concurring in the judgment.
The Court today continues on the course adopted in DelCostello v. Teamsters, 462 U. S. 151 (1983), and concludes that although Congress has enacted no federal limitations period for civil actions for damages brought under the Racketeer Influenced and Corrupt Organizations Act (RICO), 18 U. S. C. § 1964 (1982 ed. and Supp. III), it will supply one by “borrowing” the 4-year statute of limitations applicable to suits under the Clayton Act. 15 U. S. C. § 15b. While at first glance it may seem a small step from the familiar practice of borrowing state statutes of limitations to today’s decision to borrow a federal one, in my view it turns out to be a giant leap into the realm of legislative judgments. I therefore cannot join the Court’s opinion.
I
The issue presented by this case cannot arise with respect to federal criminal statutes, as every federal offense is governed by an express limitations period. If no statute specifically defines a limitations period (or prescribes the absence of a limitations period, see 18 U. S. C. § 3281) for a particular offense, a “catchall” statute operates to forbid prosecution, trial, or punishment “unless the indictment is found or the information is instituted within five years next after such offense shall have been committed.” § 3282. Congress has not provided that kind of a default limitations period, however, for federal civil suits; and since it has long been enacting civil statutes without express limitations periods, courts have long been wrestling with the problem of determining what, if any, limitations periods to apply. Prior to DelCostello, the virtually uniform practice was to look to applicable state statutes of limitations. Indeed, we departed *158from that practice only when the applicable state limitations period would have frustrated the policy of the federal statute, concluding that in such a case no limitations period governs the suit. See Occidental Life Ins. Co. of Cal. v. EEOC, 432 U. S. 355, 361, 366-372 (1977). Until DelCostello, we never responded to legislative silence by applying a limitations period drawn from a different federal statute.
To understand why this new practice differs from — and is less legitimate than — the practice of borrowing state statutes, it is necessary to understand the two-phase history through which the earlier practice developed. It is in turn essential to that understanding to recognize that certain common conceptions about the borrowing of state limitations statutes are mistaken. As Part I-A explains in more detail, the very label used to describe that practice — “borrowing”— is misleading. In its original form, during what I term the “first phase” of the borrowing doctrine, our practice of applying state law in reality involved no borrowing at all; rather, we applied state limitations periods to federal causes of action because we believed that those state statutes applied of their own force, unless pre-empted by federal law. In the “second phase” of our development of the borrowing doctrine, we approached the issue rather differently. Whereas we had originally focused on the federal statute creating the cause of action only for purposes of our pre-emption inquiry— i. e., in order to ascertain whether the otherwise applicable state statute of limitations conflicted with the federal statute’s terms or purposes — we later came to believe that the federal statute itself was the source of our obligation to apply state law. That is, instead of treating Congress’ silence on the limitations question as a failure to pre-empt state law, we came to treat it as an affirmative directive to borrow state law. In my view, that deviation from the “first phase” approach was an analytical error. It has led in turn to the further error the Court commits (or compounds) today in deciding to treat congressional silence as a directive to borrow a *159limitations period from a different federal statute. Today’s error is by far the more serious of the two. As the history outlined above (and discussed in detail below) suggests, the borrowing of state statutes on the erroneous ground of congressional intent has a basis in, and to a reasonable degree approximates the results of, the approach that I think is correct as an original matter. The same cannot be said for the borrowing of federal statutes.
