dissenting.
Because the same anomalous factor is present in both this case and in Graham County Soil & Water Conservation Dist. v. United States ex rel. Wilson, post, p. 409, and is decisive in my analysis of both cases, it is appropriate to explain *361my views in a single opinion. In each case the most natural reading of the statutory text would make it possible for the limitations period to expire before the cause of action accrues. Whether the source of this possible result is merely the use of careless wording or an incorrect assumption by Congress concerning the timing of two relevant events, I am convinced that Congress did not intend to authorize such a perverse result in either case. Thus, while I agree with much of the reasoning in the Court’s cogent opinion in Graham County, I write separately because I would agree with the Court of Appeals’ reading of the text of 31 U. S. C. § 3731(b)(1) were it not for this anomaly. In this case, however, because that same factor provides an even stronger reason for rejecting the interpretation of 28 U. S. C. § 2255, ¶ 6(3), that the Court endorses, I would reverse the judgment of the Court of Appeals.
I
In Graham County, the relator and the Government argue (and the Court of Appeals held) that the 6-year limitations period applicable to a “civil action under section 3730,” 31 U. S. C. § 3731(b)(1), applies to the retaliation action authorized by § 3730(h). That argument is supported by a literal reading of the statutory text; for § 3730(h) plainly qualifies as a “civil action under section 3730.” Moreover, that reading derives strong support from the interest in having a uniform federal statute of limitations govern the litigation of federal causes of action. Cf. Jones v. R. R. Donnelley & Sons Co., 541 U. S. 369, 377-383 (2004). Nevertheless, I agree with the Court that another reading of the text is far more plausible, and with its conclusion that when choosing between two constructions of a statute of limitations, whenever possible we should prefer the construction that starts the time limit running when the cause of action accrues.1
*362In Graham County that choice is compelled by the interaction between two relevant events: the “violation of §3729” and the retaliatory act against the whistle-blower. Section 3731(b)(1) provides that a “civil action under section 3730” must be brought within six years of a “violation of section 3729.” If this section were read to encompass retaliation claims under § 3730(h), as held by the Court of Appeals, the statute of limitations would be triggered by the “violation of section 3729”; that is, the limitations period would begin to run before the cause of action for retaliation accrues, and could potentially expire before an actionable retaliation claim even exists. See Graham County, post, at 421-422; United States ex rel. Wilson v. Graham County Soil & Water Conservation Dist., 367 F. 3d 245, 260-261 (CA4 2004) (Wilkinson, J., dissenting). Thus, the potentially prolonged time period between the two relevant events — the violation of § 3729 (triggering the limitations period) and the retaliation against the whistle-blower (giving rise to an actionable claim) — could leave the well-intentioned whistle-blower without any recourse under § 3730(h), the very statute designed to provide such protection.
*363The Court rightly avoids that harsh and counterintuitive result by adopting a construction of the statute that would generally start the running of the limitations period from the date the cause of action accrues, i. e., when the act or acts of retaliation occur. As Justice Thomas explains, that is not only the prevailing rule applied throughout the country to analogous state-law claims, Graham County, post, at 419-421, n. 3; it is also the background norm against which Congress legislates, Graham County, post, at 418-419. Because Congress surely did not intend to create a cause of action for retaliation with one hand, and impose with the other a premature trigger date for the limitations period with the potential to bar retaliation claims altogether, I concur in the judgment in Graham County.
II
The same potential for premature expiration of a statute of limitations is the primary reason why I cannot join the Court’s anomalous construction of the statute in this case. The statute we are called upon to interpret provides a 1-year period of limitation for a habeas petition that has as its basis a new rule of criminal law or criminal procedure that has retroactive application. Title 28 U. S. C. §2255, ¶ 6(3), provides that the “limitation period shall run 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.” There are two possible interpretations of when the period should start to run: from the date that this Court recognizes the new right, or from the date that both conditions2 in ¶ 6(3) are met.
