delivered the opinion of the Court.

The Armed Career Criminal Act of 1984, 18 U. S. C. § 924(e) (ACCA), raises the penalty for possession of a firearm by a felon from a maximum of 10 years in prison to a mandatory minimum sentence of 15 years and a maximum of life in prison without parole if the defendant “has three previous convictions ... for a violent felony or a serious drug offense.” We granted certiorari to determine whether a defendant in a federal sentencing proceeding may collaterally attack the validity of previous state convictions that are used to enhance his sentence under the ACCA. We hold that a defendant has no such right (with the sole exception of convictions obtained in violation of the right to counsel) to collaterally attack prior convictions.

Baltimore City Police arrested petitioner Darren J. Custis on July 1,1991. A federal grand jury indicted him on three counts: (1) possession of cocaine with intent to distribute in violation of 21 U. S. C. § 841(a)(1); (2) use of a firearm in connection with a drug trafficking offense in violation of 18 U. S. C. § 924(c); and (3) possession of a firearm by a convicted felon in violation of 18 U. S. C. § 922(g)(1). Before trial in the United States District Court for the District of Maryland, the Government notified Custis that it would seek an enhanced penalty for the § 922(g)(1) offense under § 924(e)(1). The notice charged that he had three prior felony convic*488tions: (1) a 1985 Pennsylvania state-court conviction for robbery; (2) a 1985 Maryland state-court conviction for burglary; and (3) a 1989 Maryland state-court conviction for attempted burglary.

The jury found Custis not guilty of possession with intent to distribute and not guilty of use of a firearm during a drug offense, but convicted him of possession of a firearm and simple cocaine possession, a lesser included offense in the charge of possession with intent to distribute cocaine. At the sentencing hearing, the Government moved to have Custis’ sentence enhanced under § 924(e)(1), based on the prior convictions included in the notice of sentence enhancement.

Custis challenged the use of the two Maryland convictions for sentence enhancement. He argued that his lawyer for his 1985 burglary conviction rendered unconstitutionally ineffective assistance and that his guilty plea was not knowing and intelligent as required by Boykin v. Alabama, 395 U. S. 238 (1969). He claimed that his attorney had failed to advise him of the defense of voluntary intoxication, and that he would have gone to trial, rather than pleaded guilty, had he been aware of that defense. He challenged his 1989 conviction on the ground that it had been based upon a “stipulated facts” trial. He claimed that such a “stipulated facts” trial was tantamount to a guilty plea and that his conviction was fundamentally unfair because he had not been adequately advised of his rights. Custis further asserts that he had been denied effective assistance of counsel in that case because the stipulated facts established only attempted breaking and entering rather than attempted burglary under state law.

The District Court initially rejected Custis’ collateral attacks on his two Maryland state-court convictions. The District Court’s letter ruling determined that the performance of Custis’ attorney in the 1985 case did not fall below the standard of professional competence required under Strickland v. Washington, 466 U. S. 668 (1984). Order in *489No. S 91-0334 (D. Md., Feb. 27, 1992), p. 1. It found that counsel’s recommendation of a guilty plea was not unreasonable under the circumstances. Id., at 2. The District Court also rejected Custis’ claim that the 1989 “stipulated facts” trial was the functional equivalent of a guilty plea. Id., at 2-3.

The District Court later reversed field and determined that it could not entertain Custis’ challenges to his prior convictions at all. It noted that “[ujnlike the statutory scheme for enhancement of sentences in drug cases, [§ 924(e)(1)] provides no statutory right to challenge prior convictions relied upon by the Government for enhancement.” 786 F. Supp. 533, 535-536 (Md. 1992). The District Court went on to state that the Constitution bars the use of a prior conviction for sentence enhancement only when there was a complete denial of counsel in the prior proceeding. Id., at 536, citing Gideon v. Wainwright, 372 U. S. 335 (1963); United States v. Tucker, 404 U. S. 443 (1972); and Burgett v. Texas, 389 U. S. 109 (1967). Based on Custis’ offense level of 33 and his criminal history category of VI, the District Court imposed a sentence of 235 months in prison.

The Court of Appeals affirmed. 988 F. 2d 1355 (CA4 1993).. It recognized the right of a defendant who had been completely deprived of counsel to assert a collateral attack on his prior convictions since such a defendant “has lost his ability to assert all his other constitutional rights.” Id., at 1360, citing Johnson v. Zerbst, 304 U. S. 458,465 (1938). Citing the “substantial burden” on prosecutors and the district courts, the Court of Appeals dismissed all of Custis’ challenges to his prior convictions as the “fact-intensive” type that pose a risk of unduly delaying and protracting the entire sentencing process. 988 F. 2d, at 1361. The prospect of such fact-intensive inquiries led it to express great reluctance at forcing district courts to overcome the “ 'inadequacy or unavailability of state court records and witnesses’” in trying to determine the validity of prior sentences. Ibid., *490quoting United States v. Jones, 977 F. 2d 105,109 (CA4 1992). In addition to the practical hurdles, the Court of Appeals specified concerns over comity and federalism as other factors weighing against permitting collateral attacks. “ ‘Federal courts are not forums in which to relitigate state trials.’ ” 988 F. 2d, at 1361, quoting Barefoot v. Estelle, 463 U. S. 880, 887 (1983). We granted certiorari, 510 U. S. 913 (1993), because the Court of Appeals’ decision conflicted with recent decisions from other Courts of Appeals that permitted defendants to challenge prior convictions that are used in sentencing under § 924(e)(1).1

