UNITED STATES of America, Plaintiff-Appellee, v. Raymond Roger SURRATT, Jr., Defendant-Appellant. Steven Harris Goldblatt, Court-Assigned Amicus Counsel, National Association of Criminal Defense Lawyers; Douglas A. Berman, Professor, Amici Supporting Appellant, 797 F.3d 240

Affirmed by published opinion. Judge AGEE wrote the majority opinion, in which Chief Judge TRAXLER joined. Judge GREGORY wrote a dissenting opinion. AGEE, Circuit Judge: In 2005, after pleading guilty to conspiracy to'distribute cocaine, Raymond Surratt was sentenced to life imprisonment. We affirmed his conviction and sentence on appeal, and Surratt’s motion to vacate his conviction and sentence under 28 U.S.C. § 2255 was likewise denied. Neither Sur-ratt’s direct appeal nor his § 2255 motion questioned the legality of his mandatory life sentence.

Several years later, Surratt returned to this Court and asked for permission to file a second or successive § 2255 motion. Surratt’s request was premised on United States v. Simmons, 649 F.3d 237 (4th Cir.2011) (en banc), which in turn overruled our prior decision in United States v. Harp, 406 F.3d 242 (4th Cir.2005). Had Surratt been sentenced after Simmons, he would have faced a lower mandatory minimum sentence than the mandatory life term that he actually received. Surratt maintained that this difference entitled him to be resentenced.

But Congress set out certain conditions that must be met before a successive motion may be permitted, and Surratt did not meet those required conditions. See 28 U.S.C. § 2255(h). We therefore denied him permission to file a successive motion. See In re Surratt, No. 12-283 (4th Cir. Sept. 13, 2012), ECF No. 6. In the district court, Surratt had simultaneously filed a petition for a writ of habeas corpus under 28 U.S.C. § 2241 seeking the same Simmons-based relief.

As a federal prisoner, however, Surratt cannot challenge his conviction and sentence under § 2241 unless 28 U.S.C. § 2255(e) — also called the “savings clause” — applies. The district court concluded that § 2255(

e) did not in fact confer jurisdiction to consider Surratt’s claim in a § 2241 petition, so it denied Surratt’s petition. Surratt now appeals from the judgment of the district court. We are not unsympathetic to his claim; like the dissent, we recognize the gravity of a life sentence. However, Congress has the power to define the scope of the writ of habeas corpus, and Congress has exercised that power here to narrowly limit the circumstances in which a § 2241 petition may be brought. Surratt’s petition does not present one of the permitted circumstances. Accordingly, we agree that the district court lacked jurisdiction under § 2255(

e) to consider Surratt’s § 2241 petition and affirm the judgment below. I. In 2004, a grand jury indicted Surratt on several drug-related counts, including conspiracy to possess with intent to distribute cocaine and cocaine base in violation of 21 U.S.C. §§ 841(b)(1)(

A) and 846. The Government then filed a timely information indicating that it would seek enhanced penalties based on Surratt’s criminal history. The Government’s information identified four previous drug-related convictions, each in North Carolina: (1) a 1996 conviction for felony possession of cocaine; (2) a 1997 conviction for felony possession of cocaine; (3) a 1997 conviction for felony possession of cocaine and maintaining a place for storage or sale; and (4) a 1998 conviction for sale and delivery of cocaine.

If two or more of these convictions constituted “felony drug offense[s],” then Sur-ratt faced a mandatory term of life imprisonment without release. See 21 U.S.C. § 841(b)(1)(A). Four months after his indictment, and despite the prospect of a life sentence, Surratt pleaded guilty to the conspiracy count. He acknowledged in his written plea agreement that the district court could not impose a sentence below any statutory minimum unless the United States sought a reduction for substantial assistance. He also waived any rights to further appeals, save in a few narrow circumstances.

Surratt did not stipulate, however, to any prior felony drug convictions. After Surratt’s plea, but before his sentencing, we decided United States v. Harp, 406 F.3d 242 (4th Cir.2005), which held that a North Carolina drug conviction qualified as a “felony drug offense” if “the maximum aggravated sentence that [the state court] could [have] imposed for that crime upon a defendant with the worst possible criminal history” exceeded one year. Id. at 246. Under Harp, all Sur-ratt’s prior convictions constituted felony drug offenses.

Therefore, unless Surratt offered substantial assistance to the Government, Surratt faced a mandatory life sentence. The district court sentenced Surratt to life imprisonment in October 2005. Initially, the court expressed some misgivings about a life sentence and questioned whether a different sentence would apply had it been free to consider the factors listed in 18 U.S.C. § 3553(a). But the Government had declined to move for any substantial-assistance reduction, as it regarded Surratt’s cooperation as “halfhearted ... at best.” J.A. 223.

Consequently, the district court determined that it was “required” to impose the life sentence “because of [Surratt’s] prior criminal history and his engagement in this drug trafficking and, for whatever reasons, his inability to render substantial assistance.” J.A. 222. We affirmed Surratt’s sentence on appeal. United States v. Surratt, 215 Fed.Appx. 222, 224 (4th Cir.2007). Surratt did not raise any claim in his direct appeal as to his mandatory life sentence. Surratt then moved for post-conviction relief under 28 U.S.C. § 2255 in April 2008.

In that motion, Surratt claimed that he had received ineffective assistance of counsel during his initial plea and sentence, and further sought a sentence reduction based on an amended Sentencing Guideline. He did not challenge his mandatory minimum sentence. The district court denied and dismissed the motion. Surratt v. United States, Nos. 3:08cv181, 3:04cr250, 2011 WL 815714 (W.D.N.C. Feb. 25, 2011). We denied Surratt’s request for a certificate of appealability. United States v. Surratt, 445 Fed.Appx. 640, 640 (4th Cir.2011).

More than three years after Surratt filed his first § 2255 motion, we decided Simmons. Overruling Harp, the Simmons en banc majority held that a prior North Carolina conviction will constitute a felony for purposes of an enhanced punishment only if the prior conviction was actually punishable for more than one year of imprisonment as to that defendant. 649 F.3d at 241. Surratt and the Government agree that only one of Surratt’s prior convictions would qualify as a “felony drug offense” under Simmons.

Seeking to take advantage of Simmons, Surratt sought relief in both this Court and the district court in August 2012. He first asked this Court to permit him to file a second or successive § 2225 motion. Because Surratt’s motion fell outside the statutorily enumerated exceptions permitting that type of motion, see 28 U.S.C. § 2255(h), we denied him permission to file. See In re Surratt, No. 12-283 (4th Cir. Sept. 13, 2012), ECF No. 6. On the same day that he filed that request, Sur-ratt also moved in the district court to vacate his sentence under §§ 2241 and 2255, or for writ of coram nobis.

Surratt maintained that, in light of Simmons, he was “innocent” of the career offender enhancement and was a victim of fundamental error. In particular, Surratt argued that he should not have been subject to a mandatory life sentence. The Government did not oppose Surratt’s § 2241 request. Despite the parties’ agreement, the district court denied Surratt’s motion. Surratt v. United States, Nos. 3:04-CR-250-19, 3:12-CV-513, 2014 WL 2013328 (W.D.N.C. May 16, 2014).

Although it again expressed misgivings about Surratt’s mandatory life sentence, the court determined that a federal prisoner may use § 2241 to seek relief that § 2225(

h) would otherwise bar only when a substantive change in the law coming after an initial § 2255 petition rendered the prisoner’s original offense conduct “non-criminal.” Id. at *1. Because Surratt challenged only his sentence, the court concluded that he could not establish that Simmons rendered any of his conduct “noncriminal.” Therefore, he could not obtain relief under § 2241. Surratt appealed, arguing that the district court should have permitted him to proceed under § 2241. The Government agreed and also elected not to enforce the appeal waiver in Surratt’s plea agreement.

Therefore, in light of the Government’s position, we appointed amicus curiae to defend the district court’s judgment. II. In the Antiterrorism and Effective Death Penalty Act of 1996 (“AEDPA”), Congress circumscribed the ability of federal prisoners to request post-conviction relief. See Rhines v. Weber, 544 U.S. 269, 274, 125 S.Ct. 1528, 161 L.Ed.2d 440 (2005).

Section 2555 provides the ordinary means for a federal prisoner to challenge his conviction or sentence. But in AED-PA, Congress limited the jurisdiction of federal courts to hear second or successive requests under § 2255. See, e.g., In re Weathersby, 717 F.3d 1108, 1110 (10th Cir.2013) (“Congress placed strict limitations on second or successive motions under § 2255[.]”).

Specifically, courts may hear second or successive petitions only if they pertain to (1) “newly discovered evidence ... [clearly and convincingly establishing] that no reasonable factfinder would have found the movant guilty of the offense” or (2) “a new rule of constitutional law, made retroactive to cases on collateral review by the Supreme Court, that was previously unavailable.” 28 U.S.C. § 2255(h). We have already held that Surratt’s claim does not fall into either of these categories, so § 2255(

h) does not permit him to file a second or successive motion under § 2255. If a federal prisoner cannot meet § 2255(h)’s requirements, then he may seek to file a traditional petition for writ of habeas corpus under 28 U.S.C. § 2241. But that right carries significant limits as well. Specifically, a prisoner “may file a habeas petition under § 2241 only if the collateral relief typically available under § 2255 ‘is inadequate or ineffective to test the legality of his detention.’ ” Prousalis v. Moore, 751 F.3d 272, 275 (4th Cir.2014) (quoting 28 U.S.C. § 2255(e)).

