with whom Justice Stevens, Justice Souter, and Justice Ginsburg join, dissenting.
Because Hugo Roman Almendarez-Torres illegally reentered the United States after having been convicted of an aggravated felony, he was subject to a maximum possible sentence of 20 years’ imprisonment. See 8 U. S. C. § 1326(b)(2), Had he not been convicted of that felony, he would have been subject to a maximum of only two years. See 8 U. S. C. § 1326(a). The Court today holds that § 1326(b)(2) does not set forth a separate offense, and that conviction of a prior felony is merely a sentencing enhancement for the offense set forth in § 1326(a). This causes the Court to confront the difficult question whether the Constitution requires a fact which substantially increases the maximum permissible punishment for a crime to be treated as an element of that crime — to be charged in the indictment, and found beyond a reasonable doubt by a jury. Until the Court said so, it was far from obvious that the answer to this question was no; on the basis of our prior law, in fact, the answer was considerably doubtful.
In all our prior cases bearing upon the issue, however, we confronted a criminal statute or state-court criminal ruling *249that unambiguously relieved the prosecution of the burden of proving a critical fact to the jury beyond a reasonable doubt. In McMillan v. Pennsylvania, 477 U. S. 79 (1986), the statute provided that “ fvisibl[e] possession] [of] a firearm’ ” “‘shall not be an element of the crime,’” but shall be determined at sentencing by “ ‘[t]he court... by a preponderance of the evidence,’” id., at 81, n. 1 (quoting 42 Pa. Cons. Stat. § 9712 (1982)). In In re Winship, 397 U. S. 358 (1970), it provided that determinations of criminal action in juvenile cases “‘must be based on a preponderance of the evidence,’ ” id., at 360 (quoting N. Y. Family Court Act § 744(b)). In Patterson v. New York, 432 U. S. 197 (1977), the statute provided that extreme emotional disturbance “ ‘is an affirmative defense,’ ” id., at 198, n. 2 (quoting N. Y. Penal Law § 125.25 (McKinney 1975)). And in Mullaney v. Wilbur, 421 U. S. 684 (1975), Maine’s highest court had held that in murder cases malice aforethought was presumed and had to be negated by the defendant, id., at 689 (citing State v. Lafferty, 309 A. 2d 647 (1973)).
In contrast to the provisions involved in these cases, 8 U. S. C. § 1326 does not, on its face, place the constitutional issue before us: It does not say that subsection (b)(2) is merely a sentencing enhancement. The text of the statute supports, if it does not indeed demand, the conclusion that subsection (b)(2) is a separate offense that includes the violation described in subsection (a) but adds the additional element of prior felony conviction. I therefore do not reach the difficult constitutional issue in this case because I adopt, as I think oui’ eases require, that reasonable interpretation of § 1326 which avoids the problem. Illegal reentry simplic-iter (§ 1326(a)) and illegal reentry after conviction of an aggravated felony (§ 1326(b)(2)) are separate criminal offenses. Prior conviction of an aggravated felony being an element of the latter offense, it must be charged in the indictment. Since it was not, petitioner’s sentence must be set aside.
*250H
“[W]here a statute is susceptible of two constructions, by one of which grave and doubtful constitutional questions arise and by the other of which such questions are avoided, our duty is to adopt the latter.” United States ex rel. Attorney General v. Delaware & Hudson Co., 213 U. S. 366, 408 (1909). This ‘‘cardinal principle,” which “has for so long been applied by this Court that it is beyond debate,” Edward J. DeBartolo Corp. v. Florida Gulf Coast Building & Constr. Trades Council, 485 U. S. 568, 575 (1988), requires merely a determination of serious constitutional doubt, and not a determination of unconstitutionality. That must be so, of course, for otherwise the rule would “mea[n] that our duty is to first decide that a statute is unconstitutional and then proceed to hold that such ruling was unnecessary because the statute is susceptible of a meaning, which causes it not to be repugnant to the Constitution.” United States ex rel. Attorney General v. Delaware & Hudson Co., supra, at 408. The Court contends that neither of the two conditions for application of this rule is present here: that the constitutional question is not doubtful, and that the statute is not susceptible of a construction that will avoid it. I shall address the former point first.1
*251That it is genuinely doubtful whether the Constitution permits a judge (rather than a jury) to determine by a mere preponderance of the evidence (rather than beyond a reasonable doubt) a fact that increases the maximum penalty to which a criminal defendant is subject is clear enough from our prior cases resolving questions on the margins of this one. In In re Winship, supra, we invalidated a New York statute under which the burden of proof in a juvenile delinquency proceeding was reduced to proof by a preponderance of the evidence. We held that “the Due Process Clause protects the accused against conviction except upon proof beyond a reasonable doubt of every fact necessary to constitute the crime with which he is charged,” 397 U. S., at 364, and that the same protection extends to “a juvenile ... charged with an act which would constitute a crime if committed by an adult,” id., at 359.