Custis argues that the ACCA should be read to permit defendants to challenge the constitutionality of convictions used for sentencing purposes. Looking to the language of the statute, we do not believe § 924(e) authorizes such collateral attacks. The ACCA provides an enhanced sentence for any person who unlawfully possesses a firearm in violation of 18 U. S. C. § 922(g)2 and “has three previous convictions by any court referred to in section 922(g)(1) of this title for a violent felony or a serious drug offense . . . .” Section 924(e) applies whenever a defendant is found to have suffered “three previous convictions” of the type specified. The stat*491ute focuses on the fact of the conviction and nothing suggests that the prior final conviction may be subject to collateral attack for potential constitutional errors before it may be counted.

Absent specific statutory authorization, Custis contends that an implied right to challenge the constitutionality of prior convictions exists under § 924(e). Again we disagree. The Gun Control Act of 1968, of which § 924(e) is a part, strongly indicates that unchallenged prior convictions may be used for purposes of § 924(e). At least for prior violent felonies, §921(a)(20) describes the circumstances in which a prior conviction may be counted for sentencing purposes under § 924(e):

“What constitutes a conviction of ... a crime shall be determined in accordance with the law of the jurisdiction in which the proceedings were held. Any conviction which has been expunged, or set aside or for which a person has been pardoned or has had civil rights restored shall not be considered a conviction for purposes of this chapter [18 U. S. C. §§921-930].”

The provision that a court may not count a conviction “which has been ... set aside” creates a clear negative implication that courts may count a conviction that has not been set aside.

Congress’ passage of other related statutes that expressly permit repeat offenders to challenge prior convictions that are used for enhancement purposes supports this negative implication. For example, 21 U. S. C. § 851(c), which Congress enacted as part of the Comprehensive Drug Abuse Prevention and Control Act of 1970, sets forth specific procedures allowing a defendant to challenge the validity of a prior conviction used to enhance the sentence for a federal drug offense. Section 851(c)(1) states that “[i]f the person denies any allegation of the information of prior conviction, or claims that any conviction alleged is invalid, he shall file *492a written response to the information.” Section 851(c)(2) goes on to provide:

“A person claiming that a conviction alleged in the information was obtained in violation of the Constitution of the United States shall set forth his claim, and the factual basis therefor, with particularity in his response to the information. The person shall have the burden of proof by a preponderance of the evidence on any issue of fact raised by the response. Any challenge to a prior conviction, not raised by response to the information before an increased sentence is imposed in reliance thereon, shall be waived unless good cause be shown for failure to make a timely challenge.”

The language of § 851(c) shows that when Congress intended to authorize collateral attacks on prior convictions at the time of sentencing, it knew how to do so. Congress’ omission of similar language in § 924(e) indicates that it did not intend to give defendants the right to challenge the validity of prior convictions under this statute. Cf. Gozlon-Peretz v. United States, 498 U. S. 395, 404 (1991) (“ ‘[W]here Congress includes particular language in one section of a statute but omits it in another section of the same Act, it is generally presumed that Congress acts intentionally and purposely in the disparate inclusion or exclusion’”), quoting Russello v. United States, 464 U. S. 16, 23 (1983) (internal quotation marks omitted).

Our decision in Lewis v. United States, 445 U. S. 55 (1980), also supports the conclusion that prior convictions used for sentence enhancement purposes under § 924(e) are not subject to collateral attack in the sentence proceeding. Lewis interpreted 18 U. S. C. App. § 1202(a)(1) (1982 ed.), one of the predecessors to the current felon-in-possession-of-a-firearm statute. Section 1202(a)(1) was aimed at any person who “has been convicted by a court of the United States or of a State ... of a felony.” We concluded that “ ‘[njothing on the *493face of the statute suggests a congressional intent to limit its coverage to persons [whose convictions are not subject to collateral attack].’” 445 U. S., at 60, quoting United States v. Culbert, 435 U. S. 371,373. (1978). This lack of such intent in § 1202(a)(1) also contrasted with other federal statutes that explicitly permitted a defendant to challenge the validity or constitutionality of the predicate felony. See, e. g., 18 U. S. C. § 3575(e) (note following ch. 227) (dangerous special offender) and 21 U. S. C. § 851(c)(2) (recidivism under the Comprehensive Drug Abuse Prevention and Control Act of 1970). The absence of expressed intent, and the contrast with other federal statutes, led us to determine that “the firearms prosecution [under § 1202(a)(1)] does not open the predicate conviction to a new form of collateral attack.” 445 U. S., at 67.

Similarly, § 924(e) lacks any indication that Congress intended to permit collateral attacks on prior convictions used for sentence enhancement purposes. The contrast between § 924(e) and statutes that expressly provide avenues for collateral attacks, as well as our decision in Lewis, supra, point strongly to the conclusion that Congress did not intend to permit collateral attacks on prior convictions under § 924(e).