If a federal prisoner brings a § 2241 petition that does not fall within the scope of this “savings clause,” then the district court must dismiss the “unauthorized habeas motion ... for lack of jurisdiction,” Rice v. Rivera, 617 F.3d 802, 807 (4th Cir.2010), even if the Government supports the prisoner’s position. Surratt is a federal prisoner who means to file a § 2241 petition for post-conviction relief. Therefore, we must assess whether § 2255 is “inadequate or ineffective to test the legality of his detention.” 28 U.S.C. § 2255(e). We consider that question de novo. See Yi v. Fed.

Bureau of Prisons, 412 F.3d 526, 530 (4th Cir.2005). We have jurisdiction over this appeal under 28 U.S.C. § 1291. III. A. We have determined that § 2255 was inadequate or ineffective in only one prior instance. In that case, petitioner Byron Jones argued that Bailey v. United States, 516 U.S. 137, 116 S.Ct. 501, 133 L.Ed.2d 472 (1995), rendered his convictions for “use” of a firearm during a drug offense invalid. In re Jones, 226 F.3d 328, 329 (4th Cir.2000).

Bailey had overruled this circuit’s understanding of what it meant to “use” a firearm, id. at 330 (citing Bailey, 516 U.S. at 143, 116 S.Ct. 501), and Jones contended that his conduct therefore did not amount to “use” under Bailey’s reading of the term, id. at 334. Jones, however, had already filed a § 2255 motion before Bailey, and § 2255(

h) barred him from filing another one. Id. at 330. We decided to award relief, deeming § 2255 “inadequate or ineffective to test the legality of a conviction” when: (1) at the time of conviction, settled law of this circuit or the Supreme Court established the legality of the conviction; (2) subsequent to the prisoner’s direct appeal and first § 2255 motion, the substantive law changed such that the conduct of which the prisoner was convicted is deemed not to be criminal; and (3) the prisoner cannot satisfy the gatek-eeping provisions of § 2255 because the new rule is not one of constitutional law.

Jones, 226 F.3d at 333-34. In short, Jones opened a narrow gateway to § 2241 relief for certain prisoners found actually innocent of their offenses of conviction, allowing relief only where the acts for which the defendant was convicted are not a crime. Since then, we have focused on this aspect of Jones — actual innocence of a criminal act — when characterizing the decision. See, e.g., Farrow v. Revell, 541 Fed.Appx. 327, 328-29 (4th Cir.2013); Darden v. Stephens, 426 Fed.Appx. 173, 174 (4th Cir.2011); Rice, 617 F.3d at 807; United States v. Poole, 531 F.3d 263, 267 n. 7 (4th Cir.2008).

Other circuits also give substantial attention to the actual innocence aspect of Jones when discussing it. See, e.g., Bryant v. Warden, FCC Coleman-Medium, 738 F.3d 1253, 1280 (11th Cir.2013); Abernathy v. Wandes, 713 F.3d 538, 546 n. 7 (10th Cir.2013); Stephens v. Herrera, 464 F.3d 895, 898 (9th Cir.2006); Abdullah v. Hedrick, 392 F.3d 957, 962 (8th Cir.2004); Cephas v. Nash, 328 F.3d 98, 104 n. 6 (2d Cir.2003); Reyes-Requena v. United States, 243 F.3d 893, 904 (5th Cir.2001). The dissent mistakenly tries to read the actual innocence requirement out of Jones.

That course ignores the clear limitation in Jones that, before the case can be used to invoke § 2255(e), the law must have changed “such that the conduct of which the prisoner was convicted is deemed not to be criminal.” 226 F.3d at 334. The caveat wasn’t accidental, but drew from the fundamental recognition that actual innocence of a crime is different from other circumstances. Indeed, Jones “agree[d] with the rationale” of other courts that made this principle even more explicit. Id. at 333; see, e.g., Triestman v.

United States, 124 F.3d 361, 378-79 (2d Cir.1997) (permitting, a Bailey claim to be brought via § 2241 in part because of the constitutional issues that would arise from refusing to hear an actual-innocence claim); see also Jones, 226 F.3d at 333 n. 3 (“[T]hese courts have focused on the more fundamental defect presented by a situation in which an individual is incarcerated for conduct that is not criminalf.]”).

Were the dissent’s approach right, Jones would permit any federal prisoner to bring any non-constitutional claim via § 2241 in any instance where the law relevant to that claim changed in the petitioner’s favor at any time. That hardly describes “a limited number of circumstances” in which § 2255(

e) would apply. Jones, 226 F.3d at 333. B. Jones simply does not apply here, as Surratt is not innocent of anything. All parties agree that Simmons did not decriminalize any part of Surratt’s underlying conduct. Surratt admits that he conspired to distribute drugs, the offense for which he was convicted. He nevertheless argues -that his predicate convictions constitute elements of a separate, aggravated offense for recidivists. If that proved true, we might say that Surratt was not guilty of the prior conviction “element,” and therefore innocent of the recidivist offense.

Yet we do not treat a prior conviction “as an element of [the] offense.” Almendarez-Torres v. United States, 523 U.S. 224, 247, 118 S.Ct. 1219, 140 L.Ed.2d 350 (1998); see also Alleyne v. United States, — U.S. -, 133 S.Ct. 2151, 2160 & n. 1, 186 L.Ed.2d 314 (2013). Indeed, “the [Supreme] Court has consistently held that recidivism is not an element of the triggering crime[.]” Appleby v. Warden, N. Reg’l Jail & Corr. Facility, 595 F.3d 532, 539 (4th Cir.2010).

Many good reasons justify this principle: recidivism does not directly relate to the underlying charged conduct, courts may judicially notice convictions, prior convictions could prejudice the defendant if the Government put them before the jury, and courts have treated recidivism as a sentencing factor for over 200 years. See United States v. Cheek, 415 F.3d 349, 353-54 (4th Cir.2005); see also Jones v. United States, 526 U.S. 227, 248-49, 119 S.Ct. 1215, 143 L.Ed.2d 311 (1999). Surratt insists that his sentence enhancements are different. They are not.

We have treated the predicate convictions described in § 841 as sentencing enhancements, not elements of a separate crime. See, e.g., United States v. Smith, 451 F.3d 209, 224 (4th Cir.2006); accord Jeanty v. Warden, FCI-Miami, 757 F.3d 1283, 1286 (11th Cir.2014); United States v. Ramirez, 714 F.3d 1134, 1141 (9th Cir.2013). The Government did not charge the predicate convictions here in the indictment. And, had the case gone to trial, it would not have put them before a jury and proven them beyond a reasonable doubt.

We also do not believe that Alleyne altered the longstanding rule treating predicate convictions as enhancements, not elements. Alleyne held only that a jury must find certain facts that increase a mandatory minimum sentence. 133 S.Ct. at. 2155. How Alleyne would apply here remains something of a mystery, as the parties do not identify any authority indicating that the case could apply retroactively on collateral review. Quite the contrary: courts have held that Alleyne does not apply retroactively. See United States v.

Olvera, 775 F.3d 726, 730 (5th Cir.2015) (“We reiterate that Alleyne does not apply retroactively. This decision accords with that of every circuit to have examined the issue, none of which has decided that Alleyne is retroactive.”). Regardless, Alleyne held that facts other than a prior conviction that increase a mandatory minimum amount to “elements” of the crime. 133 S.Ct. at 2160-63. The Supreme Court has specifically refused to apply this reasoning to prior convictions. Id. at 2160 & n. 1. So, “Almendarez-Torres remains good law,” even after Alleyne. United States v.

McDowell, 745 F.3d 115, 124 (4th Cir.2014). To say “that a petitioner can be ‘actually innocent’ of a sentencing enhancement,” rather than an element of the actual crime, “would require a great deal of both verbal and logical gymnastics.” Turner v. Warden Coleman FCI (Medium), 709 F.3d 1328, 1334 n. 3 (11th Cir.2013); see also, e.g., Marrero v.

Ives, 682 F.3d 1190, 1193 (9th Cir.2012) (“[For purposes of the savings clause,] a petitioner generally cannot assert a cognizable claim of actual innocence of a noncapital sentencing enhancement.”); In re Davenport, 147 F.3d 605, 609 (7th Cir.1998) (explaining that “it would make an arbitrary hole in the Anti-terrorism Act” if the court were to accept an actual-innocence argument about “a sentence-enhancement statute”); cf.

In re Bradford, 660 F.3d 226, 230 (5th Cir.2011) (“[A] claim of actual innocence of a career offender enhancement is not a claim of actual innocence of the crime of conviction and, thus, not the type of claim that warrants review under § 2241.”). The traditional view of actual innocence focuses on the elements of the crime of conviction, see United States v. Mikalajunas, 186 F.3d 490, 494 (4th Cir.1999), and nothing in Jones deviates from that settled approach. Cf. Sawyer v.

Whitley, 505 U.S. 333, 341, 112 S.Ct. 2514, 120 L.Ed.2d 269 (1992) (remarking that, “[i]n the context of a noncapital case, the concept of ‘actual innocence’ is easy to grasp,” as it means the defendant did not commit the charged crime). Jones, after all, concerned “conduct ... deemed not to be criminal,” not conduct that remains criminal but subject to a lesser penalty. Jones, 226 F.3d at 334; see also United States v.