A
The analysis representing the “first phase” of the borrowing doctrine is exemplified by McCluny v. Silliman, 3 Pet. 270 (1830), the first case presenting the question of what limitations period, if any, applies to a claim having its source in federal law when federal law does not specify the applicable time limit. Plaintiff-in-error McCluny had sought to purchase land under a federal statute providing for the sale of lands owned by the United States, but the register, a federal officer, had refused his tendered payment. McCluny then brought an action for trespass on the case in the Circuit Court for the District of Ohio, arguing that the register had wrongfully withheld the land, causing him $50,000 in damages. The register prevailed below on the ground that the Ohio statute of limitations governing actions for trespass on the case barred plaintiff’s suit. McCluny argued that the Circuit Court had erred. The Ohio limitations statute, he contended, had no application in a suit brought in federal court against a federal officer for violation of a right conferred by an Act of Congress, not because Congress did not intend so (an issue raised by neither party to the dispute) but because the Ohio Legislature did not intend so. Id., at 270-274. We agreed with McCluny that the issue was whether the statute applied as a matter of Ohio law, see id., at 276 (“The decision in this cause depends upon the construction of the statute of Ohio”), but agreed with the register that under Ohio law, the statute applied. We reasoned that while it *160was doubtless true that Ohio had not contemplated that its statute would govern such actions, by framing it to apply to all actions for trespass on the case the legislature had designed the statute to cover numerous torts not specifically within its contemplation. Id., at 277-278. At no point did we even question Ohio’s power to enact statutes of limitations applicable to federal rights, so long as Congress had not provided otherwise. Rather, we simply noted that it was “well settled” that such statutes are “the law of the forum, and operat[e] upon all who submit themselves to its jurisdiction.” Id., at 276-277. In the course of our opinion, we also mentioned the Rules of Decision Act, which provides:
“[T]he laws of the several states, except where the constitution, treaties, or statutes of the United States shall otherwise require or provide, shall be regarded as rules of decision in trials at common law in the courts of the United States in cases where they apply.” § 34, Judiciary Act of 1789, 1 Stat. 92, codified, as amended, at 28 U. S. C. § 1652.
But we discussed that statute not as the source of Ohio’s power, but as confirmation of it where “no special provision had been made by congress,” 3 Pet., at 277.
McCluny is an odd case to modern ears, because although a federal statute was clearly the source of McCluny’s claim of right, it did not expressly create his cause of action. Yet neither the parties nor the Court raised the question we would certainly ask today: whether the federal statute gave him an “implied right” to sue. Instead McCluny simply brought an action seeking a common-law writ of trespass on the case. That feature of the case leaves open the argument that our acceptance of Ohio’s power to pass limitations periods applicable to federal rights was based on the fact that the cause of action itself came from the common law rather than a federal statute.
That argument, however, was rejected in Campbell v. Haverhill, 155 U. S. 610 (1895), where we were faced with the *161question whether to apply to a suit for patent infringement a Massachusetts statute of limitations requiring actions for tort to be brought within six years. In patent infringement suits, both the right and the cause of action were created by congressional legislation, and the federal courts had exclusive jurisdiction. Accordingly, it was argued that “the States, having no power to create the right or enforce the remedy, have no power to limit such remedy or to legislate in any manner with respect to the subject matter.” Id., at 615. We replied that “this is rather to assert a distinction than to point out a difference,” ibid., and that in the absence of congressional provision to the contrary, the States had the power to pass statutes of limitations that apply neutrally to federal rights, id., at 614-615, 618-620 (although they might not have the power to enact statutes that discriminated against federal rights or provided excessively short time periods for bringing suit, id., at 614-615).1
B
These early cases provide the foundation for a reasonably coherent theory about the application of state statutes of limitations to federal statutory causes of action. First, state statutes of limitations whose terms appear to cover federal statutory causes of action apply as a matter of state law to such claims, even though the state legislature that enacted the statutes did not have those claims in mind. McCluny, supra, at 277-278. Second, imposition of limitations periods on federal causes of action is within the States’ powers, if not pre-empted by Congress. Campbell v. Haverhill, supra, at *162614-615, 618-620; McCluny, 3 Pet., at 276-277. Third, the obligation to apply state statutes of limitations does not spring from Congress’ intent in enacting the federal statute; rather, that intent is relevant only to the question whether the state limitations period had been pre-empted by Congress’ failure to provide one. Campbell v. Haverhill, supra, at 616. Fourth, congressional silence on the limitations issue is ordinarily insufficient to pre-empt state statutes; “special provision” by Congress is required to do that. Ibid,.; McCluny, supra, at 277. Fifth, the federal statute — its substantive provisions rather than its mere silence — may be sufficient to pre-empt a state statute that discriminates against federal rights or is too short to permit the federal right to be vindicated. Campbell v. Haverhill, supra, at 614-615.