Custis argues that regardless of whether § 924(e) permits collateral challenges to prior convictions, the Constitution requires that they be allowed. He relies upon our decisions in Burgett v. Texas, 389 U. S. 109 (1967), and United States v. Tucker, 404 U. S. 443 (1972), in support of this argument. Both of these decisions relied upon our earlier decision in Gideon v. Wainwright, 372 U. S. 335 (1963), holding that the Sixth Amendment of the United States Constitution required that an indigent defendant in state-court proceedings have counsel appointed for him. Gideon, in turn, overruled our earlier decision in Betts v. Brady, 316 U. S. 455 (1942), which had held that the Sixth Amendment right to counsel, long applied in federal-court proceedings, was not itself made applicable to the States by the Due Process Clause. The *494Due Process Clause, Betts had held, required the appointment of counsel for an indigent defendant in state courts only upon a showing of special circumstances. Id., at 473.

But even before Betts v. Brady was decided, this Court had held that the failure to appoint counsel for an indigent defendant in a federal proceeding not only violated the Sixth Amendment, but was subject to collateral attack in federal habeas corpus. Johnson v. Zerbst, 304 U. S. 458 (1938). At a time when the underlying habeas statute was construed to allow collateral attacks on final judgments of conviction only where the rendering court lacked “jurisdiction” — albeit a somewhat expansive notion of “jurisdiction,” see Moore v. Dempsey, 261 U. S. 86 (1923) — this Court attributed a jurisdictional significance to the failure to appoint counsel. The Court said:

“If the accused, however, is not represented by counsel and has not competently and intelligently waived his constitutional right, the Sixth Amendment stands as a jurisdictional bar to a valid conviction and sentence depriving him of his life or his liberty. . . . The judgment of conviction pronounced by a court without jurisdiction is void, and one imprisoned thereunder may obtain release by habeas corpus.” 304 U. S., at 468.

When the Court later expanded the availability of federal habeas to other constitutional violations, it did so by frankly stating that the federal habeas statute made such relief available for them, without claiming that the denial of these constitutional rights by the trial court would have denied it jurisdiction. See, e. g., Waley v. Johnston, 316 U. S. 101, 104-105 (1942) (coerced confession); Brown v. Allen, 344 U. S. 443 (1953). There is thus a historical basis in our jurisprudence of collateral attacks for treating the right to have counsel appointed as unique, perhaps because of our oft-stated view that “[t]he right to be heard would be, in many *495cases, of little avail if it did not comprehend the right to be heard by counsel.” Powell v. Alabama, 287 U. S. 45, 68-69 (1932).

Following our decision in Gideon, the Court decided Burgett v. Texas, supra. There the defendant was charged under a Texas recidivist statute with having been the subject of four previous felony convictions. 389 U. S., at 111. The prosecutor introduced certified records of one of the defendant’s earlier convictions in Tennessee. Id., at 112. The defendant objected to the admission of this conviction on the ground that he had not been represented by counsel and had not waived his right to counsel, but his objection was overruled by the trial court. Id., at 113. This Court reversed, finding that the certified records of the Tennessee conviction on their face raised a “presumption that petitioner was denied his right to counsel. . . , and therefore that his conviction was void.” Id., at 114. The Court held that the admission of a prior criminal conviction that is constitutionally infirm under the standards of Gideon is inherently prejudicial and to permit use of such a tainted prior conviction for sentence enhancement would undermine the principle of Gideon. 389 U. S., at 115.

A similar situation arose in Tucker, supra. The defendant had been convicted of bank robbery in California in 1953. At sentencing, the District Court conducted an inquiry into the defendant’s background, and, the record shows, gave explicit attention to the three previous felony convictions that the defendant had acknowledged at trial. The District Court sentenced him to 25 years in prison — the stiffest term authorized by the applicable federal statute, 18 U. S. C. § 2113(d). 404 U. S., at 444. Several years later, after having obtained a judicial determination that two of his prior convictions were constitutionally invalid, the defendant filed a writ of habeas corpus in the District Court in which he had been convicted of bank robbery. He challenged the use at *496his 1953 bank robbery trial of his three previous felony convictions. This Court sustained his challenge insofar as his sentence was concerned, saying “Gideon . . . established an unequivocal rule ‘making it unconstitutional to try a person for a felony in a state court unless he had a lawyer or had validly waived one.’” Id., at 449, quoting Burgett v. Texas, supra, at 114. The Court held that “[e]rosion of the Gideon principle can be prevented here only by affirming the judgment of the Court of Appeals remanding this case to the trial court for reconsideration of the [defendant’s] sentence.” 404 U. S., at 449.

Custis invites us to extend the right to attack collaterally prior convictions used for sentence enhancement beyond the right to have appointed counsel established in Gideon. We decline to do so. We think that since the decision in Johnson v. Zerbst more than half a century ago, and running through our decisions in Burgett and Tucker, there has been a theme that failure to appoint counsel for an indigent defendant was a unique constitutional defect. Custis attacks his previous convictions claiming the denial of the effective assistance of counsel, that his guilty plea was not knowing and intelligent, and that he had. not been adequately advised of his rights in opting for a “stipulated facts” trial. None of these alleged constitutional violations rises to the level of a jurisdictional defect resulting from the failure to appoint counsel at all. Johnson v. Zerbst, supra.