*364If, as I believe Congress thought to be the case, this Court made a decision concerning a new rule’s retroactive application at the same time it recognized the new right, the statutory scheme would make perfect sense: Petitioners, whether filing an initial habeas petition or a second or successive petition, would have one year from this Court’s decision to file a petition for a writ taking advantage of that decision. Within a relatively short amount of time, those claims would be adjudicated, and the statute’s goals of finality would be duly served. In practice, however, this Court does not ordinarily make retroactivity judgments at the time a new right is recognized.3 See, e. g., Ring v. Arizona, 536 U. S. 584 (2002) (applying Apprendi v. New Jersey, 530 U. S. 466 (2000), to determinations of death penalty eligibility); Schriro v. Summerlin, 542 U. S. 348 (2004) (concluding Ring was not retroactive). Thus, as in Graham County, the statute implicates two relevant events: this Court’s recognition of a new right (which, according to the majority, triggers the limitations period) and the declaration that the right can be applied retroactively (which allows a petitioner to proceed with the claim). Because a significant amount of time may elapse during the interval between the triggering event and the point at which a petitioner may actually be able to file an action seeking relief under the statute, there is a real risk that the 1-year limitations period will expire before the cause of action accrues. In my judgment, the probable explanation for statutory text that creates this risk is Congress’ apparent assumption that our recognition of the new right and our decision to apply it retroactively would be made at the same time. Otherwise it seems nonsensical to assume that Congress deliberately enacted a statute that recognizes a cause of action, but wrote the limitations period in a way that *365precludes an individual from ever taking advantage of the cause of action.
We are thus faced with the same decision as in Graham County: Do we interpret the statute in such a way as to allow prisoners such as Dodd to take advantage of the full year Congress provided for such claims, or do we interpret the statute in such a way that the limitations period will begin to run before a prisoner may take advantage of ¶ 6(3)? As an initial matter, the text here certainly permits both readings, just as it did in Graham County. Paragraph 6(3) requires that two prerequisites must be met before a habeas petitioner can take advantage of that date as the starting point for the statute of limitations. Both requirements are in the past tense, and both must be satisfied before ¶ 6(3) is applicable. Furthermore, just as the clause “if that right has been newly recognized by the Supreme Court” describes the date indicated in the phrase “the date on which the right asserted was initially recognized,” it is possible to read the subordinate clause “if that right has been ... made retroactively applicable to eases on collateral review” as amplifying the description of the date in the provision’s main clause, rather than adding an additional qualifier. Consequently, while the majority’s reading of ¶6(3) — requiring that the statute of limitations begin to run when this Court recognizes a new rule — may be the more natural reading of the text, that advocated by petitioner — starting the statute of limitations when the new rule is held to be retroactive — is by no means implausible.4
*366Moreover, the potential for claims to be prematurely barred by the statute of limitations is even greater than in Graham County. There, the possibility that the 6-year statute of limitations period could run before the cause of action accrued, while plausible, was not particularly likely, since in *367most cases the retaliatory conduct that would form the basis of the cause of action under 31 U. S. C. § 3730(h) would probably occur within six years of the violation of §3729.5 In this case, owing to the substantially shorter 1-year statute of limitations period, both requirements of 28 U. S. C. § 2255, ¶ 6(3), will often not be met before the statute of limitations period has expired if it is triggered by the decision of the Supreme Court announcing a new rule.
That result is certainly true for Dodd himself. Richardson v. United States, 526 U. S. 813, was decided on June 1, 1999. Under the majority’s interpretation, the statute of limitations thus expired on June 1, 2000, one year after we recognized the new rule. The Eleventh Circuit, however, did not decide whether Richardson was retroactive until April 19, 2002.6 See Ross v. United States, 289 F. 3d 677 (CA11 2002) (per curiam). Thus, Dodd would not, under the majority’s interpretation, have been able to raise his claim at all, since the statute of limitations expired before he could have taken advantage of ¶ 6(3)’s 1-year grace period.7 *368Even for those prisoners who are incarcerated in a jurisdiction in which the new rule is quickly held to be retroactive, at least part of the 1-year period in which to file a claim taking advantage of the retroactive rule will run before the petition raising the claim can be filed.8
Thus, the admonition in Graham, County that “Congress generally drafts statutes of limitations to begin when the cause of action accrues,” post, at 418, applies with special force in this case. Paragraph 6(3) both recognizes a basis for habeas relief by allowing an otherwise barred claim to go forward if certain conditions are met, and also sets forth a 1-year statute of limitations for such claims. It would make no sense for Congress, in the same provision, both to recognize a potential basis for habeas relief and also to make it highly probable that the statute of limitations would bar relief before the claim can be brought. Again, this is not simply a remote possibility: It is true for Dodd himself, and in six of the seven Circuits to have addressed whether Richardson is retroactive. See n. 7, supra. It is this absurd result that convinces me that Congress could not have in*369tended that ¶ 6(3) should be read in this manner. Even if the text is as clear as the majority claims (a proposition I reject), we should still interpret the text in a manner that would avoid such an absurd result. See, e.g., Clinton v. City of New York, 524 U. S. 417, 429 (1998); Church of Holy Trinity v. United States, 143 U. S. 457, 459 (1892).