As to the role of the Rules of Decision Act: Although Campbell v. Haverhill in particular is not clear on the question, the Rules of Decision Act plays no role in deriving the first two principles stated above. It directs federal courts to follow state laws only “in cases where they apply,” which federal courts would be required to do even in the absence of the Act. That is clear not only from the borrowing cases, but also from other early opinions of this Court displaying the clear understanding that the Act did not make state laws applicable to any new classes of cases. See Hawkins v. Barney’s Lessee, 5 Pet. 457, 464 (1831) (the Rules of Decision Act “has been uniformly held to be no more than a declaration of what the law would have been without it: to wit, that the lex loci must be the governing rule of private right, under whatever jurisdiction private right comes to be examined”); Bank of Hamilton v. Dudley’s Lessee, 2 Pet. 492, 525-526 (1829) (“The laws of the states . . . would be . . . regarded [as rules of decision in the courts of the United States] independent of that special enactment”); Hill, The Erie Doctrine in Bankruptcy, 66 Harv. L. Rev. 1013, 1026, 1035 (1953); see also Jackson v. Chew, 12 Wheat. 153, 162 (1827) (holding that the Supreme Court would follow rules of property law settled *163by state-court decisions without mentioning the Rules of Decision Act); Shelby v. Guy, 11 Wheat. 361, 367 (1826) (holding that the Court was required to follow state statutes and their construction by state courts because of its duty to administer the laws of the respective States, without mentioning the Rules of Decision Act). In fact, because the Act required application of future state laws as well as those in effect at the time of its passage, it would have been considered open to serious constitutional challenge as an improper delegation of congressional legislative power to the States had it been anything other than declaratory on that point. See Wayman v. Southard, 10 Wheat. 1, 47-48 (1825).
Thus, the Act changes the analysis of the question whether a federal court should look to state law only insofar as it provides the basis for the fourth principle. Its directive to federal courts to apply state law unless federal law otherwise “requires or provides” creates a presumption against implicit pre-emption which must be rebutted by affirmative congressional action, except for the implicit preclusion of state statutes that discriminate against federal claims or provide too short a limitations period to permit vindication of the federal right.
II
So understood, the borrowing doctrine involves no borrowing at all. Instead, it only requires us to engage in two everyday interpretive exercises: the determination of which state statute of limitations applies to a federal claim as a matter of state law, and the determination of whether the federal statute creating the cause of action pre-empts that state limitations period. We need not embark on a quest for an “appropriate” statute of limitations except to the limited extent that making those determinations may entail judgments as to which statute the State would believe “appropriate” and as to whether federal policy nevertheless makes that statute “inappropriate.” Finally, if we determine that the state limitations period that would apply under state law is pre-empted *164because it is inconsistent with the federal statute, that is the end of the matter, and there is no limitation on the federal cause of action.
In my view, that is the best approach to the question before us, and if a different historical practice had not intervened I would adhere to it. See also DelCostello v. Teamsters, 462 U. S., at 172-174 (Stevens, J., dissenting). For many years, however, we have used a different analysis. In the second phase of development of the borrowing doctrine, perhaps forgetting its origins, the Court adopted the view that we borrow the “appropriate” state statute of limitations when Congress fails to provide one because that is Congress’ directive, implied by its silence on the subject. See Automobile Workers v. Hoosier Cardinal Corp., 383 U. S. 696, 706 (1966); Holmberg v. Armbrecht, 327 U. S. 392, 395 (1946).2 As an original matter, that is not a very plausible interpretation of congressional silence. If one did not believe that state limitations periods applied of their own force, the most natural intention to impute to a Congress that enacted no limitations period would be that it wished none. However, after a century and a half of the Court’s reacting to congressional silence by applying state statutes — first for the right *165reason, then for the wrong one — by now at least it is reasonable to say that such a result is what Congress must expect, and hence intend, by its silence. The approach therefore has some legitimacy, and in any event generally produces the same results as the one I believe to be correct.3
Ill
As Justice O’Connor pointed out in her dissent in Del-Costello, however, if we are serious about this “congressional intent” justification for the borrowing doctrine, we should at least require some evidence of actual alteration of that intent before departing from it. See 462 U. S., at 174-175 (O’Connor, J., dissenting). For if the basis of the rule is, in some form, that Congress knows that we will borrow state statutes of limitations unless it directs otherwise, it also knows that it *166has to direct otherwise if it wants us to do something else. In addition, as under our former approach, should we discover that there is no appropriate state statute to borrow, because all the available ones run afoul of federal policy, we ought to conclude that there is no limitations period.
