985). As noted above, parents have the right to review all records that the school possesses in relation to their child. §1415(b)(1). They also have the right to an “independent educational evaluation of the[ir] child.” Ibid. The regulations clarify this entitlement by providing that a “parent has the right to an independent educational evaluation at public expense if the parent disagrees with an evaluation obtained by the public agency.” 34 CFR §300.502(b)(1) (2005). IDEA thus ensures parents access to an expert who can evaluate all the materials that the school must make available, and who can give an independent opinion. They are not left to challenge the government without a realistic opportunity to access the necessary evidence, or without an expert with the firepower to match the opposition.
Additionally, in 2004, Congress added provisions requiring school districts to answer the subject matter of a complaint in writing, and to provide parents with the reasoning behind the disputed action, details about the other options considered and rejected by the IEP team, and a description of all evaluations, reports, and other factors that the school used in coming to its decision. Pub. L. 108–446, §615(c)(2)(B)(i)(I), 118 Stat. 2718, 20 U. S. C. A. §1415(c)(2)(B)(i)(I) (Supp. 2005). Prior to a hearing, the parties must disclose evaluations and recommendations that they intend to rely upon. 20 U. S. C. §1415(f)(2). IDEA hearings are deliberately informal and intended to give ALJs the flexibility that they need to ensure that each side can fairly present its evidence. IDEA, in fact, requires state authorities to organize hearings in a way that guarantees parents and children the procedural protections of the Act. See §1415(a). Finally, and perhaps most importantly, parents may recover attorney’s fees if they prevail. §1415(i)(3)(B). These protections ensure that the school bears no unique informational advantage.
III
Finally, respondents and several States urge us to decide that States may, if they wish, override the default rule and put the burden always on the school district. Several States have laws or regulations purporting to do so, at least under some circumstances. See, e.g., Minn. Stat. §125A.091, subd. 16 (2004); Ala. Admin. Code Rule 290–8–9–.08(8)(c)(6) (Supp. 2004); Alaska Admin. Code tit. 4, §52.550(e)(9) (2003); Del. Code Ann., Tit. 14, §3140 (1999). Because no such law or regulation exists in Maryland, we need not decide this issue today. Justice Breyer contends that the allocation of the burden ought to be left entirely up to the States. But neither party made this argument before this Court or the courts below. We therefore decline to address it.
We hold no more than we must to resolve the case at hand: The burden of proof in an administrative hearing challenging an IEP is properly placed upon the party seeking relief. In this case, that party is Brian, as represented by his parents. But the rule applies with equal effect to school districts: If they seek to challenge an IEP, they will in turn bear the burden of persuasion before an ALJ. The judgment of the United States Court of Appeals for the Fourth Circuit is, therefore, affirmed.
It is so ordered.
Justice O’Connor delivered the opinion of the Court.
We consider whether the United States may offset Social Security benefits to collect a student loan debt that has been outstanding for over 10 years.
I
A
Petitioner James Lockhart failed to repay federally reinsured student loans that he had incurred between 1984 and 1989 under the Guaranteed Student Loan Program. These loans were eventually reassigned to the Department of Education, which certified the debt to the Department of the Treasury through the Treasury Offset Program. In 2002, the Government began withholding a portion of petitioner’s Social Security payments to offset his debt, some of which was more than 10 years delinquent.
Petitioner sued in Federal District Court, alleging that under the Debt Collection Act’s 10-year statute of limitations, the offset was time barred. The District Court dismissed the complaint, and the Court of Appeals for the Ninth Circuit affirmed. 376 F. 3d 1027 (2004). We granted certiorari, 544 U. S. ___ (2005), to resolve the conflict between the Ninth Circuit and the Eighth Circuit, see Lee v. Paige, 376 F. 3d 1179 (CA8 2004), and now affirm.
B
The Debt Collection Act of 1982, as amended, provides that, after pursuing the debt collection channels set out in 31 U. S. C. §3711(a), an agency head can collect an outstanding debt “by administrative offset.” §3716(a). The availability of offsets against Social Security benefits is limited, as the Social Security Act, 49 Stat. 620, as amended, makes Social Security benefits, in general, not “subject to execution, levy, attachment, garnishment, or other legal process.” 42 U. S. C. §407(a). The Social Security Act purports to protect this anti-attachment rule with an express-reference provision: “No other provision of law, enacted before, on, or after April 20, 1983, may be construed to limit, supersede, or otherwise modify the provisions of this section except to the extent that it does so by express reference to this section.” §407(b).