To avoid this result, I would interpret ¶ 6(3) to begin to run only when the Supreme Court has initially recognized the new right and when that right has been held to be retroactive. Under this interpretation, the statute of limitations would not begin to run until the prisoner was actually able to file a petition under ¶ 6(3), which is the only interpretation Congress could have intended. Although in enacting AEDPA Congress was clearly concerned with finality, see Duncan v. Walker, 533 U. S. 167, 179 (2001), ¶ 6(3) is an explicit exception to that general preference. Congress surely intended to allow habeas petitioners to take advantage of the new rights that this Court deems retroactive. Otherwise, there would have been no reason to include that section in the statute. That is why, “[ajbsent other indication, a statute of limitations begins to run at the time the plaintiff has the right to apply to the court for relief.” Graham County, post, at 419 (quoting TRW Inc. v. Andrews, 534 U. S. 19, 37 (2001) (Scalia, J., concurring in judgment; internal quotation marks omitted).9
*370In addition to creating the perverse result that the statute of limitations will run before a prisoner can file an initial habeas petition, the Court’s myopic reading of ¶ 6(3) effectively nullifies 28 U. S. C. § 2244(b)(2)(A), which allows prisoners to file second or successive applications based on a retroactive rule.10 As the majority recognizes in what amounts to a dramatic understatement, its interpretation of ¶ 6(3) “makes it difficult for applicants filing second or successive §2255 motions to obtain relief.” Ante, at 359. Because of the way ¶¶ 6(3) and 8(2) interact, a prisoner can only file a second or successive petition based on a newly recognized rule that has been made retroactive if this Court has held *371the rule to be retroactive within one year of recognizing it. Unfortunately for such prisoners, however, this Court has never done so since Teague v. Lane, 489 U. S. 288 (1989), was decided.11 Because of the need for percolation, and the time it takes for cases to come to this Court from the courts below, it seems unlikely (to say the least) that we would ever do so. Therefore, the majority’s interpretation of ¶ 6(3) effectively nullifies ¶ 8(2). It is, of course, a basic canon of statutory construction that we will not interpret a congressional statute in such a manner as. to effectively nullify an entire section. See, e. g., Duncan, 533 U. S., at 174 (“[A] statute ought, upon the whole, to be so construed that, if it can be prevented, no clause, sentence, or word shall be superfluous, void, or insignificant” (internal quotation marks omitted)). It is a strange principle that requires strict adherence to the text of one provision while allowing another to have virtually no real world application. It would seem far wiser to give both sections the meaning that Congress obviously intended.
Accordingly, while I concur in the judgment in Graham County, I respectfully dissent in Dodd.
Contrary to the Court’s comment in Graham County, post, at 419, n. 2, I do not suggest that a statute providing that the limitations period begins to run before the cause of action accrues is necessarily ambig*362uous. Rather, as Justice Thomas’ scholarly footnote demonstrates, Graham County, post, at 419-421, n. 3, it is so unlikely that a legislature would actually intend such an anomalous design that I would presume' that the anomaly was the product of a drafting error absent evidence in either the legislative history or elsewhere in the text that Congress specifically intended such a result. See Koons Buick Pontiac GMC, Inc. v. Nigh, 543 U. S. 50, 65-66 (2004) (Stevens, J., concurring).