Jones, 758 F.3d 579, 586 (4th Cir.2014) (“Innocence of conviction implicates the notion that a person has been incarcerated for a crime he did not commit, whereas a sentencing error does not at all implicate guilt.”). The Supreme Court, too, has not said that conduct is rendered “lawful” merely because the penalty for it is reduced. Schriro v. Summerlin, 542 U.S. 348, 354, 124 S.Ct. 2519, 159 L.Ed.2d 442 (2004), for example, found that Ring v.

Arizona, 536 U.S. 584, 122 S.Ct. 2428, 153 L.Ed.2d 556 (2002), did not “render! ] some formerly unlawful conduct lawful or vice versa,” even though Ring affected the length of sentences for some prisoners. So too here: Surratt remains as guilty today as he was in 2005. C. We recognize that we have applied broader concepts of actual innocence in other contexts. In United States v.

Pettiford, 612 F.3d 270 (4th Cir.2010), drawing on earlier eases, we said that a defendant might establish “actual innocence” sufficient to excuse his procedural default if “the defendant clearly showed that he had not committed the crime on which the calculation of his [recidivist] sentence was based.” Id. at 283-84; see also Mikalajunas, 186 F.3d at 494-95; United States v. Maybeck, 23 F.3d 888, 892 (4th Cir.1994).

In other words, Pettiford and its predecessors suggested that, as to procedural default, a defendant might be “actually innocent” even if he is only innocent of a predicate conviction, not just the offense of conviction. But we have also explained that Pettiford’s view of actual innocence should be confined to its original context. Pettiford developed in the judge-made realm of procedural default, where we enjoy more latitude to shape our own rules and doctrines. We deal here, however, with a congressionally imposed statutory constraint that is not so amenable to ad hoc, judge-made exceptions.

See Jones, 758 F.3d at 586 (holding that movant could not evade the § 2255(f)(4) statute of limitations by claiming “actual innocence of his sentence”). We are especially reluctant to blaze broad pathways to § 2255(

e) relief given that the provision constitutes only a “narrow exception.” Cleaver v. Maye, 773 F.3d 230, 232 (10th Cir.2014); accord Bryant, 738 F.3d at 1283; Hill v. Werlinger, 695 F.3d 644, 647-48 (7th Cir.2012); Trenkler v. United States, 536 F.3d 85, 98 (1st Cir.2008). Even if we apply Pettiford’s broader view, however, we still cannot call Surratt “actually innocent.” “[A]ctual innocence,” the Supreme Court has told us, “means factual innocence, not mere legal insufficiency.” Bousley v. United States, 523 U.S. 614, 623, 118 S.Ct. 1604, 140 L.Ed.2d 828 (1998); see also United States v.

Tyler, 732 F.3d 241, 246 (3d Cir.2013) (“While Bousley addressed the standard that a petitioner must meet for claims brought under § 2255, this standard applies equally to actual innocence claims brought under § 2241.”). So, “actual innocence applies in the context of habitual offender provisions only where the challenge to eligibility stems from factual innocence of the predicate crimes, and not from the legal classification of the predicate crimes.” Pettiford, 612 F.3d at 284; accord Damon v. United States, 732 F.3d 1, 5-6 (1st Cir.2013); McKay v. United States, 657 F.3d 1190, 1199 (11th Cir.2011).

Surratt does not suggest that he did not commit the crimes listed in the Government’s original information. He only says Simmons changed the relevance of those crimes in determining his later sentence. Surratt’s argument constitutes the sort of argument about “legal classification” that we have deemed insufficient. IV. Although Jones does not present a portal through which Surratt may pass to obtain post-conviction relief, Jones is not the exclusive route to § 2255(

e) relief in all situations. As the Government observes, we “had no occasion” in Jones “to consider whether sentencing errors” like the one claimed here “are redressable under the savings clause.” Gov’t Br. 23; see also Illinois v. Lidster, 540 U.S. 419, 424, 124 S.Ct. 885, 157 L.Ed.2d 843 (2004) (explaining that “general language in judicial opinions” must be read “as referring in context to circumstances similar to those then before the Court”). We must examine the text of the savings clause to decide whether the statute can afford Surratt any relief. See United States v. Neuhauser, 745 F.3d 125, 128 (4th Cir.2014) (starting with the text in a question of statutory

interpretation). The parties principally discuss whether Surratt’s case presents a “fundamental miscarriage of justice” and then assume that this characterization could trigger the savings clause. Certainly, in deciding whether to provide relief in an initial § 2255 motion, we have considered whether a non-constitutional error “involves a fundamental defect which inherently results in a complete miscarriage of justice.” Mikalajunas, 186 F.3d at 495; see also, e.g., United States v. Martinez, 139 F.3d 412, 417 n. * (4th Cir.1998).

But we have never tied this “miscarriage” standard to the entirely separate question of relief via the savings clause. Such an approach would confuse the question of cognizability (that is, what motions may be brought via § 2241 generally) with the separate question that we address here (that is, what § 2241 claims are barred by § 2255(e)). And the “miscarriage” standard addresses a different question in still another way.

In particular, the Court adopted the “fundamental defect” and “miscarriage” standard as a way of identifying those cases where a petitioner’s failure to bring his claim in his initial direct appeal could be excused. See Reed v. Farley, 512 U.S. 339, 354, 114 S.Ct. 2291, 129 L.Ed.2d 277 (1994). In contrast, § 2255(

e) applies to prisoners who failed to bring their claims twice before: on direct appeal and in their initial § 2255 motion. Thus, an

interpretation of § 2255(

e) focusing on a “miscarriage of justice” “is less an argument for interpreting § 2255 than it is one for amending § 2255.” Prost v. Anderson, 636 F.3d 578, 597 (10th Cir.2011); accord Williams v. Warden, Fed. Bureau of Prisons, 713 F.3d 1332, 1346 (11th Cir.2013) (calling the “miscarriage of justice standard” “plainly inapplicable” to the question of whether § 2255(

e) forecloses a particular § 2241 petition); Perez v. Stephens, 593 Fed.Appx. 402, 403 (5th Cir.2015) (“Perez’s argument that the actual innocence and miscarriage of justice standards provide an exception to the § 2255 savings clause requirement fails[.]”). As an appellate court, we are not in the business of amending statutes. See Dodd v. United States, 545 U.S. 353, 359, 125 S.Ct. 2478, 162 L.Ed.2d 343 (2005) (indicating that the court was “not free to rewrite the statute that Congress has enacted” even though the restrictions on § 2255 petitions created “the potential for harsh results in some cases”). And holding that the “miscarriage of justice” standard opens the § 2255(

e) portal would not just ignore the text of that

section — it would also erase the limitations on initial motions and “second or successive” motions found in §§ 2255(

f) and (h). We hold that § 2255(e)’s text does not permit Surratt to raise his claim under § 2241. And although the dissent insists that we act “without any textual basis,” dissenting op. at 271, we observe at least four separate textual factors — in addition to the relevant provisions’ context and purpose — that establish that Surratt’s claim is foreclosed. A. The savings clause first tells us that “the remedy by [a § 2255] motion [must be] inadequate or ineffective to test the legality of [Surratt’s] detention” before he can proceed under § 2241. 28 U.S.C. § 2255(e) (emphasis added).

We find Congress’s deliberate use of the word “test,” rather than a more expansive term like “guarantee” or “ensure,” very meaningful. To “test” means to “put to the proof.” Oxford English Dictionary (2d ed.1989). In other words, the statutory text anticipates the opportunity to raise a challenge or argument. See, e.g., Wag More Dogs, Ltd. Liab. Corp. v. Cozart, 680 F.3d 359, 365 n. 3 (4th Cir.2012) (equating the government’s ability “to test the plausibility of the claim” with its ability to raise the relevant argument and offer particular evidence); NLRB v.

Inter-City Advertising Co., 154 F.2d 244, 246 (4th Cir.1946) (saying that an employer had no way “to test” an administrative determination where it had no legal avenue to object). Therefore, “the clause is concerned with process— ensuring the petitioner an opportunity to bring his argument — not with substance— guaranteeing nothing about what the opportunity promised will ultimately yield in terms of relief.” Prost, 636 F.3d at 584; see also Taylor v. Gilkey, 314 F.3d 832, 835 (7th Cir.2002) (agreeing that § 2255(e)’s use of the term “test” “implies a focus on procedures rather than outcomes”).

We too have focused on the “opportunity” or “unobstructed procedural shot” afforded by § 2255, not on the ultimate result of a motion under that section, in defining the scope of the savings clause. Rice, 617 F.3d at 807. We have stressed that § 2255(

e) will not permit an individual’s § 2241 petition “merely because [the] individual has been unable to obtain relief under [§ 2255] or because an individual is procedurally barred from filing a § 2255 motion.” In re Vial, 115 F.3d 1192, 1194 n. 5 (4th Cir.1997). In § 2255, Congress meant to provide a chance to be heard, not a right to prevail on any particular argument. See, e.g., Cleaver, 773 F.3d at 233 (offering an example of permissible savings clause relief in which “the original sentencing court has been abolished or dissolved and the petitioner has nowhere to file a § 2255 petition”).

Surratt clearly could have “test[ed]” the legality of his detention in his initial § 2255 motion. Surratt also chose to fore-go any challenge to his mandatory sentence on direct appeal. At both steps, then, he had the opportunity to claim that a lower mandatory sentence should have applied because he did not qualify as a two-time drug felon.