Ease of administration also supports the distinction. As revealed in a number of the cases cited in this opinion, failure to appoint counsel at all will generally appear from the judgment roll itself, or from an accompanying minute order. But determination of claims of ineffective assistance of counsel, and failure to assure that a guilty plea was voluntary, would require sentencing courts to rummage through frequently nonexistent or difficult to obtain state-court transcripts or records that may date from another era, and may come from any one of the 50 States.

*497The interest in promoting the finality of judgments provides additional support for our constitutional conclusion. As we have explained, “[i]nroads on the concept of finality tend to undermine confidence in the integrity of our procedures” and inevitably delay and impair the orderly administration of justice. United States v. Addonizio, 442 U. S. 178, 184, n. 11 (1979). We later noted in Parke v. Raley, 506 U. S. 20 (1992), that principles of finality associated with habeas corpus actions apply with at least equal force when a defendant seeks to attack a previous conviction used for sentencing. By challenging the previous conviction, the defendant is asking a district court “to deprive [the] [state-court judgment] of [its] normal force and effect in a proceeding that ha[s] an independent purpose other than to overturn the prior judgmen[t].” Id., at 30. These principles bear extra weight in cases in which the prior convictions, such as one challenged by Custis, are based on guilty pleas, because when a guilty plea is at issue, “the concern with finality served by the limitation on collateral attack has special force.” United States v. Timmreck, 441 U. S. 780, 784 (1979) (footnote omitted).

We therefore hold that § 924(e) does not permit Custis to use the federal sentencing forum to gain review of his state convictions. Congress did not prescribe and the Constitution does not require such delay and protraction of the federal sentencing process. We recognize, however, as did the Court of Appeals, see 988 F. 2d, at 1363, that Custis, who was still “in custody” for purposes of his state convictions at the time of his federal sentencing under § 924(e), may attack his state sentences in Maryland or through federal habeas review. See Maleng v. Cook, 490 U. S. 488, 492 (1989)., If Custis is successful in attacking these state sentences, he may then apply for reopening of any federal sentence enhanced by the state sentences. We express no opinion on the appropriate disposition of such an application.

The judgment of the Court of Appeals is accordingly

Affirmed.

dissenting.

The Court answers a difficult constitutional question that I believe the underlying statute does not pose. Because in my judgment the Armed Career Criminal Act of 1984, 18 U. S. C. § 924(e) (ACCA), does not authorize sentence enhancement based on prior convictions that a defendant can show at sentencing to have been unlawfully obtained, I respectfully dissent.

I

A

The ACCA mandatory minimum sentence applies to defendants with “three previous convictions ... for a violent felony or a serious drug offense.” 18 U. S. C. § 924(e). The Court construes “convictio[n]” to refer to the “fact of the conviction,” ante, at 491 (emphasis in original), and concludes that “Congress did not intend to permit collateral attacks [during sentencing] on prior convictions under § 924(e),” ante, at 493.1 This interpretation of the ACCA will come as a surprise to the Courts of Appeals, which (with the one exception of the court below) have understood “convictio[n]” in the ACCA to mean “lawful conviction,” and have permitted defendants to show at sentencing that a prior conviction offered for enhancement was unconstitutionally obtained, whether as violative of the right to have appointed counsel, see Gideon v. Wainwright, 372 U. S. 335 (1963), the right to effective assistance of counsel, see Strickland v. Washington, 466 U. S. 668 (1984), the right against conviction based on an unknowing or involuntary guilty plea, see Boykin v. *499Alabama, 395 U. S. 238 (1969), or other constitutional rights.2 The weight of appellate authority, in my opinion, reflects the proper construction of the ACCA.

The Court’s contrary reading ignores the legal framework within which Congress drafted the ACCA, a framework with which we presume Congress was familiar. See, e. g., Cannon v. University of Chicago, 441 U. S. 677, 696-698 (1979). When the language that became the ACCA was first proposed in 1982, when it was enacted in 1984 (codified at § 1202(a)(1)) and when it was reenacted in 1986 (codified at § 924(e)), this Court’s decisions in Burgett v. Texas, 389 U. S. 109 (1967), and United States v. Tucker, 404 U. S. 443 (1972), were on the books. Even under the narrow reading the Court accords those decisions today, they recognize at least a right to raise during sentencing Gideon challenges to prior convictions used for enhancement. See ante, at 495-496. Unless Congress intended to snub that constitutional right (and we ordinarily indulge a “strong presumption . . . that Congress legislated in accordance with the Constitution,” Textile Workers v. Lincoln Mills of Ala., 353 U. S. 448, 477 (1957) (Frankfurter, J., dissenting)), “convictio[n]” in § 924(e) simply cannot refer to the mere fact of conviction, and the provision must have been meant to allow during sentencing at least some challenges to prior convictions offered for enhancement.