The literal text of § 3731(b)(1), which uses an event that is not an element of the retaliation cause of action to start the limitations period running, produces two anomalies: (1) The statute may never begin to run, and (2) it may expire before the cause of action accrues. The Court argues that the first anomaly makes the statute ambiguous and that the second justifies resort to a default rule to resolve the ambiguity. In my judgment, the latter anomaly would provide a sufficient justification for resort to the default rule whether or not some other feature of the statute would support an argument that the text was “ambiguous.”
Section 2255, ¶ 6(3), technically has three requirements: that a right be “initially recognized,” that it be “newly recognized,” and that it be “made retroactively applicable.” In practice, however, the first two requirements are one and the same. Hence, in this opinion I will refer only to the requirements (1) that a right be newly recognized and (2) that it be made retroactively applicable.
The retroactivity issue is not normally argued in the same case that announces a new rule because the prisoner is only interested in the outcome of his own case. Moreover, in order to minimize the impact of the new rule at issue, he actually has an incentive to minimize its consequences.
I should note an additional point of disagreement with the majority (and with petitioner). In reaching its result, the Court relies on an assumption made by both parties and not challenged in this Court: namely, that the decision to make a new rule retroactive for purposes of this section can be made by any lower court. While I recognize that every Circuit to have addressed the issue has made the same assumption, I am satisfied that the Government’s initial interpretation of this provision is the correct one. See Brief for United States as Amicus Curiae in Tyler v. Cain, O. T. 2000, No. 00-5961, p. 16, n. 7. Under that interpretation, the requirement that the “right has been newly recognized by the Su*366preme Court and made retroactively applicable to cases on collateral review” is met only if the Supreme Court has made the right retroactive.
Courts that have reached the contrary conclusion have principally relied on the fact that 28 U. S. C. § 2244(b)(2)(A) contains an explicit requirement that a new rule be “made retroactive .. .by the Supreme Court.” (Emphasis added.) See Ashley v. United States, 266 F. 3d 671, 674 (CA7 2001). Thus, the argument goes, the absence of “by the Supreme Court” after “made retroactive” must have some meaning. However, in that clause there is only one verb that the prepositional phrase “by the Supreme Court” can modify, whereas in the relevant clause of §2255, ¶ 6(3), there are two: newly recognized and made retroactive. The more natural reading of ¶ 6(3) is that the prepositional phrase “by the Supreme Court” modifies both verbs of the subordinate clause. This reading comports with Congress’ general direction that this Court, and not the lower courts, should provide the final answer to questions of interpretation arising under the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA). See, e. g., 28 U. S. C. § 2254(d)(1) (requiring that a state-court decision be contrary to “clearly established Federal law, as determined by the Supreme Court of the United States” (emphasis added)). Additionally, it avoids difficult questions of which court can make a retroactivity determination, sets a uniform date by which lower courts can make determinations as to whether a petition is timely, and m'eans that only those cases made retroactive by this Court can form the basis for a petition that can gain the benefit of tolling under § 2255, ¶ 6(3). Finally, it is the only interpretation that gives full effect to § 2255, ¶ 8(2), which allows prisoners who have already completed one round of federal habeas review to seek additional relief on the basis of such a new rule.
Ultimately, this reading has no direct bearing on the question presented in this case. While my view that this Court must make the retroactivity determination informs my belief that Congress had a mistaken understanding of how ¶ 6(3) would operate in practice, I would conclude that the 1-year limitations period begins to run when both requirements of ¶ 6(3) are met regardless of which court makes the retroactivity decision.
Justice Souter, Justice Ginsburg, and Justice Breyer do not join this footnote.
As the majority in Graham County noted, however, in almost every ease the statute of limitations would begin to run before the cause of action actually accrued. See post, at 421.
This assumes that the Eleventh Circuit is the relevant “court” to decide the retroactivity question, an issue the majority fails to address. Even if a district court, as opposed to the Court of Appeals, could make that determination for purposes of ¶ 6(3), the District Court for the Southern District of Florida has not decided the issue in a published opinion.