His choice to focus his § 2255 motion on claims of ineffective assistance of counsel proves most important here, as he cannot complain of his inability to “test” his detention in a post-conviction proceeding when § 2255 offered him that chance and he declined to take it. “A prisoner cannot be permitted to lever his way into

section 2241 by making his

section 2255 remedy inadequate[J” Morales v. Bezy, 499 F.3d 668, 672 (7th Cir.2007). But Surratt protests that “the basis for [his] claim was foreclosed by circuit precedent at the time[J” Opening Br. 20. That argument fails for at least two reasons. As a factual matter, we cannot know whether Surratt’s claim might have led to relief because he never made it. We do know that “many defendants ... filed suits prior to Simmons asserting the exact same substantive claim that [Surratt] now raises, including, of course, Simmons himself.” Whiteside v. United States, 775 F.3d 180, 186 (4th Cir.2014) (en banc) (collecting cases). And indeed, “[t]he nature of statutory

interpretation requires that someone present the argument before the courts can define the law” or, as in Simmons, change it. Bryan Florendo, Prost v. Anderson and the Enigmatic Savings Clause of § 2555: When Is A Remedy By Motion “Inadequate or Inejfective”?, 89 Denv. U.L. Rev. 435, 454 (2012). As a legal matter, courts do not permit petitioners special favors because the petitioners misjudged their claims as futile and chose not to present them in the first instance.

In fact, the Supreme Court has held a “dismissive” attitude toward “the contention that an argument’s legal futility is the same as. a court’s inability to entertain the argument[J” Brown v. Caraway, 719 F.3d 583, 598 (7th Cir.2013) (Easter-brook, J., statement concerning the circulation under Circuit Rule 40(e)). In the procedural default context, for instance, “futility cannot constitute cause if it means simply that a claim was unacceptable to that particular court at that particular time.” Bousley, 523 U.S. at 623, 118 S.Ct. 1604. Futility also provides.no basis to toll § 2255’s statute of limitations.

See Whiteside, 775 F.3d at 185; Minter v. Beck, 230 F.3d 663, 666-67. (4th Cir.2000) (same). The same principle applies here: “[t]he § 2255 remedial vehicle was fully available and amply sufficient to test the argument, whether or not [Surratt] thought to raise it.” Prost, 636 F.3d at 589. Admittedly, through Simmons, “a shift in the law ... has given increased relevance to a point made at the trial but not pursued on appeal.” Sunal v. Large, 332 U.S. 174, 182, 67 S.Ct. 1588, 91 L.Ed. 1982 (1947).

But to permit this shift to open the gate to habeas relief would render “litigation in these criminal cases ... interminable.” Id.; cf. Whiteside, 775 F.3d at 186 (noting that the “myriad of substantive changes in law past the point of finality” would cause “a tectonic shift of resources” if courts treated them as enough to evade procedural bars).

Section 2255(

e) means to preserve a prisoner’s “remedy,” not his “relief.” See, e.g., Tolliver v. Dobre, 211 F.3d 876, 878 (5th Cir.2000) (“We join our sister circuits that have held that a prior unsuccessful § 2255 motion ... do[es] not make § 2255 inadequate or ineffective.”). This distinction is reflected in other portions of the habeas statutes, where “remedy” similarly means an avenue or mechanism for relief, not the relief itself. See § 2254(b)(1)(A) (referring to exhaustion of state “remedies”); see also Vial, 115 F.3d at 1194 n. 5 (referring to § 2255’s “remedy” as separate from its “relief’).

That distinction likewise implies that what matters is the ability to make the request, not the ability to win it. These reasons aside, we note that substantial practical difficulties could follow if we held that a prisoner could not “test” his detention because adverse circuit precedent once “foreclosed” his claim. Neither Surratt nor the Government informs us, for instance, how clear circuit precedent must be before it “forecloses” the prisoner’s claim. Perhaps, under the parties’ approach, analogous precedent might do; perhaps not.

Nor do they tell us why our precedent alone “forecloses” an argument even though Supreme Court review remains available. And they do not explain when the foreclosure must arise. We find it better to avoid these questions by adhering to a more direct understanding of “test.” As Jones reflects, our

interpretation of “test” yields in cases involving actual innocence of the crime of conviction. “[A] thorny constitutional issue” could result if “no other avenue of judicial review [were] available for a party who claims that s/he is factually or legally innocent as a result of a previously unavailable statutory

interpretation^]” In re Dorsainvil, 119 F.3d 245, 248 (3d Cir.1997); accord Triestman, 124 F.3d at 378-79; cf. Davis v. United States, 417 U.S. 333, 346-47, 94 S.Ct. 2298, 41 L.Ed.2d 109 (1974). “For under our federal system it is only Congress, and not the courts, which can make conduct criminal.” Bousley, 523 U.S. at 620-21, 118 S.Ct. 1604. Generally, “courts will construe [a] statute to avoid [constitutional] problems unless such construction is plainly contrary to the intent of Congress.” United States v. Hamilton, 699 F.3d 356, 367 (4th Cir.2012).

Therefore, we read Jones as an outgrowth of this constitutional avoidance doctrine. But because Surratt presents no genuine claim of actual innocence of the crime of conviction, no “thorny constitutional issues” arise here. B. In much the same way as “test,” the statute’s references to inadequacy and ineffectiveness tell us that § 2255(

e) preserves only the chance to request relief, not the ultimate and absolute right to obtain it. We see this notion in cases evaluating whether other “substitute” habeas remedies are adequate and effective. In Swain v. Pressley, 430 U.S. 372, 97 S.Ct. 1224, 51 L.Ed.2d 411 (1977), the Supreme Court examined a provision of the District of Columbia Code that channeled collateral review of convictions in the Superior Court of the District of Columbia back into that court (rather than through federal courts via habeas review).

In finding the provision adequate and effective, the Court focused on the opportunity that the statute afforded, including the relief that the prisoner could request and the manner in which the court would consider his request. Id. at 381-84, 97 S.Ct. 1224. Similarly, in Boumediene v. Bush, 553 U.S. 723, 128 S.Ct. 2229, 171 L.Ed.2d 41 (2008), the Court analyzed whether combatant status review tribunals afforded an adequate and effective substitute to habeas corpus for enemy combatants held at Guantanamo Bay, Cuba.

The Court again concentrated on the “opportunity” that the substitute offered — not on the outcome, id. at 779, 128 S.Ct. 2229, asking whether the “sum total of procedural protections” sufficed, id. at 783, 128 S.Ct. 2229. These cases both make the same fundamental point: the chance to argue a claim is the relevant criterion for adequacy and effectiveness, not any particular disposition of that claim. When discussing “effectiveness” and “adequacy” in other areas of the law, we likewise focus on processes, not outcomes.

We do not declare that counsel rendered “ineffective” assistance any time his particular argument fails and the client loses. Similarly, a state’s appellate procedures satisfy the Fourteenth Amendment’s guarantee to an “adequate and effective” appeal so long as those procedures provide certain minimum safeguards. See Smith v. Robbins, 528 U.S. 259, 276-84, 120 S.Ct. 746, 145 L.Ed.2d 756 (2000). When deciding whether we can enjoin a state criminal prosecution, we also must pass on whether the party seeking relief “has an adequate remedy at law.” Trainor v.

Hernandez, 431 U.S. 434, 440, 97 S.Ct. 1911, 52 L.Ed.2d 486 (1977). In doing so, we consider whether the movant has an “opportunity to raise and have timely decided ... the federal issues involved.” Id. at 441, 97 S.Ct. 1911. These and other examples suggest that we should not assess § 2255’s effectiveness by weighing the historic outcome under that motion against the outcome we would reach today in an initial appeal from trial. We should instead ask only whether the procedures in § 2255 provided a genuine opportunity for the petitioner to raise his present claim.

Surratt never suggests that the § 2255 mechanism denied him a chance to make his present argument. At best, he says only that our pre-Simmons reading of § 841(b)(1)(

A) did. See Prost, 636 F.3d at 590 (“[Wjhenever legal error occurs it may very well mean circuit law is inadequate or deficient. But that does not mean the § 2255 remedial vehicle is inadequate or ineffective to the task of testing the argument.”). We understand Surratt’s reluctance to attack § 2255 head-on, as courts have upheld the statute as an adequate and effective means to challenge the detention of a federal prisoner for some 60 years. United States v. Hayman, 342 U.S. 205, 219, 72 S.Ct. 263, 96 L.Ed. 232 (1952).

But the point remains: no deficiency in § 2255 itself precluded Surratt from lodging his sentencing challenge, so we cannot call § 2255 “inadequate” or “ineffective” here. C. The savings clause also focuses on the “legality” of the relevant detention. Especially in the post-conviction context, courts have recognized “unlawful” or “illegal” sentences in a narrow subset of cases.

Actual innocence of the crime of conviction may present that sort of a case, as courts have long understood that “[a]n imprisonment under a judgment” becomes “unlawful” if “that judgment be an absolute nullity.” Ex parte Watkins, 28 U.S. (3 Pet.) 193, 203, 7 L.Ed. 650 (1830). And a sentence imposed above the proper statutory maximum might present another instance of an unlawful sentence, as “the power ... to prescribe the punishments to be imposed upon those found guilty of [federal crimes] resides wholly with the Congress.” Whalen v.

United States, 445 U.S. 684, 689, 100 S.Ct. 1432, 63 L.Ed.2d 715 (1980); see also United States v. Addonizio, 442 U.S. 178, 186, 99 S.Ct. 2235, 60 L.Ed.2d 805 (1979); accord Sun Bear v. United States, 644 F.3d 700, 705 (8th Cir.2011) (“An unlawful or illegal sentence is one imposed without, or in excess of, statutory authority.”); United States v. Gonzalez-Huerta, 403 F.3d 727, 739 n. 10 (10th Cir.2005) (“[T]he term ‘illegal sentence’ is reserved for those instances where the term of incarceration exceeds the statutory maximum[.]”).