Five years later, in Mullaney v. Wilbur, supra, we unanimously extended Winship’s protections to determinations that went not to a defendant’s guilt or innocence, but simply to the length of his sentence. We invalidated Maine’s homicide law, under which all intentional murders were presumed to be committed with malice aforethought (and, as such, were punishable by life imprisonment), unless the defendant could rebut this presumption with proof that he acted in the heat of passion (in which case the conviction would be reduced to manslaughter and the maximum sentence to 20 years). We acknowledged that “under Maine law these facts of intent [were] not general elements of the crime of felonious homieideC, but] [i]nstead, [bore] only on the appropriate punishment category.” 421 U. S., at 699. Nonetheless, we rejected this distinction between guilt and punishment. “[I]f Winship,” we said, “were limited to those facts that constitute a crime as defined by state law, a State could undermine many of the interests that decision sought to protect without effecting any substantive change in its law. ■ It would only be necessary to redefine the elements that constitute differ*252ent crimes, characterizing them as factors that bear solely on the extent of punishment.” Id., at 697-698.
In Patterson v. New York, we cut back on some of the broader implications of Mullaney. Although that case contained, we acknowledged, “some language... that ha[dj been understood as perhaps construing the Due Process Clause to require the prosecution to prove beyond a reasonable doubt any fact affecting 'the degree of criminal culpability/ ” we denied that we “intended]... such far-reaching effect.” 432 U. S., at 214-215, n. 15. Accordingly, we upheld in Patterson New York’s law casting upon the defendant the burden of proving as an “affirmative defense” to second-degree murder that he “'acted under the influence of extreme emotional disturbance for which there was a reasonable explanation or excuse,'” id., at 198-199, n. 2, which defense would reduce his crime to manslaughter. We explained that “[pjroof of the nonexistence of all affirmative defenses has never been constitutionally required,” id., at 210, and that the State need not “prove beyond a reasonable doubt every fact, the existence or nonexistence of which it is willing to recognize as an exculpatory or mitigating circumstance affeeting the degree of culpability or the severity of the punishment.” Id., at 207. We cautioned, however, that while our decision might “seem to permit state legislatures to reallocate burdens of proof by labeling as affirmative defenses at least some elements of the crimes now defined in their statutes[,]... there are obviously constitutional limits beyond which the States may not go in this regard.” Id., at 210.
Finally, and most recently, in McMillan v. Pennsylvania, 477 U. S., at 81, we upheld Pennsylvania’s Mandatory Minimum Sentencing Act, which prescribed a mandatory minimum sentence of five years upon a judge’s finding by a preponderance of the evidence that the defendant “visibly possessed a firearm” during the commission of certain enumerated offenses which all carried maximum sentences of *253more than five years. We observed that "we [had] never attempted to define precisely the constitutional limits noted in Patterson, i. e., the extent to which due process forbids the reallocation or reduction of burdens of proof in criminal cases,” but explained that, whatever those limits, Pennsylvania’s law did not transgress them, id., at 86, primarily because it “neither alter[ed] the maximum penalty for the crime committed nor ereate[d] a separate offense calling for a separate penalty; it operate[d] solely to limit the sentencing court’s discretion in selecting a penalty within the range already available to it without the special finding of visible possession of a firearm,” id., at 87-88.
The feebleness of the Court’s contention that here there is no serious constitutional doubt is evidenced by the degree to which it must ignore or distort the analysis of McMillan. As just described, that opinion emphasized — and emphasized repeatedly — that an increase of the maximum penalty was not at issue. Beyond that, it specifically acknowledged that the outcome might have been different (i. e., the statute might have been unconstitutional) if the maximum sentence had been affected:
“Petitioners’ claim that visible possession under the Pennsylvania statute is ‘really5 an element of the offenses for which they are being punished' — that Pennsylvania has in effect defined a new set of upgraded felonies — would have at least more superficial appeal if a finding of visible possession exposed them to greater or additional punishment, cf. 18 U. S. C. § 2118(d) (providing separate and greater punishment for bank robberies accomplished through ‘use of a dangerous weapon or device’), but it does not.” Id., at 88.
The opinion distinguished one of our own precedents on this very ground, noting that the Colorado Sex Offenders Act invalidated in Specht v. Patterson, 386 U. S. 605 (1967), increased a sex offender’s sentence from a 10-year maximum *254to an indefinite term up to and including life imprisonment. 477 U. S., at 88.
Despite all of that, the Court would have us believe that the present statute’s alteration of the maximum permissible sentence — which it acknowledges is “the major difference between this case and McMillan,” ante, at 244 — militates in favor of, rather than against, this statute’s constitutionality, because an increase of the minimum sentence (rather than the permissible maximum) is more disadvantageous to the defendant. Ibid. That is certainly an arguable position (it was argued, as the Court has the temerity to note, by the dissent in McMillan). But it is a position which McMillan not only rejected, but upon the converse of which McMillan rested its judgment.