Moreover, the Debt Collection Act’s offset provisions generally do not authorize the collection of claims which, like petitioner’s debts at issue here, are over 10 years old. 31 U. S. C. §3716(e)(1). In 1991, however, the Higher Education Technical Amendments, 105 Stat. 123, sweepingly eliminated time limitations as to certain loans: “Notwithstanding any other provision of statute… no limitation shall terminate the period within which suit may be filed, a judgment may be enforced, or an offset, garnishment, or other action initiated or taken,” 20 U. S. C. §1091a(a)(2), for the repayment of various student loans, including the loans at issue here, §1091a(a)(2)(D).
The Higher Education Technical Amendments, by their terms, did not make Social Security benefits subject to offset; these were still protected by the Social Security Act’s anti-attachment rule. Only in 1996 did the Debt Collection Improvement Act—in amending and recodifying the Debt Collection Act—provide that, “[n]otwithstanding any other provision of law (including [§407] … ),” with a limited exception not relevant here, “all payment due an individual under … the Social Security Act … shall be subject to offset under this section.” 31 U. S. C. §3716(c)(3)(A)(i).
II
The Government does not contend that the “notwithstanding” clauses in both the Higher Education Technical Amendments and the Debt Collection Improvement Act trump the Social Security Act’s express-reference provision. Cf. Marcello v. Bonds, 349 U. S. 302, 310 (1955) (“Exemptions from the terms of the … Act are not lightly to be presumed in view of the statement … that modifications must be express[.] But … [u]nless we are to require the Congress to employ magical passwords in order to effectuate an exemption from the … Act, we must hold that the present statute expressly supersedes the … provisions of that Act” (citation omitted)); Great Northern R. Co. v. United States, 208 U. S. 452, 465 (1908).
We need not decide the effect of express-reference provisions such as §407(b) to resolve this case. Because the Debt Collection Improvement Act clearly makes Social Security benefits subject to offset, it provides exactly the sort of express reference that the Social Security Act says is necessary to supersede the anti-attachment provision.
It is clear that the Higher Education Technical Amendments remove the 10-year limit that would otherwise bar offsetting petitioner’s Social Security benefits to pay off his student loan debt. Petitioner argues that Congress could not have intended in 1991 to repeal the Debt Collection Act’s statute of limitations as to offsets against Social Security benefits—since debt collection by Social Security offset was not authorized until five years later. Therefore, petitioner continues, the Higher Education Technical Amendments’ abrogation of time limits in 1991 only applies to then-valid means of debt collection. We disagree. “The fact that Congress may not have foreseen all of the consequences of a statutory enactment is not a sufficient reason for refusing to give effect to its plain meaning.” Union Bank v. Wolas, 502 U. S. 151, 158 (1991).
Petitioner points out that the Higher Education Technical Amendments, unlike the Debt Collection Improvement Act, do not explicitly mention §407. But §407(b) only requires an express reference to authorize attachment in the first place—which the Debt Collection Improvement Act has already provided.
III
Nor does the Debt Collection Improvement Act’s 1996 recodification of the Debt Collection Act help petitioner. The Debt Collection Improvement Act, in addition to adding offset authority against Social Security benefits, retained the Debt Collection Act’s general 10-year bar on offset authority. But the mere retention of this previously enacted time bar does not make the time bar apply in all contexts—a result that would extend far beyond Social Security benefits, since it would imply that the Higher Education Technical Amendments’ abrogation of time limits was now a dead letter as to any kind of administrative offset. Rather, the Higher Education Technical Amendments retain their effect as a limited exception to the Debt Collection Act time bar in the student loan context.
Finally, we decline to read any meaning into the failed 2004 effort to amend the Debt Collection Act to explicitly authorize offset of debts over 10 years old. See H. R. 5025, 108th Cong., 2d Sess., §642 (Sept. 8, 2004); S. 2806, 108th Cong., 2d Sess., §642 (Sept. 15, 2004). “[F]ailed legislative proposals are ‘a particularly dangerous ground on which to rest an interpretation of a prior statute.’ ” United States v. Craft, 535 U. S. 274, 287 (2002) (quoting Pension Benefit Guaranty Corporation v. LTV Corp., 496 U. S. 633, 650 (1990)). In any event, it is unclear what meaning we could read into this effort even if we were inclined to do so, as the failed amendment—which was not limited to offsets against Social Security benefits—would have had a different effect than the interpretation we advance today.