In the case before us, however, the Court does not require any showing of actual congressional intent at all before departing from our practice of borrowing state statutes, prowling hungrily through the Statutes at Large for an appetizing federal limitations period, and pouncing on the Clayton Act. Of course, a showing of actual congressional intent that we depart from tradition and borrow a federal statute is quite impossible. Under ordinary principles of construction, the very identity between the language and structure of the Clayton Act’s and RICO’s private civil-remedy provisions relied on by the Court as arguments for borrowing 15 U. S. C. § 15b, would, when coupled with Congress’ enactment of a limitations period for the former and failure to enact one for the latter, demonstrate — if any intent to depart from the state borrowing rule — a desire for no limitations period at all. The same is suggested by the legislative history discussed by the Court, showing that Congress has passed up several opportunities to impose a federal limitations period on civil RICO claims, ante, at 154-155. The Court avoids the troublesome requirement of finding a congressional intent to depart from state borrowing by the simple expedient of reformulating the rule, transforming it from a presumption that congressional silence means that we should apply the appropriate state limitations period into a presumption that congressional silence means we should apply the appropriate limitations period, state or federal. I cannot go along with this, for two reasons.
First, I can find no legitimate source for the new rule. Whereas our prior practice provides some basis for arguing that when Congress creates a civil cause of action without a limitations period, it expects and intends application of an *167appropriate state statute, there is no basis whatsoever for arguing that its silence signifies that the most appropriate statute, state or federal, should be borrowed. To the contrary, all available evidence indicates that when Congress intends a federal limitations period for a civil cause of action, it enacts one — for example, 15 U. S. C. § 15b itself. The possibility of borrowing a federal statute of limitations did not occur to any of the parties in this litigation until it was suggested by a concurring judge in the Court of Appeals, see 792 F. 2d 341, 356 (CA3 1986), and all of the Federal Courts of Appeals that have passed on the issue of the appropriate RICO limitations period have applied state statutes. See 792 F. 2d 341 (1986) (case below); Cullen v. Margiotta, 811 F. 2d 698 (CA2 1987); La Porte Construction Co. v. Bayshore National Bank, 805 F. 2d 1254 (CA5 1986); Silverberg v. Thomson McKinnon Securities, Inc., 787 F. 2d 1079 (CA6 1986); Tellis v. United States Fidelity & Guaranty Co., 805 F. 2d 741 (CA7 1986); Alexander v. Perkin Elmer Corp., 729 F. 2d 576 (CA8 1984); Compton v. Ide, 732 F. 2d 1429 (CA9 1984); Hunt v. American Bank & Trust Co., 783 F. 2d 1011 (CA11 1986). It is extremely unlikely that Congress expected anything different. Moreover, had our prior rule been that a federal statute should be borrowed if appropriate, the considerations the Court advances as to why that is the right course here — that it will promote uniformity and avoid litigation, and that there are differences between the federal action and the actions covered by state statutes — would have been sufficient to warrant selection of a federal limitations period for almost any federal statute, a conclusion plainly inconsistent with the results of our cases.4
*168Second, as the case before us demonstrates, the new rule involves us in a very different kind of enterprise from that required when we borrow state law. In general, the type of decision we face in the latter context is how to choose among various statutes of limitations, each of which was intended by the state legislature to apply to a whole category of causes of action. Federal statutes of limitations, on the other hand, are almost invariably tied to specific causes of action. The first consequence of this distinction is that in practice the inquiry as to which state statute to select will be very close to the traditional kind of classification question courts deal with all the time. Thus, for example, if a federal statute creates a cause of action that has elements of tort