Nor is it likely that Congress’s intent was informed by as narrow a reading of Burgett and Tucker as the Court adopts *500today. In the legal environment of the ACCA’s enactment, Burgett and Tucker were thought to stand for the broader proposition that “[n]o consideration can be given [at sentencing] to a conviction that was unconstitutionally obtained,” 3 C. Wright, Federal Practice and Procedure §526, p. 102 (1982), and Courts of Appeals consistently read the decisions as requiring courts to entertain claims that prior convictions relied upon for enhancement were unconstitutional for reasons other than Gideon violations.3 The Congress that enacted the ACCA against this backdrop must be presumed to have intended to permit defendants to attempt to show at sentencing that prior convictions were “unconstitutionally obtained.”

That presumption is strongly bolstered by the fact that Congress, despite the consistent interpretation of the ACCA as permitting attacks on prior convictions during sentencing, and despite amending the law several times since its enactment (see note following 18 U. S. C. § 924 (1988 ed. and Supp. V) (listing amendments)), left the language relevant here untouched. Congress’s failure to express legislative disagreement with the appellate courts’ reading of the ACCA cannot be disregarded, especially since Congress has acted in this area in response to other Courts of Appeals decisions that it thought revealed statutory flaws requiring “correction].” S. Rep. No. 98-583, p. 7, and n. 17 (1984); see id., at 8, and n. 18, 14, and n. 31; see also Herman & MacLean v. Huddleston, 459 U. S. 375, 385-386 (1983) (“In light of [a] well-established judicial interpretation [of a statutory provision], Congress’ decision to leave [the provision] intact suggests that Congress *501ratified” the interpretation). Accordingly, absent clear indication that Congress intended to preclude all challenges during sentencing to prior convictions relied upon for enhancement, the ACCA must be read as permitting such challenges.

B

The Court fails to identify any language in the ACCA affirmatively precluding collateral attacks on prior convictions during sentencing, as there is none. Instead, the Court hears a clear message in the statutory silence, but I find none of its arguments persuasive. The Court first invokes 18 U. S. C. § 921(a)(20), under which a conviction “which has been expunged, or set aside or for which a person has been pardoned or has had civil rights restored shall not be considered a conviction for purposes of this chapter.” According to the Court, this “exemption clause” (as we have elsewhere called it, see Beecham v. United States, ante, at 369, “creates a clear negative implication that courts may count a conviction that has not been set aside,” ante, at 491. Expressio unius, in other words, est exclusio alterius.

Even if the premise of the Court’s argument is correct,4 the bridge the Court crosses to reach its conclusion is notoriously unreliable and does not bear the weight here. While “often a valuable servant,” the maxim that the inclusion of something negatively implies the exclusion of everything else {expressio unius, etc.) is “a dangerous master to follow in the construction of statutes.” Ford v. United States, 273 U. S. 593, 612 (1927) (internal quotation marks and citation omitted). It rests on the assumption that all omissions in *502legislative drafting are deliberate, an assumption we know to be false. See Posner, Statutory Interpretation — in the Classroom and in the Courtroom, 50 U. Chi. L. Rev. 800, 813 (1983); Radin, Statutory Interpretation, 43 Harv. L. Rev. 863, 873-874 (1930). As a result, “[sjcholars have long savaged the expressio canon,” Cheney R. Co. v. ICC, 902 F. 2d 66, 68 (CADC 1990) (Williams, J.), at least when it is made to do the work of a conclusive presumption, and our decisions support the proposition that “[sjometimes [the canon] applies and sometimes it does not, and whether it does or does not depends largely on context.” R. Dickerson, Interpretation and Application of Statutes 47 (1975); see also id., at 234-235.

In this case, the “contemporary legal context,” Cannon v. University of Chicago, 441 U. S., at 699, in which Congress drafted the ACC A requires rejecting the negative implication on which the Court relies. That context, as I have described, understood defendants to have a constitutional right to attack at sentencing prior convictions that had not previously been invalidated, and in that legal setting it would have been very odd for Congress to have intended to establish a constitutionally controversial rule by mere implication. See Lowe v. SEC, 472 U. S. 181, 206, n. 50 (1985) (“In areas where legislation might intrude on constitutional guarantees, we believe that Congress, which has always sworn to protect the Constitution, would err on the side of fundamental constitutional liberties when its legislation implicates those liberties”) (internal quotation marks and citation omitted). And in fact the legislative history indicates that quite a different intention informed the addition to §921(a)(20) in 1986, two years after the ACCA’s enactment, of the exemption clause (and the related “choice-of-law clause,” Beecham v. United States, ante, at 369. Congress simply intended to clarify that the law of the convicting jurisdiction should be the principal reference point in determining what counts as a “conviction” for purposes of the federal “felon in posses*503sion” law, and to correct an oversight that had resulted in the omission of exemption language from one of two parallel provisions. See S. Rep. No. 98-583, supra, at 7; H. R. Rep. No. 99-495, p. 20 (1986). In amending § 921(a)(20), Congress was not addressing the question of where, in the course of federal litigation, a conviction could be challenged. Indeed, the legislative history of the amendment reveals no hint of any intention at all with respect to § 924(e)’s sentence-enhancement provision, but rather an exclusive focus on the federal firearms disability in §922. Cf. Miles v. Illinois Central R. Co., 315 U. S. 698, 714-715 (1942) (Frankfurter, J., dissenting) (relying on legislative history to counter a negative implication from a statute’s text). As a result, the Court’s argument by negative implication from § 921(a)(20)’s exemption clause must fail. The fact that Congress in the exemption clause expressly precluded reliance upon unconstitutional convictions that have been set aside simply does not reveal an intent with respect to § 924(e) to require reliance at sentencing on unconstitutional convictions that have not yet been set aside.