In addition to inverting the consequence of this distinction (between statutes that prescribe a minimum sentence and those that increase the permissible maximum sentence) the Court seeks to minimize the importance of the distinction by characterizing it as merely one of five factors relied on in McMillan, and asserting that the other four factors here are the same. Ante, at 242-243. In fact, however, McMillan did not set forth any five-factor test; the Court selectively recruits “factors” from various parts of the discussion. Its first*factor, for example, that “‘the statute plainly does not transgress the limits expressly set.out in Patterson,’” ante, at 242, quoting McMillan, 477 U. S, at 86 — viz., that it does not “disear[d] the presumption of innocence” or “relieve the prosecution of its burden of proving guilt,” id., at 87 — merely narrows the issue to the one before the Court, rather than giving any clue to the resolution of that issue. It is no more a factor in solving the constitutional problem before us than is the observation that § 1326 is not an ex post facto law and does not effect an unreasonable search or seizure. The Court’s second, fourth, and part of its fifth “factors” are in fact all subparts of the crucial third factor (the one that is absent here), since they are all culled from the general dis*255cussion in McMillan of how the Pennsylvania statute simply limited a sentencing judge’s discretion. We said that, whereas in Mullaney the State had imposed “‘a differential in sentencing ranging from a nominal fine to a mandatory life sentence’” (the Court’s “second” faetor), Pennsylvania’s law “neither alter[ed] the maximum penalty for the crime committed [the Court’s‘third’ faetor] nor ereate[d] a separate offense calling for a separate penalty [the Court’s ‘fourth’ factor]; it operate[d] solely to limit the sentencing court’s discretion in selecting a penalty within the range already available to it without the special finding of visible possession of a firearm [the Court’s ‘third’ faetor]. . . . The statute gives no impression of having been tailored to permit the visible possession finding to be a tail which wags the dog of the substantive offense [part of the Court’s ‘fifth’ factor].” 477 U. S., at 87-88.
The Court’s recruitment of “factors” is, as I have said, selective. Omitted, for example, is McMillan’s statement that “petitioners do not contend that the particular faetor made relevant [by the statute]... has historically been treated ‘in the Anglo-American legal tradition’ as requiring proof beyond a reasonable doubt.” Id., at 90, quoting Patterson, 432 U. S., at 226. Petitioner does make such an assertion in the present case — correctly, as I shall discuss. But even with its selective harvesting, the Court is incorrect in its assertion that “most” of the “factors” it recites, ante, at 243 (and in its implication that all except the third of them) exist in the present ease as well. The second of them contrasted the consequence of the fact assumed in Mullaney (extension of the permissible sentence from as little as a nominal fine to as much as a mandatory life sentence) with the consequence of the fact at issue in McMillan (no extension of the permissible sentence at all, but merely a “limit[ation of] the sentencing court’s discretion in selecting a penalty within the range already available,” 477 U. S., at 88). The present case resembles Mullaney rather than McMillan in this regard, *256since the fact at issue increases the permissible sentence tenfold. And the only significant part of the fifth “factor”— that the statute in McMillan “'dictated the precise weight to be given [the statutory] factor,' ” ante, at 243, quoting McMillan, supra, at 89-90 — is likewise a point of difference and not of similarity.
But this parsing of various factors is really beside the point. No one can read our pre-McMillan eases, and especially Mullaney (whose limits were adverted to in Patterson but never precisely described), without entertaining a serious doubt as to whether the statute as interpreted by the Court in the present case is constitutional. And no one can read McMillan, our latest opinion on the point, without perceiving that the determinative element in our validation of the Pennsylvania statute was the fact that it merely limited the sentencing judge’s discretion within the range of penalty already available, rather than substantially increasing the available sentence. And even more than that: No one can read McMillan without learning that the Court was open to the argument that the Constitution requires a fact which does increase the available sentence to be treated as an element of the crime (such an argument, it said, would have “at least . . . superficial appeal,” 477 U. S., at 88). If all that were not enough, there must be added the fact that many State Supreme Courts have concluded that a prior conviction which increases maximum punishment must be treated as an element of the offense under either their State Constitutions, see, e. g., State v. McClay, 146 Me. 104, 112, 78 A. 2d 347, 352 (1951); Tuttle v. Commonwealth, 68 Mass. 505, 506 (1854) (prior conviction increasing maximum sentence must be set forth in indictment); State v. Furth, 5 Wash. 2d 1, 11-19, 104 P. 2d 925, 930-933 (1940); State ex rel. Lockmiller v. Mayo, 88 Fla. 96, 98-99, 101 So. 228, 229 (1924); Roberson v. State, 362 P. 2d 1115, 1118-1119 (Okla. Crim. App. 1961), or as a matter of common law, see, e. g., People ex rel. Cosgriff v. Craig, 195 N. Y. 190, 194-195, 88 N. E. 38, 39 (1909); People *257v. McDonald, 233 Mich. 98, 102, 105, 206 N. W. 516, 518, 519 (1925); State v. Smith, 129 Iowa 709, 710-715, 106 N.W. 187, 188-189 (1906) (“By the uniform current of authority, the fact of the prior convictions is to be taken as part of the offense instantly charged, at least to the extent of aggravating it and authorizing an increased punishment”); State v. Pennye, 102 Ariz. 207, 208-209, 427 P. 2d 525, 526-527 (1967); State v. Waterhouse, 209 Ore. 424, 428-433, 307 P. 2d 327, 329-331 (1957); Robbins v. State, 219 Ark. 376, 380-381, 242 S. W. 2d 640, 643 (1951); State v. Eichler, 248 Iowa 1267, 1270-1273, 83 N. W. 2d 576, 577-579 (1957).2
In the end, the Court cannot credibly argue that the question whether a fact which increases maximum permissible punishment must be found by a jury beyond a reasonable doubt is an easy one. That, perhaps, is why the Court stresses, and stresses repeatedly, the limited subject matter that § 1326(b) addresses — recidivism. It even tries, with utter lack of logic, to limit its rejection of the fair reading of McMillan to recidivism cases. “For the reasons just given,” it says, “and in light of the particular sentencing factor at issue in this case — recidivism—we should take *258McMillan’s statement [regarding the “superficial appeal” the defendant’s argument would have had if the factor at issue increased his maximum sentence] to mean no more than it said, and therefore not to make a determinative difference here.” Ante, at 245 (emphasis added). It is impossible to understand how McMillan could mean one thing in a later case where recidivism is at issue, and something else in a later ease where some other sentencing factor is at issue. One might say, of course, that recidivism should be an exception to the general rule set forth in McMillan — but that more forthright characterization would display how doubtful the constitutional question is in light of our prior case law.