As we explained in another context, though, “not every [proceeding] alleging a legal error in sentencing challenges that sentence as ‘illegal.’ ” United States v. Thornsbury, 670 F.3d 532, 539 (4th Cir.2012). We have said that “illegal” sentences “involve[] much more fundamental issues — such as challenges claiming a district court exceeded its authority, claiming that a sentence was based on a constitutionally impermissible factor such as race, or claiming a post-plea violation of the right to counsel.” Id.

In like manner, the Supreme Court explained that a sentence was “not illegal” where, among other things, “[t]he punishment meted out was not in excess of that prescribed by the relevant statutes [and] multiple terms were not imposed for the same offense.” Hill v. United States, 368 U.S. 424, 430, 82 S.Ct. 468, 7 L.Ed.2d 417 (1962); accord United States v. Pavlico, 961 F.2d 440, 443 (4th Cir.1992). In accordance with this view, courts largely have not recognized an “illegal” detention — one that would trigger the savings clause — where the defendant challenges a sentence within the correct statutory maximum.

See, e.g., United States v. Powell, 691 F.3d 554, 563 (4th Cir.2012) (King, J., dissenting in part and concurring in part) (indicating that a sentence would be “lawful” if it “falls within the unen-hanced statutory maximum”). The Sixth Circuit noted this general rule, for example, in holding that the defendants’ sentence-focused claims did not “fall within any arguable construction” of the savings clause. United States v. Peterman, 249 F.3d 458, 462 (6th Cir.2001).

The Eleventh Circuit likewise held that a federal prisoner could not bring a claim that “would otherwise be barred by § 2255(h)” in a § 2241 petition where the asserted error “resulted in a longer sentence not exceeding the statutory maximum.” Gilbert v. United States, 640 F.3d 1293, 1323 (11th Cir.2011) (en banc); see also Bryant, 738 F.3d at 1288 (“[A]ny § 2241 challenge to a sentence that is already below the authorized statutory maximum could not open the § 2255(

e) portal.”); accord Trenkler, 536 F.3d at 99 (labeling petitioner’s claim “incompatible with engagement of the savings clause” where he did “not charge that [his] life sentence [wa]s beyond the- statutory maximum for the crimes of conviction”). For a time, at least, even the Government agreed that a federal prisoner could not attack a sub-maximum sentence through a § 2241 petition. See Wilson v. Wilson, No. 1:11cv645 (TSE/TCB), 2012 WL 1245671, at *3 n. 2 (E.D.Va.

Apr. 12, 2012) (quoting the “position of the United States” that “savings clause relief is foreclosed for an erroneous sentence within the statutory maximum”). That view also comports with § 2255’s scope, which provides sentencing relief only in cases involving “violations- of statutes establishing maximum sentences.” United States v. Pregent, 190 F.3d 279, 284 (4th Cir.1999). This widely held view of “legality” should foreclose Surratt’s claim. He never suggests that he received a sentence above the applicable statutory maximum. He cannot.

All parties agree that Surratt has at least one prior felony drug conviction,, regardless of Simmons. A defendant like Surratt who violates 21 U.S.C. §§ 841 and 846 after a single “prior conviction for a felony drug offense has become final ... shall be sentenced to a term of imprisonment which may not be less than 20 years and no more than life imprisonment[.]” 21 U.S.C. § 841(b)(1)(A). Surratt often emphasizes that his sentence is “without parole,” but that does not create any special issue., In the present federal system, life imprisonment is equivalent to life without parole in every case. See Richmond v.

Polk, 375 F.3d 309, 316 (4th Cir.2004) (“[T]he Sentencing Reform Act of 1984 abolished parole for federal offenses committed after November 1, 1984[.]”); see also 18 U.S.C. § 3624(b) (stating that prisoners serving life -sentences are not entitled to a deduction for good time credits). When Congress abolished parole in 1984, it essentially “translate[d] every life sentence into life without possibility of parole.”. United States v. Gonzalez, 922 F.2d 1044, 1051 (2d Cir.1991).

Thus, Surratt’s life sentence does not exceed the maximum that the district court could have imposed had Simmons been issued before Surratt’s sentencing. In other words, Surratt’s sentence does not offend the limits of punishment, such that we could question his detention’s “legality.” Tacitly recognizing the weakness of his position under the majority view, Surratt urges us to adopt a distinct minority view that he considers “persuasive.” Opening Br. 22 (citing Brown, 719 F.3d 583). But even under this outlier decision, Surratt cannot prevail.

In Brown, the Seventh Circuit permitted a prisoner to challenge the length of his sentence through a § 2241 petition even though that sentence fell below the statutory maximum. Id. at 588. But Brown concerned a pr e-Booker sentence imposed under the formerly mandatory Guidelines regime. See id. (stressing that the defendant was sentenced in “the pr e-Booker era”) (citing United States v. Booker, 543 U.S. 220, 125 S.Ct. 738, 160 L.Ed.2d 621 (2005)).

For the Seventh Circuit, this historical fact made all the difference, as it believed that — at least when courts treated the Guidelines as mandatory—“the only lawful sentence was a [G]uidelines sentence.” Id. Regardless of whether the Seventh Circuit correctly characterized the former Guidelines regime (and we need not say here whether it did, see Gilbert, 640 F.3d at 1306-07), its view on the savings-clause issue exists in harmony with the approach taken by all other courts.

Quite simply, the Seventh Circuit believed that it could permit a § 2241 petition because, at least pr e-Booker, an unjustified above-Guidelines sentence equated to a sentence above the statutory maximum. The top of the Guidelines range provided the relevant maximum sentence, and Brown’s sentence exceeded the pre-Booker maximum Guidelines sentence. In contrast, in cases involving the gost-Booker advisory Guidelines, even the Seventh Circuit recognizes that courts cannot correct sentencing errors in post-conviction proceedings, under § 2241 or otherwise, if the sentence “is below the statutory maximum.” Hawkins v.

United States, 706 F.3d 820, 823 (7th Cir.2013). Here, the district court sentenced Sur-ratt under the post-Booker, advisory Guidelines scheme. As a result, his life sentence does not exceed the statutory maximum sentence, and even Brown would afford him no relief. Thus, we cannot deem his detention “illegal.” D. Pointing to the savings clause’s reference to “detention,” Surratt suggests Congress meant to permit claims like his. But Congress used the term “sentence” in some portions of § 2255, see 28 U.S.C. § 2255(a)-(b), and “conviction” or “offense” in other parts, id. §§ 2255(f)(1), (h)(1).

It chose not to use those terms in the savings clause, id. § 2255(e), and we “are obligated to give effect to Congress’s decision to use different language in proximate subsections of the same statute.” United States v. Brandon, 247 F.3d 186, 190 (4th Cir.2001). We think the statutory term “detention” is yet another textual indication that § 2255(

e) bars Surratt’s petition; it draws a line not between conviction and sentencing challenges, but between permissible challenges to executive acts and impermissible challenges to the acts of other branches of government. “Detention” concerns the act of physically confining or restraining an individual. See, e.g., Oxford English Dictionary (2d ed.1989) (“Keeping in custody or confinement.”); Black’s Law Dictionary (10th ed. 2014) (“The act or an instance of holding a person in custody; confinement or compulsory delay.”).

Physical confinement stands separate and apart from the sentence, and some detentions that may give rise to a habeas, petition do not follow criminal sentences at all. Pre-trial detentions provide one common example, see 18 U.S.C. § 3142(e), but other instances come to mind, including detentions in instances of civil contempt, detentions in connection with immigration proceedings, detentions of material witnesses, or civil detention of sexually dangerous persons, see Neuhauser, 745 F.3d at 125-31.

Rather than being tied in all cases to a sentence as criminal punishment, then, § 2255(e)’s reference to “detention” speaks to physical restraint of a person’s liberty. In contrast to the rendering of a criminal sentence by the judiciary, the physical holding of a federal prisoner represents an executive function, a “keeping back or withholding” by the executive branch. See Samak v. Warden, FCC Coleman-Medium, 766 F.3d 1271, 1280 (11th Cir.2014) (Pryor, J., concurring).

The Attorney General, acting through the Bureau of Prisons, acts as the “detainer,” and the sentencing court does not involve itself in keeping the prisoner restrained. See 18 U.S.C. § 3621(a) (explaining that the Bureau of Prisons has the responsibility to detain a person “who has been sentenced to a term of imprisonment ... until the expiration of the term imposed, or until earlier released for satisfactory behavior”); see also, e.g., Janko v. Gates, 741 F.3d 136, 141 (D.C.Cir.2014) (agreeing with a party’s concession that “courts ordinarily don’t detain people”).

We see this concept in action in those instances where Congress precluded the sentencing court from involving itself too much in the ongoing aspects of a detention. See United States v. Wilson, 503 U.S. 329, 334, 112 S.Ct. 1351, 117 L.Ed.2d 593 (1992) (holding that a sentencing court may not calculate credit for time already spent in custody, as that task belongs to the Attorney General); see also, e.g., United States v. Ceballos, 671 F.3d 852, 855 (9th Cir.2011) (“[T]he [sentencing] court has no jurisdiction to select the place where the sentence will be served.