The Court’s second statutory argument also seeks to establish congressional intent through negative implication, but is no more successful. The Court observes that Congress in other statutes expressly permitted challenges to prior convictions during sentencing, see ante, at 491-493 (citing 21 U. S. C. § 851(c)(2) and 18 U. S. C. § 3575(e)), which is said to show that “when Congress intended to authorize collateral attacks on prior convictions at the time of sentencing, it knew how to do so,” ante, at 492. But surely the Court does not believe that, if Congress intended to preclude collateral attacks on prior convictions at the time of sentencing, it did not know how to do that. And again, the Court’s effort to infér intent from the statutory silence runs afoul of the context of the statute’s enactment; within a legal framework forbidding sentencing on the basis of prior convictions a de*504fendant can show to be invalid, a Congress that intended to require sentencing on the basis of such convictions can be expected to have made its intention explicit.

Finally, the Court turns for support to Lewis v. United States, 445 U. S. 55 (1980), which held that the federal “felon in possession” law does not permit a defendant, during his prosecution, to challenge the constitutional validity of the predicate felony conviction. The Court’s reliance on Lewis, however, assumes an equivalence between two different types of laws that Lewis itself disclaimed: between a law disabling convicted felons from possessing firearms (at issue in Lewis), and a law requiring sentence enhancement based on prior convictions (at issue here, as well as in Burgett and Tucker). Lewis explained that the “felon in possession” law is “a sweeping prophylaxis” designed “to keep firearms away from potentially dangerous persons,” 445 U. S., at 63, 67, whereas a sentence-enhancement law “depend[s] upon the reliability of a past . . . conviction,” id., at 67. While the unlawfulness of a past conviction is irrelevant to the former, it is not to the latter, or so the Lewis Court thought in expressly distinguishing Burgett and Tucker: “[enforcement of [the federal gun disability] does not ‘support guilt or enhance punishment’ ... on the basis of a conviction that is unreliable.” 445 U. S., at 67 (quoting Burgett, 389 U. S., at 115).

Because of the material way in which a “felon in possession” law differs from a sentence-enhancement law, Burgett and Tucker were not part of the relevant legal backdrop against which Congress enacted the law interpreted in Lewis, and the Lewis Court could thus fairly presume that “conviction” in the statute before it was used as shorthand for “the fact of a felony conviction.” 445 U. S., at 60, 67. As Lewis itself recognized, however, Burgett and Tucker are part of the backdrop against which sentence-enhancement laws are enacted, and against that backdrop Congress must be presumed to have used “conviction” in § 924(e) to mean “lawful conviction,” and to have permitted defendants to *505show at sentencing that prior convictions offered for enhancement were unconstitutionally obtained.

II

A

Even if I thought the ACCA was ambiguous (the most the Court’s statutory arguments could establish), I would resolve the ambiguity in petitioner’s favor in accordance with the “‘cardinal principle’” of statutory construction that “‘this Court will first ascertain whether a construction of the statute is fairly possible by which [a constitutional] question may be avoided.’” Ashwander v. TVA, 297 U. S. 288, 348 (1936) (Brandéis, J., concurring) (quoting Crowell v. Benson, 285 U. S. 22, 62 (1932)); see also Edward J. DeBartolo Corp. v. Florida Gulf Coast Building & Constr. Trades Council, 485 U. S. 568, 575 (1988); NLRB v. Catholic Bishop of Chicago, 440 U. S. 490, 499-501, 504 (1979); Blodgett v. Holden, 275 U. S. 142, 148 (1927) (Holmes, J., concurring in result). The Ashwander principle, to be sure, comes into play only when the constitutional question to be avoided is a difficult one, but that designation easily fits the question that the Court’s reading of the ACCA requires it to decide, the question whether the Constitution permits courts to enhance a defendant’s sentence on the basis of a prior conviction the defendant can show was obtained in violation of his right to effective assistance of counsel, see Strickland v. Washington, 466 U. S. 668 (1984), or that the defendant can show was based on an unknowing or involuntary guilty plea, see Boy-kin v. Alabama, 395 U. S. 238 (1969).