In any event, there is no rational basis for making recidivism an exception. The Court is of the view that recidivism need not be proved to a jury beyond a reasonable doubt (a view that, as I shall discuss, is precisely contrary to the common-law tradition) because it “ ‘goes to the punishment only.’” It relies for this conclusion upon our opinion in Graham v. West Virginia, 224 U. S. 616 (1912). See ante, at 243, quoting Graham, supra, at 624; see also ante, at 247. The holding of Graham provides no support for the Court’s position. It upheld against due process and double jeopardy objections a state recidivism law under which a defendant’s prior convictions were charged and tried in a separate proceeding after he was convicted of the underlying offense. As the Court notes, ante, at 243, the prior convictions were not charged in the same indictment as the underlying offense; but they were charged in an “information” before the defendant was tried for the prior convictions, and, more importantly, the law explicitly preserved his right to a jury determination on the recidivism question. See Graham, supra, at 622-623; see also Oyler v. Boles, 368 U. S. 448, 453 (1962) (same). It is true, however, that if the basis for Graham’s holding were accepted, one would have to conclude that recidivism need not be tried to the jury and found beyond a reasonable doubt. The essence of Graham’s reason*259ing was that in the recidivism proceeding the defendant “was not held to answer for an offense,” 224 U. S., at 624, since the recidivism charge “ ‘goes to the punishment only/ ” ibid., quoting McDonald v. Massachusetts, 180 U. S. 311, 313 (1901).
But that basis for dispensing with the protections of jury trial and findings beyond a reasonable doubt was explicitly rejected in Mullaney, which accorded these protections to facts that were “not general elements of the crime of felonious homicide . . . [but bore] only on the appropriate punishment category,” 421 U. S., at 699. Whatever else Mulla-ney stands for, it certainly stands for the proposition that what Graham used as the line of demarcation for double jeopardy and some due process purposes (the matter “goes only to the punishment”) is not the line of demarcation for purposes of the right to jury trial and to proof beyond a reasonable doubt. So also does McMillan, which even while narrowing Mullaney made it very clear that the mere fact that a certain finding “goes only to the penalty” does not end the inquiry. The Court is certainly correct that the distinctive treatment of recidivism determinations for double jeopardy purposes takes some explaining; but it takes some explaining for the Court no less than for me. And the explanation assuredly is not (what the Court apparently suggests) that recidivism is never an element of the crime. It does much less violence to our jurisprudence, and to the traditional practice of requiring a jury finding of recidivism beyond a reasonable doubt, to explain Graham as a recidivism exception to the normal double jeopardy rule that conviction of a lesser included offense bars later trial for the greater erime. Our double jeopardy law, after all, is based upon traditional American and English practice, see United States v. Dixon, 509 U. S. 688, 704 (1993); United States v. Wilson, 420 U. S. 332, 339-344 (1975), and that practice has allowed recidivism to be charged and tried separately, see Spencer v. Texas, 385 U. S. 554, 566-567 (1967); Graham, supra, at 623, *260625-626, 631; McDonald, supra, at 312-313. It has not allowed recidivism to be determined by a judge as more likely than not.
While I have given many arguments supporting the position that the Constitution requires the recidivism finding in this ease to be made by a jury beyond a reasonable doubt, I do not endorse that position as necessarily correct. Indeed, that would defeat my whole purpose, which is to honor the practice of not deciding doubtfu.1 constitutional questions unnecessarily. What I have tried to establish — and all that I need to establish — is that on the basis of our jurisprudence to date, the answer to the constitutional question is not clear. It is the Court’s burden, on the other hand, to establish that its constitutional answer shines forth clearly from our eases. That burden simply cannot be sustained. I think it beyond question that there was, until today’s unnecessary resolution of the point, “serious doubt” whether the Constitution permits a defendant’s sentencing exposure to be increased tenfold on the basis of a fact that is not charged, tried to a jury, and found beyond a reasonable doubt. If the Court wishes to abandon the doctrine of constitutional doubt, it should do so forthrightly, rather than by declaring certainty on a point that is clouded in doubt.