Authority to determine place of confinement resides in the executive branch of government[.]”). Even the venue provisions of §§ 2241 and 2255 reflect an executive versus non-executive dichotomy. A petitioner must bring his § 2241 petition in the district of confinement, where the court can summon the warden to defend his actions. In contrast, a petitioner must move for § 2255 relief before the district court that sentenced him, where the sentencing court and the original prosecuting United States Attorney may bring their expertise to bear. A challenge to “detention” through § 2255(

e) should therefore focus on acts of the executive branch, rather than the judicial branch. See Munaf v. Geren, 553 U.S. 674, 693, 128 S.Ct. 2207, 171 L.Ed.2d 1 (2008) (“Habeas is at its core a remedy for unlawful executive detention.”); INS v. St. Cyr, 533 U.S. 289, 301, 121 S.Ct. 2271, 150 L.Ed.2d 347 (2001) (same); accord Samak, 766 F.3d at 1291 (Pryor, J., concurring); Trenkler, 536 F.3d at 96. Hence, a § 2241 attack on “detention” lodged through § 2255(

e) should entail a challenge to (1) the right and authority of the executive to keep the individual in custody; or (2) the manner in which the executive executes the^ detention. That reading serves the historic purposes of the writ as well. Boumediene, 553 U.S. at 745, 128 S.Ct. 2229 (“The [Suspension] Clause protects the rights of the detained by affirming the duty and authority of the Judiciary to call the jailer to account.”). Focusing on the executive’s power of restraint draws a clear line — a line consistent with long-accepted petitions under § 2241.

For example, we have previously noted that § 2241 petitioners can challenge the Bureau of Prison’s calculation of good time credits. See Yi, 412 F.3d at 530. That makes sense: if a prisoner deserves good-time credits but the Bureau of Prisons denies them, then an error by the executive may cause the warden to hold the prisoner without authority — that is, beyond the time prescribed by Congress and the sentencing court. See Preiser v.

Rodriguez, 411 U.S. 475, 487, 93 S.Ct. 1827, 36 L.Ed.2d 439 (1973) (finding that prisoners’ petition concerning good-time credits “fell squarely within th[e] traditional scope of habeas corpus” because “[t]hey alleged that the deprivation of their good-conduct-time credits was causing or would cause them to be in illegal physical confinement” (emphasis added)). Similarly, when federal sentencing allowed for parole, § 2241 provided a vehicle to challenge parole revocation. See Vial, 115 F.3d at 1194 n. 5.

A pre-trial detainee might also invoke § 2241 if he is not timely arraigned or tried; in that situation, the warden’s acts of continuing to detain the prisoner could render the detention improper under the Fifth and .Sixth Amendments. See United States v. Tootle, 65 F.3d 381, 383 (4th Cir.1995). And the same principle may explain our decision in Jones: if a change in law renders the prisoner’s conduct “non-criminal,” then the warden would have no authority to hold the now-innocent defendant.

Contrast these commonly accepted claims with Surratt’s present one, which has little to do with any executive-branch action. By all accounts, the warden is holding. Surratt under a validly entered sentencing order. Surratt does not suggest that the warden is violating his statutory or constitutional rights. And the warden is detaining Surratt pursuant to a sentence that Congress empowered the district court to impose — that is, a sentence within the statutory limits. Sur-ratt’s petition does not therefore attack any act of detention by the executive branch.

Thus, Congress would not have expected federal prisoners to bring this type of a claim in a § 2241 petition via the savings clause. E. The broader context of § 2255 further convinces us that Surratt cannot invoke § 2241. “[Statutory construction is a holistic endeavor, and we therefore must evaluate the statutory language itself, the specific context in which such statutory language is used, and the broader context of the statute as a whole.” R.H. Donnelley Corp. v. United States, 641 F.3d 70, 76 (4th Cir.2011). Particularly in post-conviction proceedings like this one, context often proves decisive.

See, e.g., Taylor, 314 F.3d at 836 (relying in part on the “history” of § 2255(

e) to reject a putative § 2241 petition). If Surratt were permitted to raise his sentencing challenge in a § 2241 petition, we would thwart almost every one of the careful limits that Congress placed on post-conviction challenges to a federal prisoner’s sentence. And we would do so in a broad class of cases. Quite simply, “Congress would have accomplished nothing at all in its attempts — through statutes like AEDPA — to place limits on federal collateral review.” Triestman, 124 F.3d at 376.

This comprehensive conflict between Sur-ratt’s view and so many different parts of the statute convinces us that we are not “piek[ing] and choos[ing]” some parts of the statute over others, as the dissent alleges. Instead, we are simply “construing] all parts to have meaning.” United States v. Briley, 770 F.3d 267, 273 (4th Cir.2014). If we embraced Surratt’s position, then the limits found in § 2255(

h) would first fall. That

section provides that a prisoner may bring a second or successive § 2255 motion only if the prisoner (1) produces newly discovered evidence of actual innocence; or (2) relies upon “a new rule of constitutional law, made retroactive to cases on collateral review by the Supreme Court, that was previously unavailable.” 28 U.S.C. § 2255(h). The prisoner cannot bring his second or successive motion until we grant him a certificate of authorization indicating that the motion satisfies one of these two conditions.

Id. “When Congress adopted § 2255(h), it was undoubtedly aware that prisoners might wish to press other sorts of arguments,” that is, arguments other than actual innocence or constitutional error, “in second or successive motions.” Prost, 636 F.3d at 585. Nevertheless, Congress decided to bar those “other” arguments outright. Surratt, however, means to bring one of the very claims that Congress foreclosed, albeit under a different caption.

That course would produce an illogical, if not absurd, result: § 2255(h)’s restrictions would continue to apply to prisoners bringing Congressionally preferred claims, while no restrictions would apply to prisoners bringing claims that Congress did not contemplate at all. Surratt’s version of § 2255(

e) would then return us, at least in part, to the pre-AEDPA world, where “prisoners [could] file as many collateral attacks as they please[d], provided that they [did]n’t abuse the writ.” Taylor, 314 F.3d at 836. Moreover, Surratt premises his claim on a circuit-level decision, even though § 2255(

h) specifically states that only a retroactive Supreme Court decision should open the door to successive relief. No court has permitted a petitioner to bring a claim via the savings clause premised only-on a change in circuit-level authority. Instead,- the available authority all requires Supreme Court precedent. See Tyler, 732 F.3d at 246 (listing an “intervening Supreme Court decision” as one of several facts required to bring a claim via the savings clause); Kenemore v.

Roy, 690 F.3d 639, 641 (5th Cir.2012) (stating that a claim brought via the savings clause must be “based on a retroactively applicable Supreme Court decision”); Davenport, 147 F.3d at 611 (noting an “obvious” “qualification[ ]” that a “change of law” could not qualify a prisoner to seek § 2241 relief unless “the change of law [w]as :.. made retroactive by the Supreme Court”); Wofford v.

Scott, 177 F.3d 1236, 1245 (11th Cir.1999) (“[T]he only sentencing claims that may conceivably be covered by the savings clause are those based upon a retroactively applicable Supreme Court decision overturning circuit precedent.”). Holding otherwise would elevate circuit court precedent to the level of retroactive Supreme Court precedent without any congressional authority to do so. The Government once recognized this point, too. See Brief for the Respondent in Opposition at 14, Williams v.

Hastings, — U.S. -, 135 S.Ct. 52, 190 L.Ed.2d 29 (2014) (No. 13-1221), 2014 WL 3749512 (arguing that the savings clause applies to petitions raising certain sentencing errors “when the error is revealed by a retroactively applicable decision of [the Supreme] Court”). In pressing a contrary position, the best case the dissent can summon is one in which another circuit court permitted a § 2241 petition to be brought based on an assumedly retroactive Supreme Court decision. See Light v. Caraway, 761 F.3d 809, 814 (7th Cir.2014).

Given that Surratt’s case involves no Supreme Court decision at all, that case strikes us as irrelevant. Surratt’s concept of retroactivity produces other anomalous results. Under his approach, a petitioner invoking a new statutory

interpretation from the Fourth Circuit could file immediately for § 2241 relief, with no need to wait for a declaration of retroactivity. Meanwhile, a petitioner invoking a new constitutional rule promulgated by the Supreme Court would have to wait for an explicit statement from that court rendering the rule retroactive. See Tyler v. Cain, 533 U.S. 656, 663, 121 S.Ct. 2478, 150 L.Ed.2d 632 (2001).(“The new rule becomes retroactive, not by the decisions of the lower court or by the combined action of the Supreme Court and the lower courts, but simply by the action of the Supreme Goúrt.”).

We “resist attributing to Congress an intention to render a statute so internally inconsistent.” Greenlaw v. United States, 554 U.S. 237, 251, 128 S.Ct. 2559, 171 L.Ed.2d 399 (2008). Movants under § 2255 must also comply with a one-year statute of limitations, 28 U.S.C. § 2255(f), and this provision would also fall victim to Surratt’s view. A prisoner hamstrung by § 2255(

f) could file a § 2241 petition, as § 2241 contains no statute of limitations. We only recently held, en banc, that the supposed futility of raising a Simmons — like claim before we decided Simmons would not permit an untimely § 2255 motion raising that sort of claim. See Whiteside, 775 F.3d at 184-87. But if Surratt proved correct, then that decision would mean next to nothing. A different caption on his otherwise time-barred petition would permit the federal prisoner to go forward.

Were we to adopt Surratt’s view, the differing venue provisions in § 2241 and § 2255 would also invite mischief in circumstances like these. A prisoner sentenced in this circuit would file his initial § 2255 motion here. If that motion failed, and the prisoner happened to be held in a different circuit (as so often happens in the federal prison system), then the prisoner could seek relief via § 2241 in that circuit.