This is a difficult question, for one thing, because the language and logic of Burgett and Tucker are hard to limit to claimed violations of the right, recognized in Gideon v. Wainwright, to have a lawyer appointed if necessary. As indicated by the uniformity of lower court decisions interpreting them, see supra, at 500, and n. 3, Burgett and Tucker are easily (if not best) read as announcing the broader principle *506that a sentence may not be enhanced by a conviction the defendant can show was obtained in violation of any “ ‘specific federal right’ ” (or, as Tucker put it, that a sentence may not be “founded [even] in part upon misinformation of constitutional magnitude,” 404 U. S., at 447) because to do so would be to allow the underlying right to be “denied anew” and to “suffer serious erosion,” Burgett, supra, at 116 (citation omitted); see also Tucker, supra, at 449. The Court’s references in both Burgett and Tucker to the right discussed in Gideon is hardly surprising; that was the “specific federal right” (and the record of the conviction obtained in violation of it the “misinformation of constitutional magnitude”) that the defendants before it invoked. The opinions in both cases, moreover, made it quite clear that the discussion of Gideon was not meant to supply a limitation. Burgett described Gideon not as unique but as “illustrative of the limitations which the Constitution places on state criminal procedures,” and it recounted as supportive of its holding cases involving coerced confessions, denials of the confrontation right, and illegal searches and seizures, 389 U. S., at 114; and Tucker made it clear that “the real question” before the Court was whether the defendant’s sentence might have been different if the sentencing judge had known that the defendant’s “previous convictions had been unconstitutionally obtained,” 404 U. S., at 448.5

*507Even if, consistently with principles of stare decisis, Burgett and Tucker could be read as applying only to some class of cases defined to exclude claimed violations of Strickland or Boykin, the question whether to confine them so is not easily answered for purposes of the Ashwander rule. Burgett and Tucker deal directly with claimed violations of Gideon, and distinguishing for these purposes between violations of Gideon and Strickland would describe a very fine line. To establish a violation of the Sixth Amendment under Strickland, a defendant must show that “counsel’s performance was deficient,” and that “the deficient performance prejudiced the defense” in that “counsel’s errors were so serious as to deprive the defendant of a fair trial, a trial whose result is reliable.” 466 U. S., at 687. It is hard‘to see how such a defendant is any better off than one who has been denied counsel altogether, and why the conviction of such a defendant may be used for sentence enhancement if the conviction of one who has been denied counsel altogether may not. The Sixth Amendment guarantees no mere formality of appointment, but the “assistance” of counsel, cf. Strickland, supra, at 685, 686 (“That a person who happens to be a lawyer is present at trial alongside the accused . . . is not enough to satisfy the [Sixth Amendment]” because “ ‘the right to counsel is the right to the effective assistance of counsel’”), and whether the violation is of Gideon or Strickland, the defendant has been denied that constitutional right.

It is also difficult to see why a sentencing court that must entertain a defendant’s claim that a prior conviction was obtained in violation of the Sixth Amendment’s right to counsel need not entertain a defendant’s claim that a prior conviction was based on an unknowing or involuntary guilty plea. *508That claim, if meritorious, would mean that the defendant was convicted despite invalid waivers of at least one of two Sixth Amendment rights (to trial by jury and to confront adverse witnesses) or of a Fifth Amendment right (against compulsory self-incrimination). See Boykin, 395 U. S., at 243. It is, to be sure, no simple task to prove that a guilty plea was the result of “[i]gnorance, incomprehension, coercion, terror, inducements, [or] subtle or blatant threats,” id., at 242-243, but it is certainly at least a difficult question whether a defendant who can make such a showing ought to receive less favorable treatment than the defendants in Burgett and Tucker.

Though the Court offers a theory for drawing a line between the right claimed to have been violated in Burgett and Tucker and the rights claimed to have been violated here, the Court’s theory is itself fraught with difficulty. In the Court’s view, the principle of Burgett and Tucker reaches only “constitutional violations ris[ing] to the level of a jurisdictional defect resulting from the failure to appoint counsel at all.” Ante, at 496 (citing Johnson v. Zerbst, 304 U. S. 458 (1938)). But nowhere in Burgett or Tucker is a distinction drawn between “jurisdictional” and “nonjurisdictional” rights, a fact giving no cause for surprise since long before (in Waley v. Johnston, 316 U. S. 101 (1942)) “the Court openly discarded the concept of jurisdiction — by then more a fiction than anything else — as a touchstone of the availability of federal habeas review.” Wainwright v. Sykes, 433 U. S. 72, 79 (1977). Nor was Johnson v. Zerbst, on which the Court today places much reliance, a ringing endorsement of a jurisdiction theory. For many years prior to that case, “the concept of jurisdiction ... was subjected to considerable strain,” Fay v. Noia, 372 U. S. 391, 450 (1963) (Harlan, J., dissenting), and Johnson v. Zerbst was actually the very last case to mention the idea, offering just “token deference to the old concept that the [habeas] writ could only reach jurisdictional defects,” Wechsler, Habeas Corpus and the *509Supreme Court: Reconsidering the Reach of the Great Writ, 59 U. Colo. L. Rev. 167,174 (1988).