II
The Court contends that the doctrine of constitutional doubt is also inapplicable because § 1326 is not fairly susceptible of the construction which avoids the constitutional problem — i e., the construction whereby subsection (b)(2) sets forth a separate criminal offense. Ante, at 238. The Court begins its statutory analysis not by examining the text of § 1326, but by demonstrating that the “subject matter [of the statute] — prior commission of a serious crime — is as typical a sentencing factor as one might imagine.” Ante, at 230. That is eminently demonstrable, sounds powerfully good, but in fact proves nothing at all. It is certainly true that a *261judge (whether or not bound by the Federal Sentencing Guidelines) is likely to sentence nearer the maximum permitted for the offense if the defendant is a repeat offender. But the same can be said of many, perhaps most, factors that are used to define aggravated offenses. For example, judges will “typically” sentence nearer the maximum that a statute allows if the crime of conviction is committed with a firearm, or in the course of another felony; but that in no way suggests that armed robbery and felony murder are sentencing enhancements rather than separate crimes.
The relevant question for present purposes is not whether prior felony conviction is “typically” used as a sentencing factor, but rather whether, in statutes that provide higher maximum sentences for crimes committed by convicted felons, prior conviction is “typically” treated as a mere sentence enhancement or rather as an element of a separate offense. The answer to that question is the latter. That was the rule at common law, and was the near-uniform practice among the States at the time of the most recent study I am aware of. See Note, Recidivist Procedures, 40 N. Y. U. L. Rev. 332, 333-334 (1965); Note, The Pleading and Proof of Prior Convictions in Habitual Criminal Prosecutions, 33 N. Y. U. L. Rev. 210, 215-216 (1958). At common law, the fact of prior convictions had to be charged in the same indictment charging the underlying crime, and submitted to the jury for determination along with that crime. See, e.g., Spencer v. Texas, supra, at 566; Massey v. United States, 281 F. 293, 297 (CA8 1922); Singer v. United States, 278 F. 415, 420 (CA3 1922); People v. Sickles, 156 N. Y. 541, 545, 51 N. E. 288, 289 (1898). While several States later altered this procedure by providing a separate proceeding for the determination of prior convictions, at least as late as 1965 all but eight retained the defendant’s right to a jury determination on this issue. See Note, 40 N. Y. U. L. Rev., at 333-334, 347. I am at a loss to explain the Court’s assertion that it has “found no statute that clearly makes recidivism an offense *262element” added to another crime, ante, at 230. There are many such.3
It is interesting that the Court drags the red herring of recidivism through both parts of its opinion — the “constitutional doubt” part and the “statutory interpretation” part alike. As just discussed, logic demonstrates that the nature of that charge (the fact that it is a “typical” sentencing factor) has nothing to do with what this statute means. And as discussed earlier, the text and reasoning of McMillan, and of the cases McMillan distinguishes, provide no basis for saying that recidivism is exempt from the Court’s clear acknowledgment that taking away from the jury facts that increase the maximum sentence is constitutionally questionable. One wonders what state courts, and lower federal courts, are supposed to do with today’s mysterious utterances. Are they to pursue logic, and conclude that all ambiguous statutes adding punishment for factors accompanying the principal offense are mere enhancements, or are they illogieally to give this special treatment only to recidivism? Are they to deem the reasoning of McMillan superseded for all eases, or does it remain an open and doubtful question, for all eases except those involving recidivism, whether statutory máximums can be increased without the benefit of jury trial? Whatever else one may say about today’s opinion, there is no doubt that it has brought to this area of the law more confusion than clarification.
Passing over the red herring, let me turn now to the statute at issue — §1326 as it stood when petitioner was con-*263vieted. The author of today’s opinion for the Court once agreed that the “language and structure” of this enactment “are subject to two plausible readings/’ one of them being that recidivism constitutes a separate offense. United States v. Forbes, 16 F. 3d 1294, 1298 (CA1 1994) (opinion of Coffin, J., joined by Breyer, C. J.).4 This would surely be enough to satisfy the- requirement expressed by Justice Holmes, see United States v. Jin Fuey Moy, 241 U. S. 394, 401 (1916), and approved by the Court, ante, at 237-238, that the constitutional-doubt-avoiding construction be “fairly possible.” Today, however, the Court relegates statutory language and structure to merely two of five “factors” that “help courts determine a statute’s objectives and thereby illuminate its text,” ante, at 228.
The statutory text reads, in relevant part, as follows:
“Reentry of deported alien; criminal penalties for reentry of certain deported aliens
“(a) Subject to subsection (b) of this section, any alien who [has been deported and thereafter reenters the United States]... shall be fined under title 18, or imprisoned not more than 2 years, or both.
“(b) Notwithstanding subsection (a) of this section, in the ease of any alien described in such subsection—
“(1) whose deportation was subsequent to a conviction for commission of three or more misdemeanors involving drugs, crimes against the person, or both, or a felony (other than an aggravated felony), such alien shall be fined under title 18, imprisoned not more than 10 years, or both; or
“(2) whose deportation was subsequent to a conviction for commission of an aggravated felony, such alien *264shall be fined under such title, imprisoned not more than 20 years, or both.” 8 U. S. C. § 1326(b).
One is struck at once by the parallel structure of subsections (a) and (b). Neither subsection says that the individual it describes “shall be guilty of a felony,” and both subsections say that the individuals they describe “shall be fined under title 18, or imprisoned not more than [2, 10, or 20] years.” If this suffices to define a substantive offense in subsection (a) (as all agree it does), it is hard to see why it would not define a substantive offense in each paragraph of subsection (b) as well. Cl, for example, 21 U. S. C. § 841, which has a subsection (a) entitled “Unlawful acts,” and a subsection (b) entitled “Penalties.”