One of two unusual things would then happen: the § 2241 court would either apply the law of a different circuit to assess where the prisoner’s petition stands, see, e.g., Samak, 766 F.3d at 1275 n. 3, or it would apply its own law, see, e.g., United States v. Wyatt, 672 F.3d 519, 523 (7th Cir.2012). Either way, any number of inconsistent results could flow from this disjointed approach. By contrast, if challenges like Surratt’s are confined to the § 2255 motions, then only the sentencing court will consider them, and the motion will only implicate our circuit’s law. Surratt’s

interpretation of the savings clause would also have strange implications for appeal rights. “Congress has the power to preclude any appeal from an order dismissing a writ of habeas corpus, since a party to a suit has no vested right to an appeal.” United States v. Brooks, 245 F.3d 291, 293 (3d Cir.2001). And Congress exercised that power: a § 2255 mov-ant cannot appeal the district court’s decision on his motion unless the district or circuit court first grants a certificate of appealability. See 28 U.S.C. § 2253(c)(1)(B).

That certificate only issues when the movant has “made a substantial showing of the denial of a constitutional right.” See id. § 2253(c)(2); see also Jennings v. Stephens, — U.S. -, 135 S.Ct. 793, 807, 190 L.Ed.2d 662 (2015) (Thomas, J., dissenting) (“One of the key ways in which AEDPA encourages finality is to narrow the scope of appellate review by requiring habeas petitioners to obtain [certificates of appealability].”). So, a § 2255 movant who raises a Simmons claim might be unable to appeal because his claim does not present constitutional issues. See United States v.

Hadden, 475 F.3d 652, 665 (4th Cir.2007) (“[BJecause a [certificate of appealability] will not issue for allegations of non-constitutional error, he would not be able to obtain appellate review of any purely statutory errors at his sentencing[.]”); accord Young v. United States, 523 F.3d 717, 718 (7th Cir.2008); United States v. Christensen, 456 F.3d 1205, 1206 (10th Cir.2006); Mateo v. United States, 310 F.3d 39, 41-42 (1st Cir.2002); United States v. Brooks, 230 F.3d 643, 646 (3d Cir.2000). But § 2241 contains no such limits on appeal.

Therefore, applying Surratt’s reading of the savings clause, a petitioner could evade Congress’s conscious constraint on appellate review just by filing a § 2241 petition, losing, and then exercising his automatic entitlement to appellate review. We should not allow the savings clause to “create a detour” around § 2255’s restrictions, as those restrictions would then be “rendered a nullity.” Reyes-Requena, 243 F.3d at 901 n. 19; cf Trenkler, 536 F.3d at 97 (“The strictures of

section 2255 cannot be sidestepped by the simple expedient of resorting to some more exotic writ.”). Accordingly, we decline to dismantle the protections that Congress created in § 2255 by adopting Surratt’s expansive reading of the savings clause. F. Lastly, we must take account of Congress’ general purpose in enacting the various provisions at issue. See Broughman v. Carver, 624 F.3d 670, 677 (4th Cir.2010); accord McCreary Cnty. v. ACLU, 545 U.S. 844, 861, 125 S.Ct. 2722, 162 L.Ed.2d 729 (2005) (“Examination of purpose is a staple of statutory

interpretation that makes up the daily fare of every appellate court in the country.”). Congressional purpose confirms that § 2255(

e) does not apply to cases like this one. Our narrower reading of the savings clause preserves one of the most important purposes that AEDPA sought to serve: finality. Although AEDPA serves multiple objectives, including “comity, finality, and federalism,” Williams v. Taylor, 529 U.S. 420, 436, 120 S.Ct. 1479, 146 L.Ed.2d 435 (2000), the. foremost AEDPA goal — the act’s “central concern” — is Congress’ intent that “the merits of concluded criminal proceedings not be revisited in the absence of a strong showing of actual innocence,” Calderon v. Thompson, 523 U.S. 538, 558, 118 S.Ct. 1489, 140 L.Ed.2d 728 (1998). Surratt and the dissent both seek to untether § 2255(

e) from any actual innocence requirement. Instead, they would have us create a new rule contradicting Congress’ chief goal in enacting AEDPA — one that would apply to most any imaginable sentencing enhancement that we later believe could have erroneously increased a petitioner’s sentence. We see few limiting principles in that kind of rule, and take little comfort in Surratt’s assurances that it would apply in only a few cases (so long as he receives relief here).

Further, even before AEDPA, “the concern with finality served by the limitation on collateral attack ha[d] special force with respect to convictions,” like Sur-ratt’s, “based on guilty pleas.” United States v. Timmreck, 441 U.S. 780, 784, 99 S.Ct. 2085, 60 L.Ed.2d 634 (1979); see also United States v. Fugit, 703 F.3d 248, 252-53 (4th Cir.2012) (discussing the historical importance of finality, especially as to guilty pleas). Plea agreements work “only if dispositions by guilty plea are accorded a great measure of finality.” Blackledge v. Allison, 431 U.S. 63, 71, 97 S.Ct. 1621, 52 L.Ed.2d 136 (1977).

Thus, “a voluntary and intelligent plea of guilty made by an accused person, who has been advised by competent counsel, [usually] may not be collaterally attacked.” Mabry v. Johnson, 467 U.S. 504, 508, 104 S.Ct. 2543, 81 L.Ed.2d 437 (1984). This rule holds even when the law later changes in the petitioner’s favor, as “the possibility of a favorable change in the law occurring after a plea is one of the normal risks that accompan[y] a guilty plea.” United States v. Archie, 771 F.3d 217, 222 (4th Cir.2014).

Undeniably, “the Federal Government, no less than the States, has an interest in the finality of its criminal judgments.” United States v. Frady, 456 U.S. 152, 166, 102 S.Ct. 1584, 71 L.Ed.2d 816 (1982). For “[w]thout finality, the criminal law is deprived of much of its deterrent effect.” Teague v. Lane, 489 U.S. 288, 309, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989). “A procedural system which permits an endless repetition of inquiry into facts and law ... implies a lack of confidence about the possibilities of justice that cannot but war with the effectiveness of underlying substantive commands.” McCleskey v.

Zant, 499 U.S. 467, 492, 111 S.Ct. 1454, 113 L.Ed.2d 517 (1991); accord Custis v. United States, 511 U.S. 485, 497, 114 S.Ct. 1732, 128 L.Ed.2d 517 (1994) (“[I]nroads on the concept of finality tend to undermine confidence in the integrity of our procedures [.]”); see also Henry J. Friendly, Is Innocence Irrelevant? Collateral Attack on Criminal Judgments, 38 U. Chi. L.Rev. 142, 145 (1970) (“The proverbial man from Mars would surely think we must consider our system of criminal justice terribly bad if we are willing to tolerate such efforts at undoing judgments of conviction.”).

A waning of confidence would follow if we revisited convictions repeatedly, especially when the defendant conceded his guilt. It seems even odder to revisit convictions where, as here, the petitioner accepted an appeal waiver in his plea agreement. Even though the Government has chosen not to enforce that waiv-' er, its presence means that Surratt expected to have no later chance for any kind of review when he first agreed to plea guilty.

Although one might find it tempting to put finality concerns aside for the sake of self-designed notions of fairness, we have recognized that “there are dangers” in “subordinating] [finality] to the equities of the individual case,” “especially if so vague a term as ‘fairness’ is to be the touchstone.” In re Under Seal, 749 F.3d 276, 289 (4th Cir.2014).

Courts also have “never ... defined the scope of the writ [of habeas corpus] simply by reference to a perceived need to assure that an individual accused of crime is afforded a trial free of constitutional,” let alone statutory, “error.” Teague, 489 U.S. at 308, 109 S.Ct. 1060. No doubt, if we put finality aside when it suited us, we would invite a disquieting indeterminacy into our judicial system. And it’s hard to see where this “fairness” exception would end. So, “[w]e do not pause to consider whether a statute differently conceived and framed would yield results more consonant with fairness and reason.

We take the statute as we find it.” Anderson v. Wilson, 289 U.S. 20, 27, 53 S.Ct. 417, 77 L.Ed. 1004 (1933). Finality offers tangible benefits for the judicial system. At the very least, it ensures that court resources focus on the initial trial and appeal stage. In contrast, “[i]f every favorable precedential decision could become ... a ticket to being resen-tenced, the criminal justice system would need to continually marshal resources in order to keep in prison defendants whose trials and appeals and sentences conformed to then-existing constitutional and statutory standards.” Whiteside, 775 F.3d at 186.

Finality also provides closure to victims and the defendant: it assures the victim that his assailant will be punished, while it directs the defendant to move on with his life. And finality provides important incentives to litigants. If defendants know that they cannot raise arguments more than once, then they will exercise greater diligence and invoke whatever rights they may have early on. See Wainwright v. Sykes, 433 U.S. 72, 90, 97 S.Ct. 2497, 53 L.Ed.2d 594 (1977).