Therefore, we affirm the judgment of the Ninth Circuit.
It is so ordered.
We do not revisit our abortion precedents today, but rather address a question of remedy: If enforcing a statute that regulates access to abortion would be unconstitutional in medical emergencies, what is the appropriate judicial response? We hold that invalidating the statute entirely is not always necessary or justified, for lower courts may be able to render narrower declaratory and injunctive relief.
O’CONNOR, Associate Justice (Ret.).
The Bureau of Alcohol, Tobacco, Firearms, and Explosives appeals from the district court’s grant of summary judgment holding that the Bureau lacks authority to issue a letter requiring a small percentage of licensed firearms dealers to submit portions of their records relating to secondhand firearms. Because we find that the Bureau acted within its statutory authority under 18 U.S.C. § 923(g)(5)(A), we reverse the district court’s grant of summary judgment. We affirm, however, the district court’s determination that the Bureau did not act in an arbitrary and capricious fashion in deciding which dealers should receive the disputed letter.
I.
The Gun Control Act of 1968, 18 U.S.C. § 921 et seq., requires persons wishing to “engage in the business of importing, manufacturing, or dealing in firearms” to apply for and obtain a license from the Bureau of Alcohol, Tobacco, Firearms, and Explosives.1 18 U.S.C. § 923(a). Successful applicants, known as federal firearms licensees (“FFLs”), must create and maintain detailed records documenting the firearms transactions that they conduct. When FFL dealers receive a firearm they must record “the date of receipt, the name and address or the name and license number of the person from whom received, the name of the manufacturer and importer (if any), the model, serial number, type, and the caliber or gauge.” 27 C.F.R. § 478.125(e). After selling a firearm, FFL dealers must further record the name and address
O’CONNOR, Associate Justice (Ret.).
A jury convicted Raul Jiminez of conspiracy to distribute and to possess with intent to distribute more than 500 grams of methamphetamine. Jiminez challenges the proceedings below on two grounds. First, Jiminez contends that the district court improperly allowed certain evidence to be introduced which was unfairly prejudicial. Second, Jiminez contends that there was insufficient evidence to support his conviction. Because the district court did not abuse its discretion in the contested evidentiary rulings and a reasonable jury could have found Jiminez guilty beyond a reasonable doubt, we affirm the judgment of the district court.
I.
Using a federal wiretap to intercept phone calls, law enforcement officials learned of a drug shipment that was to be delivered to Norfolk, Nebraska. On the morning of March 3, 2005, a Nebraska State Patrol official informed Trooper Timothy Stopak of suspicions that a shipment of methamphetamine was being transported in a green Plymouth minivan, with the Minnesota license plate number NHB030, traveling north on Highway 81 toward Norfolk, Nebraska. Later that morning, Trooper Stopak saw the green minivan and pulled the vehicle over for speeding.
Trooper Stopak approached the vehicle and asked Jiminez, who was the only occupant of the minivan, for his driver’s license, registration, and proof of insurance. As Jiminez handed over some of the materials, Trooper Stopak observed that Jiminez’s hand was shaking. When
O’CONNOR, Associate Justice (Ret.).
This case considers the response of the Department of Labor’s Office of Federal Contract Compliance Programs (OFCCP) to Donald Greer’s complaint filed under the Vietnam Era Veterans’ Readjustment Assistance Act of 1974 (VEVRAA), as amended. Because the OFCCP promptly discharged its duty to conduct an investigation into Greer’s administrative complaint, we conclude that the Secretary of Labor’s response in this case represents a decision committed to agency discretion and is, therefore, immune from judicial review.
I.
The Vietnam Era Veterans Readjustment Assistance Act of 1974 (VEVRAA), as amended, provides that the federal government shall require its contractors to “take affirmative action to employ and advance in employment qualified covered veterans.” 38 U.S.C. § 4212(a). In addition, the statute provides that a covered veteran who believes that a government contractor has not complied with VEVRAA “may file a complaint with the Secretary of Labor, who shall promptly investigate such complaint and take appropriate action in accordance with the terms of the contract and applicable laws and regulations.” 38 U.S.C. § 4212(b). The Secretary of Labor has charged the Office of Federal Contract Compliance Programs (OFCCP) with investigating complaints made against contractors. 41 C.F.R. 60–250.60, 60–250.61(a). After the OFCCP receives such a complaint, it is directed to “prompt[ly] investigat[e],” 41 C.F.R. 60–250.61(d), and determine whether