and contract, we may frame the question of which statute to apply as whether it is more “appropriate” to apply the State’s tort or contract limitations period. In reality, however, rather than examine whether the policies of the federal statute are better served by one limitations period than the other, we will generally an*169swer that question by determining whether the federal cause of action should be classified as sounding in tort or contract. See, e. g., Goodman v. Lukens Steel Co., 482 U. S. 656, 662 (1987) (42 U. S. C. § 1981 actions sound in tort); id., at 670 (Brennan, J., dissenting) (§ 1981 actions sound in contract). In deciding whether to borrow a federal statute that clearly does not apply by its terms, however, we genuinely will have to determine whether, for example, the Clayton Act’s limitations period will better serve the policies underlying civil actions under RICO than the limitations period covering criminal actions under RICO, or whether either will do the job better than state limitations upon actions for economic injury. That seems to me to be quintessentially the kind of judgment to be made by a legislature. See generally Wilcox v. Fitch, 20 Johns. *472, *475 (N. Y. 1823) (limitations are creatures of statute, and did not exist at common law); Wall v. Robson, 2 Nott & McCord 498, 499 (S. C. 1820) (same); 2 E. Coke, Institutes 95 (6th ed. 1680).
The second consequence of the generality of state statutes of limitations versus the particularity of federal ones is that in applying a state statute, we do not really have to make a new legislative judgment. The state legislature will already have made the judgment that, for example, in contract actions, a certain balance should be struck between “protecting valid claims . . . [and] prohibiting the prosecution of stale ones.” Johnson v. Railway Express Agency, 421 U. S. 454, 464 (1975). That judgment will have been made in the. knowledge that it will apply to a broad range of contractual matters, some of which the legislature has not specifically contemplated. That is not true of a federal statute enacted with reference to a particular cause of action, such as the one for the Clayton Act. The Court is clearly aware of this difficulty. It declines to apply 18 U. S. C. § 3282, the general 5-year criminal statute of limitations, on the ground that it “does not reflect any congressional balancing of the competing equities unique to civil RICO actions.” Ante, at 156. *170That objection should also, however, lead it to reject a 4-year limitations period, which clearly reflects only the balance of equities Congress deemed appropriate to the Clayton Act.
* * *
Thus, while I can accept the reasons the Court gives for refusing to apply state statutes of limitations to the civil RICO claim at issue here, ante, at 152-154, they lead me to a very different conclusion from that reached by the Court. I would hold that if state codes do not furnish an “appropriate” limitations period, there is none to apply. Such an approach would promote uniformity as effectively as the borrowing of a federal statute, and would do a better job of avoiding litigation over limitations issues than the Court’s approach. That was the view we took in Occidental Life Ins. Co. of Cal. v. EEOC, 432 U. S. 355 (1977), as to Title VII civil enforcement actions, unmoved by the fear that that conclusion might prove “‘“repugnant to the genius of our laws.’”” Ante, at 156, quoting Wilson v. Garcia, 471 U. S. 261, 271 (1985), in turn quoting Adams v. Woods, 2 Cranch 336, 342 (1805).5 See also 18 U. S. C. § 3281 (no limitations period for federal capital offenses). Indeed, it might even prompt Congress to enact a limitations period that it believes “appropriate,” a judgment far more within its competence than ours.
Although the opinion states that the Rules of Decision Act requires us to apply state statutes, 155 U. S., at 614, and therefore appears to suggest that the Act rather than the state laws themselves was the source of our obligation to do so, a careful reading of the opinion belies that interpretation. Because the Act directs the federal courts to regard state laws as rules of decision only “in cases where they apply,” the parties and the Court treated the questions of the applicability of the Act and the applicability of state law of its own force as interchangeable.