We decline to dispense with these many benefits — deriving from AEDPA’s central objective of finality — by accepting Sur-ratt’s view of § 2255(e). Y. Invoking the canon of constitutional avoidance, Surratt suggests that we must permit his petition because a failure to do so would present substantial constitutional concerns. Specifically, he argues that denying him relief under § 2255(

e) would raise questions related to due process, equal protection, separation-of-powers principles, and the Suspension Clause. We disagree. The canon of constitutional avoidance is an ill fit to this case. “The canon is a tool for choosing between competing plausible

interpretations of a provision.” Warger v. Shauers, - U.S. -, 135 S.Ct. 521, 529, 190 L.Ed.2d 422 (2014). “It has no application in the absence of ambiguity.” Id. As we have already explained, though, we see no ambiguity in § 2255(

e) when it comes to Surratt’s sentence. Further, Surratt’s attacks read like backdoor constitutional challenges to § 2255 itself. But here again, “[t]he canon of constitutional avoidance is not a method of adjudicating constitutional questions by other means.” United States v. Apel, — U.S. -, 134 S.Ct. 1144, 1153, 186 L.Ed.2d 75 (2014). Nor does avoidance grant a permission slip to “ ‘interpret’ statutes by gerrymandering them with a list of exceptions that happen to describe a party’s case,” id., which strikes us as just the sort of thing that Surratt means us to do.

Ultimately, though, “if the statute does not raise constitutional concerns, then there is no basis for employing the canon of constitutional avoidance.” Rodriguez v. Robbins, 715 F.3d 1127, 1140 (9th Cir.2013); accord United States v. Dwinells, 508 F.3d 63, 70-71 (1st Cir.2007); United States ex rel. Long v. SCS Bus. & Tech. Instit., Inc., 173 F.3d 870, 886 (D.C.Cir.1999). With no need for avoidance, we can rest assured that we have given § 2255(

e) the correct reading. A. We first conclude that our

interpretation of § 2255(

e) presents no color-able issue under the Suspension Clause. That provision of the Constitution states that “[t]he Privilege of the Writ of Habeas Corpus shall not be suspended, unless when in Cases of Rebellion or Invasion the public Safety may require it.” U.S. Const. art. I, § 9, cl.2. The Suspension Clause “ensures that, except during periods of formal suspension, the Judiciary will have a time-tested device, the writ, to maintain the delicate balance of governance that is itself the surest safeguard of liberty.” Boumediene, 553 U.S. at 745, 128 S.Ct. 2229.

As the Government observes, the Suspension Clause may not apply here at all. Neither the Supreme Court nor this Court has decided whether the Suspension Clause protects the writ as it existed in 1789 or the writ as it exists today. See Boumediene, 553 U.S. at 746, 128 S.Ct. 2229 (reserving the issue); St. Cyr, 533 U.S. at 300-01, 121 S.Ct. 2271 (same). Sitting en banp, we hinted that the Clause might very well protect only the former and not the latter. See Vial, 115 F.3d at 1197 n. 11.

If that proves to be the case (that is, if the Suspension Clause protects only the 1789 version), then Surratt’s Suspension Clause argument fails. See Morales, 499 F.3d at 670; Lindh v. Murphy, 96 F.3d 856, 867 (7th Cir.1996) (en banc), rev’d on other grounds, 521 U.S. 320, 117 S.Ct. 2059, 138 L.Ed.2d 481 (1997); United States v. Anselmi, 207 F.2d 312, 314 (3d Cir.1953). The original 1789 form of the writ did not afford post-conviction relief of the type sought here. See Ex Parte Yerger, 75 U.S. (8 Wall.) 85, 101, 19 L.Ed. 332 (1868); accord Brown v.

Allen, 344 U.S. 443, 533, 73 S.Ct. 397, 97 L.Ed. 469 (1953) (Jackson, J., concurring in result); LaGuerre v. Reno, 164 F.3d 1035, 1038 (7th Cir.1998). In the end, we need not settle the particular question of whether the Suspension Clause protects the writ in its 18th or 21st Century form. Our reading of § 2255(

e) sufficiently respects the writ in either event. “[T]he substitution of a collateral remedy which is neither inadequate nor ineffective to test, the legality of a person’s detention does not constitute a suspension of the writ of habeas corpus.” Swain, 430 U.S. at 381, 97 S.Ct. 1224. Here, § 2255 itself serves as the relevant “substitute.” And by holding that § 2255 presents an adequate and effective substitute means to test Surratt’s detention, we have likewise ensured that no Suspension Clause issue exists.

See Hayman, 342 U.S. at 223, 72 S.Ct. 263 (refusing to consider any constitutional attack on § 2255 because the presence of the savings clause avoids those issues); see also Medberry v. Crosby, 351 F.3d 1049, 1057 (11th Cir.2003) (“[Section 2255(e) ] avoided any serious question about whether the replacement of the writ of habeas corpus for a federal prisoner with the new § 2255 caused an unconstitutional suspension of the writ.”); Reyes-Requena, 243 F.3d at 901 n. 19 (noting that. § 2255(

e) ameliorates any Suspension Clause issues). As our preceding discussion should have made plain, Surratt had a full opportunity under § 2255 to make the claim that he makes here, but simply chose not to do so. Perhaps Surratt means to attack our reading of § 2255(

e) together with other limits found in § 2255, such as § 2255(h)’s limits on second or successive petitions. But that sort of attack would fare no better. Stressing that “judgments about the proper scope of the writ are normally for Congress to make,” the Supreme Court has held that § 2254’s analogous restrictions “do not amount to a ‘suspension’ of the writ.” Felker v. Turpin, 518 U.S. 651, 664, 116 S.Ct. 2338, 135 L.Ed.2d 827 (1996). “[Restrictions on successive petitions constitute a modified res judicata rule, a restraint on what is called in habeas corpus practice ‘abuse of the writ.’ ” Id.

In short, restrictions on multiple motions for post-conviction relief “amount[ ] to an entirely proper exercise of Congress’ judgment regarding the proper scope of the writ.” Vial, 115 F.3d at 1197. And “the reasoning of the [Supreme] Court with respect to limitations on second and successive habeas petitions pursuant to § 2254 applies with equal force to the identical language in § 2255.” Id. at 1198; accord Gilbert, 640 F.3d at 1317; United States v. Barrett, 178 F.3d 34, 53 (1st Cir.1999); see also Justin F.

Marceau, Challenging the Habeas Process Rather Than the Result, 69 Wash. & Lee L.Rev. 85, 144 n. 207 (2012) (“[T]he Suspension Clause is much less directly implicated by efforts to file multiple habeas petitions[.]”). Nor does a Suspension Clause problem develop whenever a prisoner is unable to take advantage of a later, favorable case. “Congress can, without offending the Suspension Clause, ... narrow the source of law cognizable on habeas review.” Green v. French, 143 F.3d 865, 876 (4th Cir.1998), abrogated on other grounds, Williams v. Taylor, 529 U.S. 362, 376-77, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000).

And the narrowing here flows from an oft-made distinction in the law “between actions previously taken and those yet to come.” Armour v. City of Indianapolis, Ind., — U.S. -, 132 S.Ct. 2073, 2082, 182 L.Ed.2d 998 (2012); cf. Solem v. Stumes, 465 U.S. 638, 642, 104 S.Ct. 1338, 79 L.Ed.2d 579 (1984) (“[Retroactive application is not compelled, constitutionally or otherwise.”); Wainwright v. Stone, 414 U.S. 21, 23-24, 94 S.Ct. 190, 38 L.Ed.2d 179 (1973) (holding that state prisoner was not constitutionally entitled to the benefit of a new

interpretation of a state criminal statute). Therefore, the Suspension Clause is not offended in this case. B. The approach yfe take to the savings clause also does not give rise to the sort of due process concerns identified in Hicks v. Oklahoma, 447 U.S. 343, 100 S.Ct. 2227, 65 L.Ed.2d 175 (1980). In Hicks, a state court jury was erroneously required to impose a sentence of forty years once they found the defendant guilty. Id. at 344-45, 100 S.Ct. 2227. On direct appeal, the state appellate court deemed the mandatory prison-term unconstitutional but declined to vacate the sentence. Id. at 345, 100 S.Ct. 2227.

The Supreme Court determined that decision to be a due process violation, as the state court’s unwillingness to correct the error deprived Hicks of a state-provided liberty interest in a jury determination. Id. at 346, 100 S.Ct. 2227. Given the somewhat unusual circumstances of the case, courts have read it to provide “a rather narrow rule.” Simpson v. Norris, 490 F.3d 1029, 1034 (8th Cir.2007). Several factors indicate that the due process concerns addressed in Hicks do not arise here. First, Hicks involved sentencing by a jury; this case does not.

Courts aggressively protect the jury’s “historic role” as “an intermediary between the State and criminal defendants.” Alleyne, 133 S.Ct. at 2161; see also Blakely v. Washington, 542 U.S. 296, 306, 124 S.Ct. 2531, 159 L.Ed.2d 403 (2004) (“Just as suffrage ensures the people’s ultimate control in the legislative and executive branches, jury trial is meant to ensure their control in the judiciary.”); Jones, 526 U.S. at 244-48, 119 S.Ct. 1215 (describing the historic development of the jury as a check on overly severe sentences).

The Framers, too, chose to emphasize the importance of the jury by separately enshrining its role in the Sixth Amendment. In other words, we show a special solicitousness towards claims that a mistake has tread upon the jury’s crucial role. Hicks reflects that concept. It would not therefore apply in cases, like this one, that do not implicate the jury right.

Indeed, the Supreme Court has narrowly read Hicks to hold “only that where state law creates for the defendant a liberty interest in having the jury make particular findings, the Due Process Clause implies that appellate findings do not suffice to protect that entitlement.” Cabana v. Bullock, 474 U.S. 376, 387 n. 4, 106 S.Ct. 689, 88 L.Ed.2d 704 (1986) (emphasis added), abrogated on other grounds by Pope v. Illinois, 481 U.S. 497, 503 n. 7, 107 S.Ct. 1918, 95 L.Ed

[…]