The opening phrase of subsection (b) certainly does not indicate that what follows merely supplements or enhances the penalty provision of subsection (a); what follows is to apply “notwithstanding” all of subsection (a), 1 e., “in spite of” or “without prevention or obstruction from or by” subsection (a). See, e.g., Webster’s New International Dictionary 1669 (2d ed. 1949). The next phrase (“in the case of any alien described in . . . subsection [(a)]”) imports by reference the substantive acts attributed to the hypothetical alien (deportation and unauthorized reentry) in subsection (a). Significantly, this phrase does not apply subsection (b) to any alien “convicted under” subsection (a) — which is what one would expect if the provision was merely increasing the penalty for certain subsection (a) convictions. See, e. g., United States v. Davis, 801 F. 2d 754, 755-756 (CA5 1986) (noting that “predicat[ing] punishment upon conviction” of another offense is one of the “common indicia of sentence-enhancement provisions”). Instead, subsection (b) applies to an alien “described in” subsection (a) — one who has been deported and has reentered illegally. And finally, subsection (a)’s provision that it applies “[sjubject to subsection (b)” means that subsection (a) is inapplicable to an alien covered by subsection (b), just as subsection (b) applies “not*265withstanding” that the alien would otherwise be covered by-subsection (a).5
The Court relies on an earlier version of § 1826 to support its interpretation of the statute in its current form. Ante, at 232. While I agree that such statutory history is a legitimate tool of construction, the statutory history of §1326 does not support, but rather undermines, the Court’s interpretation. That earlier version contained a subsection (a) that, in addition to setting forth penalties (as did the sub-parts of subsection (b)), contained the phrase (which the subparts of subsection (b) did not) “shall be guilty of a felony, and upon conviction thereof....” With such a formulation, of course, it would be easier to conclude that subsection (a) defines the crime and sets forth the basic penalty, and subsection (b) sets forth merely penalty enhancements. But if that was what the additional language in subsection (a) of the 1988 statute connoted, then what was the elimination of that additional language (in the 1990 version of the statute at issue here) meant to achieve? See § 543(b)(3), 104 Stat. 5059. The more strongly the “shall be guilty of a felony” language suggests that subsection (b) of the 1988 statute contained only enhancements, the more strongly the otherwise inexplicable elimination of that language sug*266gests that subsection (b) of the 1990 statute was meant to be parallel with subsection (a) — i. e., that both subsections were meant to set forth not merely penalties but also offenses.6
After considering the subject matter and statutory language, the third factor the Court considers in arriving at its determination that this statute can only be read as a sentencing enhancement is the title of the 1988 amendment that added subsection (b)(2): “Criminal Penalties for Reentry of Certain Deported Aliens.” See § 7345, 102 Stat. 4471, cited ante, at 234. Of course, this title pertains to a subsection (b)(2) which, unlike the (b)(2) under which petitioner was convicted, was not parallel with the preceding subsection (a). But even disregarding that, the title of the amendment proves nothing at all. While “Criminal Penalties for Reentry” might normally be more suggestive of an enhancement than of a separate offense, there is good reason to believe it imports no such suggestion here. For the very next provision of the same enactment, which adjusts the substantive requirements for the crime of aiding and abetting the unlawful entry of an alien, is entitled “Criminal Penalties for Aiding or Assisting Certain Aliens to Enter the United States.” See § 7346, 102 Stat. 4471. Evidently, new substantive offenses that were penalized were simply entitled “Criminal Penalties” for the relevant offense. Moreover, *267the 1988 amendment kept the original title of § 1326 (“Reentry of Deported Alien”) intact, leaving it to apply to both subsection (a) and subsection (b). See § 7345, supra; § 276, 66 Stat. 229.
The Court’s fourth factor leading it to conclude that this statute cannot reasonably be construed as establishing substantive offenses is legislative history. See ante, at 234. It is, again, the legislative history of the provision as it existed in 1988, before subsection (a) was stripped of the language “shall be guilty of a felony,” thereby making subsections (a) and (b) parallel. Even so, it is of no help to the Court’s case. The stray statements that the Court culls from the Congressional Record prove only that the new subsection (b) was thought to increase penalties for unlawful reentry. But there is no dispute that it does that! The critical question is whether it does it by adding penalties to the subsection (a) offense, or by creating additional, more severely punished, offenses. That technical point is not alluded to in any of the remarks the Court recites.
The Court’s fifth and last argument in support of its interpretation of the statute is the contention that “the contrary interpretation . .. risks unfairness,” ibid., because it would require bringing the existence of the prior felony conviction to the attention of the jury. But it is also “unfair,” of course, to deprive the defendant of a jury determination (and a beyond-a-reasonable-doubt burden of proof) on the critical question of the prior conviction. This Court’s own assessment of which of those disadvantages is the greater can be of relevance here only insofar as we can presume that that perception would have been shared by the enacting Congress. We usually presume, however, not that an earlier Congress agreed with our current policy judgments, but rather that it agreed with the disposition provided by traditional practice or the common law. See United States v. Texas, 507 U. S. 529, 534 (1993); Astoria Fed. Sav. & Loan Assn. v. Solimino, 501 U. S. 104, 108 (1991); Norfolk Redevel*268opment and Housing Authority v. Chesapeake & Potomac Telephone Co. of Va., 464 U. S. 30, 35 (1983); Morissette v. United States, 342 U. S. 246, 263 (1952). As noted earlier, the Court’s hostility to jury determination of prior convictions is quite simply at odds with the manner in which recidivism laws have historically been treated in this country.