Thus, although we did not squarely reject our earlier approach until DelCostello v. Teamsters, 462 U. S. 151 (1983), the Court correctly argued in that case that our way of analyzing the issue had changed before then. Id., at 159-160, n. 13. Contrary to the DelCostello Court’s claim, however, neither our decision in Erie R. Co. v. Tompkins, 304 U. S. 64 (1938), nor the Rules of Decision Act scholarship underlying it in any way required that change. Neither remotely established that that statute applies only in diversity cases. See Hill, The Erie Doctrine in Bankruptcy, 66 Harv. L. Rev. 1013, 1033-1034 (1953); see also DelCostello v. Teamsters, supra, at 173, n. 1 (Stevens, J., dissenting) (noting that “‘the [Act] itself neither contains nor suggests ... a distinction’” between diversity and federal-question cases, quoting Campbell v. Haverhill, 155 U. S. 610, 616 (1895)); Friendly, In Praise of Erie—And of the New Federal Common Law, 39 N. Y. U. L. Rev. 383, 408, n. 122 (1964) (characterizing the view that Erie requires application of state law only in diversity cases as an “oft-encountered heresy”).
It need not always produce the same results, because the implicit directive attributed to Congress is not (as the old approach provided) that the courts apply the statute of limitations that the State deemed appropriate, but rather that the courts instead determine which state limitations period will best serve the policies of the federal statute. See, e. g., Automobile Workers v. Hoosier Cardinal Corp., 383 U. S. 696, 706 (1966); cf. Wilson v. Garcia, 471 U. S. 261, 268-269 (1985). Imagine, for example, a federal statute with no limitations period creating a cause of action in favor of handicapped persons discriminated against in the making of contracts. If a State had two statutes of limitations, one covering tortious personal injury, and one covering tortious economic injury, under the old approach the question would have been whether the federal statutory cause of action was an action for personal or economic injury. Under the new approach the question, at least in theory, is whether application of the personal injury or economic injury statute best serves the policies of the federal Act.
Second, even before conducting pre-emption analysis, the old approach can lead to the conclusion that state law supplies no statute of limitations. For example, that would be true in the case of our hypothetical federal statute if a State had limitations periods only for assault and battery. The new approach, however, should never lead to that conclusion, because we have already made the determination that federal law directs us to borrow some limitations period, and the only question is which one.
In fact, however, our analysis under the new approach has not been ruthlessly faithful to its logic, so that it has turned out in practice to be almost indistinguishable from the old approach. See infra, at 168-169.
Even DelCostello does not fully support the Court’s reformulation in the present opinion. It specifically noted that “our holding today should not be taken as a departure from prior practice in borrowing limitations periods for federal causes of action” and that it did “not mean to suggest that federal courts should eschew use of state limitations periods anytime state law fails to provide a perfect analogy.” 462 U. S., at 171. It also *168involved borrowing a federal statute that was arguably applicable by its own terms. Id., at 170. In any event, to the extent our decision here rests on our interpretation of congressional intent, the Court’s conclusion in that case that Congress intended § 10(b) of the National Labor Relations Act, 29 U. S. C. § 160(b), to be borrowed for suits claiming breach of the duty of fair representation tells us nothing as to what Congress intended in enacting RICO.
Because we claimed in DelCostello not to have abandoned our prior practice, that decision did not place Congress on notice that henceforth we would interpret its silence as a directive to borrow federal statutes of limitations. Any decision that the lower federal courts, whose regular task involves interpreting our opinions, did not understand to have worked a change in the law, see supra, at 167, is certainly not clear enough to form the basis for a presumption that Congress’ expectations were transformed. In any event, even if that decision had announced a general change of approach, to which it could be expected that Congress would adapt, it would only be appropriate to make the assumption that it had done so with respect to statutes passed after the decision came down. RICO was passed in 1970, well before our opinion in DelCostello. Pub. L. 91-452, 84 Stat. 943, 18 U. S. C. § 1963.
In Adams v. Woods, that argument was advanced not as a reason why the Court should apply a clearly inapplicable statute of limitations, but as a reason why it should interpret an arguably ambiguous one to apply to the claim at issue. 2 Cranch, at 341-342.