In reviving the “jurisdiction” theory, the Court skips over the very difficulty that led to its abandonment, of devising a standard to tell whether or not a flaw in the proceedings leading to a conviction counts as a “jurisdictional defect.” “Once the concept of ‘jurisdiction’ is taken beyond the question of the court’s competence to deal with the class of offenses charged and the person of the prisoner” (as it must be if the concept is to reach Gideon violations) “it becomes a less than luminous beacon.” Bator, Finality in Criminal Law and Federal Habeas Corpus for State Prisoners, 76 Harv. L. Rev. 441, 470 (1963). Thus, if being denied appointed counsel is a “jurisdictional defect,” why not being denied effective counsel (treated as an equivalent in Strickland)? If a conviction obtained in violation of the right to have appointed counsel suffers from a “jurisdictional defect” because the right’s “purpose... is to protect an accused from conviction resulting from his own ignorance of his legal and constitutional rights,” Johnson v. Zerbst, supra, at 465, how distinguish a conviction based on a guilty plea resulting from a defendant’s own ignorance of his legal and constitutional rights?6 It was precisely due to the futility of providing principled answers to these questions that more than 50 years ago, and a quarter of a century before Burgett and Tucker, “[t]he Court finally abandoned the kissing of the jurisdictional book.” P. Bator, D. Meltzer, P. Mishkin, & D. Shapiro, Hart and Wechsler’s The Federal Courts and the Federal System 1502 (3d ed. 1988). The Court nevertheless finds itself compelled to reembrace the concept of “jurisdic*510tional defect,” fraught as it is with difficulties, in order to answer the constitutional question raised by its reading of the ACCA. Because it is “fairly possible,” Ashwander, 297 U. S., at 348, to construe the ACCA to avoid these difficulties and those associated with the other constitutional questions I have discussed, the Ashwander rule of restraint provides sufficient reason to reject the Court’s construction of the ACCA.

B

The rule of lenity, “which applies not only to interpretations of the substantive ambit of criminal prohibitions, but also to the penalties they impose,” Albernaz v. United States, 450 U. S. 333, 342 (1981), drives me to the same conclusion. Though lenity is usually invoked when there is doubt about whether a legislature has criminalized particular conduct, “[the] policy of lenity [also] means that the Court will not interpret a federal criminal statute so as to increase the penalty that it places on an individual when such an interpretation can be based on no more than a guess as to what Congress intended.” Ibid, (internal quotation marks and citation omitted); cf. Bell v. United States, 349 U. S. 81, 83 (1955) (“It may fairly be said to be a presupposition of our law to resolve doubts in the enforcement of a penal code against the imposition of a harsher punishment”). Because I “cannot say with assurance,” United States v. Granderson, ante, at 53, that Congress intended to require courts to enhance sentences on the basis of prior convictions a defendant can show to be invalid, the rule of lenity independently requires interpreting the ACCA to permit defendants to present such challenges to the sentencing judge before sentence is imposed.

C

The Court invokes “[e]ase of administration” to support its constitutional holding. Ante, at 496. While I doubt that even a powerful argument of administrative convenience *511would suffice to displace the Ashwander rule, cf. Stanley v. Illinois, 405 U. S. 645, 656 (1972), the burden argument here is not a strong one. The burdens of allowing defendants to challenge prior convictions at sentencing are not so severe, and are likely less severe than those associated with the alternative avenues for raising the very same claims.

For more than 20 years, as required by 21 U. S. C. §§ 851(c)(1) and (2), federal courts have entertained claims during sentencing under the drug laws that prior convictions offered for enhancement are “invalid” or were “obtained in violation of the Constitution,” the unamended statute reflecting a continuing congressional judgment that any associated administrative burdens are justified and tolerable. For almost a decade, federal courts have done the same under the ACCA, see n. 2, supra, again without congressional notice of any judicial burden thought to require relief. See also Parke v. Raley, 506 U. S., at 32 (“In recent years state courts have permitted various challenges to prior convictions” during sentencing). As against this, the Court sees administrative burdens arising because “sentencing courts [would be required] to rummage through frequently nonexistent or difficult to obtain state-court transcripts or records that may date from another era, and may come from any of the 50 States.” Ante, at 496. It would not be sentencing courts that would have to do this rummaging, however, but defendants seeking to avoid enhancement, for no one disagrees that the burden of showing the invalidity of prior convictions would rest on the defendants.

Whatever administrative benefits may flow from insulating sentencing courts from challenges to prior convictions will likely be offset by the administrative costs of the alternative means of raising the same claims. The Court acknowledges that an individual still in custody for a state conviction relied upon for enhancement may attack that conviction through state or federal habeas review and, if successful, “may . . . apply for reopening any federal sentence enhanced by the *512state sentences.” Ante, at 497. And the Court does not disturb uniform appellate case law holding that an individual serving an enhanced sentence may invoke federal habeas to reduce the sentence to the extent it was lengthened by a prior unconstitutional conviction. See J. Liebman & R. Hertz, Federal Habeas Corpus Practice and Procedure § 8.2, pp. 62-64, and n. 13.2, and §8.4, p. 89, n. 27 (1993 Supp.) (collecting cases).7 From the perspective of administrability, it strikes me as entirely sensible to resolve any challenges to the lawfulness of a predicate conviction in the single sentencing proceeding, especially since defendants there will normally be represented by counsel, who bring efficiency to the litigation (as well as equitable benefits).

Ill

Because I cannot agree that Congress has required federal courts to impose enhanced sentences on the basis of prior convictions a defendant can show to be constitutionally invalid, I respectfully dissent.