Moreover, even if we were free to resolve this matter according to our current views of what is fair, the Court’s judgment that avoiding jury “infection” is more important than affording a jury verdict (beyond a reasonable doubt) does not seem to me sound. The Court is not correct, to begin with, that the fact of prior conviction is “almost never contested,” ante, at 235, particularly in unlawful-entry cases. That is clear from the very legislative history of the present statute. Senator Chiles explained that “identifying and prosecuting . . . illegal alien felons is a long and complex process” because “[i]t is not uncommon for an alien who has committed a certain felony to pay his bond and walk, only to be apprehended for a similar crime in the next county but with a new name and identification.” 133 Cong. Rec. 8771 (1987). He went on to describe two specific aliens, one from whom police “seized 3 passports issued to him in 3 different names, 11 drivers licenses, immigration cards and numerous firearms and stolen property,” and the other on whom immigration officials had “5 alien files ... with 13 aliases, different birth dates and different social security cards.” Id., at 8771, 8772. He said that “these aliens [were] not exceptions but rather common amongst the 100,000 illegal alien felons in the United States.” Id., at 8772. Representative Smith stated that aliens arrested for felonies “often are able to pay expensive bonds and disappear under a new identity often to reappear in court with a different name and a new offense. In some eases, they may return to their native lands and reenter the United States with new names and papers but committing the same crimes.” Id., at 28840. And on the other side of the ledger, I doubt whether “infection” of the jury *269with knowledge of the prior crime is a serious problem. See, e. g., Spencer, 385 U. S., at 561 (“The defendants’ interests [in keeping prejudicial prior convictions from the jury] are protected by limiting instructions and by the discretion residing with the trial judge to limit or forbid the admission of particularly prejudicial evidence even though admissible under an accepted rule of evidence” (eitation omitted)); Old Chief v. United States, 519 U. S. 172, 191 (1997) (it is an abuse of discretion under Federal Rulé of Evidence 408 to disallow defendant’s stipulation to prior felony convictions where such convictions are an element of the offense); cf. Brief for National Association of Criminal Defense Lawyers as Amicus Curiae 30 (“In 1996, 98.2% of all Section 1326 defendants pleaded guilty”). If it is a problem, however, there are legislative and even judicial means for dealing with it, short of what today’s decision does: taking the matter away from the jury in all eases. See Note, 40 N. Y. U. L. Rev., at 333-334 (describing commonly used procedures under which defendant’s right to a jury is invoked only “[i]f [he] denies the existence of prior convictions or stands mute”); Spencer, supra, at 567 (describing the English rule, under which the indictment alleges both the substantive offense and prior conviction, but the jury is not charged on the prior conviction until after it convicts the defendant of the substantive offense).
In sum, I find none of the four nontextual factors relied upon by the Court to support its interpretation (“typicality” of recidivism as a sentencing factor; titles; legislative history; and risk of unfairness) persuasive. What does seem to me significant, however, is a related statutory provision, introduced by a 1996 amendment, which explicitly refers to subsection (b)(2) as setting forth “offenses.” See §334, 110 Stat. 3009-635 (instructing United States Sentencing Commission to amend sentencing guidelines “for offenses under . . . 1326(b)”). This later amendment can of course not cause subsection (b)(2) to have meant, at the time of petitioner’s conviction, something different from what it then *270said. But Congress's expressed understanding that subsection (b) creates separate offenses is surely evidence that it is “fairly possible” to read the provision that way.7
I emphasize (to conclude this part of the discussion) that “fairly possible” is all that needs to be established. The doctrine of constitutional doubt does not require that the problem-avoiding construction be the preferable one — the one the Court would adopt in any event. Such a standard would deprive the doctrine of all function. “Adopt the interpretation that avoids the constitutional doubt if that is the right one” produces precisely the same result as “adopt the right interpretation.” Rather, the doctrine of constitutional doubt comes into play when the statute is “susceptible of” the problem-avoiding interpretation, Delaware & Hudson Co., 213 U. S., at 408 — when that interpretation is reasonable, though not necessarily the best. I think it quite impossible to maintain that this standard is not met by the interpretation of subsection (b) which regards it as creating separate offenses.
* * *
For the foregoing reasons, I think we must interpret the statute before us here as establishing a separate offense rather than a sentence enhancement. It can be argued that, once the constitutional doubts that require this course have been resolved, statutes no less ambiguous than the one before us here will be interpretable as sentence enhancements, *271so that not much will have been achieved. That begs the question, of course, as to how the constitutional doubt will be resolved. Moreover, where the doctrine of constitutional doubt does not apply, the same result may be dictated by the rule of lenity, which would preserve rather than destroy the criminal defendant’s right to jury findings beyond a reasonable doubt. See, e. g., People ex rel. Cosgriff v. Craig, 195 N. Y., at 197, 88 N. E., at 40 (“It is unnecessary in this case to decide how great punishment the legislature may constitutionally authorize Courts of Special Sessions to impose on a conviction without a common-law jury. It is sufficient to say that in cases of doubtful construction or of conflicting statutory provisions, that interpretation should be given which best protects the rights of a person charged with an offense, to a trial according to the common law”). Whichever doctrine is applied for the purpose, it seems to me a sound principle that whenever Congress wishes a fact to increase the maximum sentence without altering the substantive offense, it must make that intention unambiguously clear. Accordingly, I would find that § 1826(b)(2) establishes a separate offense, and would reverse the judgment below.