This would be true for prisoners in every Circuit except the Sixth Circuit, in which a prisoner would have had six months to file his petition. See Murr v. United States, 200 F. 3d 895 (Jan. 7, 2000). In the five other Circuits besides the Eleventh to have decided the issue, all held Richardson v. United States, 526 U. S. 813 (1999), to be retroactive more than one year after Richardson was decided; in all of those Circuits, prisoners’ claims under Richardson would be time barred before they were able to file under ¶ 6(3). See Santana-Madera v. United States, 260 F. 3d 133 (CA2, Aug. 3, 2001); United States v. Lopez, 248 F. 3d 427 (CA5, Apr. 16, 2001); Lanier v. United States, 220 F. 3d 833 (CA7, June 12, 2000); United *368States v. Montalvo, 331 F. 3d 1052 (CA9, June 9, 2003) (per curiam); United States v. Barajas-Diaz, 313 F. 3d 1242 (CA10, Dec. 3, 2002); Ross v. United States, 289 F. 3d 677 (CA11, Apr. 19, 2002) (per curiam). The Eighth Circuit appears to have assumed the retroactive application of Richardson, but that too was decided more than a year after Richardson itself. See United States v. Scott, 218 F. 3d 835 (July 7, 2000). Of course, if any of the other four Circuits that have not yet decided the issue were to conclude Richardson was retroactive, the statute of limitations would have long since expired, and prisoners would be similarly barred from taking advantage of any such decision.
In Tyler v. Cain, 533 U. S. 656 (2001), it appeared that a majority of the Court recognized that the Court could make a new rule retroactive “through multiple holdings that logically dictate the retroactivity of the new rule.” Id., at 668 (O’Connor, J., concurring). In such a case, a prisoner could file a petition under ¶ 6(3) immediately. Since there was much disagreement over when that would be the ease, however, that potential exception holds small comfort in this case.
The approach that the Court takes in Graham County and the approach I would take here has support in our prior case law. In Fleischmann Constr. Co. v. United States ex rel. Forsberg, 270 U. S. 349 (1926), the Court was faced with the interpretation of the Materialmen’s Act of 1894, as amended, which allowed a private creditor to bring suit against a party contracting with the United States, provided that the United States did not itself bring suit “within six months from the completion and final settlement” of the contract. 33 Stat. 812. Such a creditor had one year from the completion of the contract and final settlement to bring a suit, giving him a 6-month window within which to file his claims. If any other creditors wanted to bring suit, they had to join the action of the original *370creditor, but under the statute had only one year from “the completion of the work” in which to do so. Ibid. As the Court in Fleischmann recognized, if taken literally this last section would have meant that in a case in which the “final settlement” of the contract occurred more than six months after work was completed on the contract — as “frequently” happened — only the initial creditor to file suit would have been able to meet the requirement of the statute of limitations; any subsequent creditor would have been barred under the second statute of limitations that did not reference the final settlement as a start date, but rather only the completion of work. 270 U. S., at 361. Rather than permit these “unjust or absurd consequences,” id., at 360, the Court interpreted “within one year from the completion of the work” to mean “ ‘within one year after the performance and final settlement of the contract,’ ” id., at 362.
Fleischmann thus presents the identical situation as in Graham County and Dodd: Because of the unforeseen possibility that two relevant events might occur far apart in time, the most natural reading of the statute would cause the statute of limitations to expire before the suit may be brought. As we did in Fleischmann and in Graham County, we should construe the statute of limitations in Dodd to avoid this unnatural result.
Our cases make clear that when interpreting a particular section of a statute, we look to the entire statutory scheme rather than simply examining the text at issue. See Koons, 543 U. S., at 60. “A provision that may seem ambiguous in isolation is often clarified by the remainder of the statutory scheme — because the same terminology is used elsewhere in a context that makes its meaning clear, or because only one of the permissible meanings produces a substantive effect that is compatible with the rest of the law.” United Sav. Assn. of Tex. v. Timbers of Inwood Forest Associates, Ltd., 484 U. S. 365, 371 (1988) (citation omitted and emphasis added).
Again, it is possible that a combination of our decisions has effectively-done this, see n. 8, supra, but we have never actually recognized an instance in which that has occurred.