The Court asserts that we have declined to apply the doctrine “in circumstances similar to those here — where a constitutional question, while lacking an obvious answer, does not lead a majority gravely to doubt that the statute is constitutional.” Ante, at 239. The cases it dtes, however, do not support this contention. In Rust v. Sullivan, 500 U. S. 173 (1991), the Court believed that “[t]here [was] no question but that the statutory prohibition... [was] constitutional,” id., at 192 (emphasis added). And in United States v. Locke, 471 U. S. 84 (1985), the Court found the doctrine inapplicable not because of lack of constitutional doubt, but because the statutory language did not permit an interpretation that would “avoid a constitutional question,” id., at 96. Similarly, in United States v. Monsanto, 491 U. S. 600 (1989), “the language of [the statute was] plain and unambiguous,” id., at 606.
It would not be, as the Court claims, “anomalous” to require jury trial for a factor increasing the maximum sentence, “in light of existing ease law that permits a judge, rather than a jury, to determine the existence of factors that can make a defendant eligible for the death penalty ... Ante, at 247, citing Walton v. Arizona, 497 U. S. 639 (1990); Hildwin v. Florida, 490 U. S. 638 (1989) (per curiam); and Spaziano v. Florida, 468 U. S. 447 (1984). Neither the cases dted, nor any other ease, permits a judge to determine the existence of a factor which makes a crime a capital offense. What the cited cases hold is that, once a jury has found the defendant guilty of all the elements of an offense which carries as its maximum penalty the sentence of death, it may be left to the judge to decide whether that maximum penalty, rather than a lesser one, ought to be imposed — even where that decision is constrained by a statutory requirement that certain “aggravating factors” must exist. The person who is charged with actions that expose him to the death penalty has an absolute entitlement to jury trial on all the elements of the charge.
For federal statutes of this sort, see, e.g., 15 U. S. C. § 1264(a), 18 U. S. C. § 924(e), and § 2114(a). In each of these provisions, recidivism is recited in a list of sentence-increasing aggravators that include, for example, intent to defraud or mislead (15 U. S. C. § 1264(a)), use of a firearm that is a machine gun, or a destructive device, or that is equipped with a silencer (18 U. S. C. § 924(c)), and wounding or threatening life with a dangerous weapon (§ 2114(a)). It would do violence to the text to treat recidivism as a mere enhancement while treating the parallel provisions as aggravated offenses, which they obviously are.
The statutory text at issue in Forbes was in all relevant respects identical to the statute before us here, except that the years of imprisonment for the offenses were less; they were increased by a 1994 amendment, see § 130001(b), 108 Stat. 2023.
The Court contends that treating subsection (b) as establishing substantive offenses renders the “notwithstanding” and “subject to” provisions redundant, because even without them our lesser included-offense jurisprudence would prevent a defendant from being convicted under both subsections (a) and (b). Ante, at 231. Kedundaney, however, consists of the annoying practice of saying the same thing twice, not the sensible practice of saying once, with clarity and conciseness, what the law provides. The author of today’s opinion once agreed that “[t]he fact that each subsection makes reference to the other is simply the logical way of indicating the relationship between the arguably two separate crimes.” United States v. Forbes, 16 F. 3d 1294, 1298 (CA1 1994). But if this be redundancy, it is redundancy that the Court’s alternative reading does not cure — unless one believes that, without the “notwithstanding” and “subject to” language, our interpretive jurisprudence would permit the subsection (a) penally to be added to the subsection (b) penalties.
Immediately after stressing the significance of the 1988 version of § 1326(a), the Court dismisses the 1990 amendment that eliminated the 1988 language upon which it relies, as a “housekeeping measure” by which “Congress [did not] inten[d] to change, or to clarify, the fundamental relationship between” subsections (a) and (b). Ante, at 234. The Court offers no support for this confident characterization, unless it is the mistaken assumption that statutory changes or clarifications unconfirmed by legislative history are inoperative. “Suffice it to say that legislative history need not confirm the details of changes in the law effected by statutory language before we will interpret that language according to its natural meaning.” Morales v. Trans World Airlines, Inc., 504 U. S. 374, 385, n. 2 (1992).
The Court is incorrect in its contention that the effective-date provision of the 1996 amendments reflects the opposite congressional understanding. See ante, at 237. That provision states that the amendments “apply under [subsection (b)] . . . only to violations of [subsection (a)],” occurring on or after the date of enactment. § 321(c), 110 Stat. 3009-628. There is no dispute, of course, that if subsection (b) creates separate offenses, one of the elements of the separate offenses is the lesser offense set forth in subsection (a). The quoted language is the clearest and simplest way of saying that that element of the subsection (b) offenses must have occurred after the date of enactment in order for the amendments to be applicable.