Justice Breyer

delivered the opinion of the Court

Subsection (a) of 8 U. S. C. § 1326 defines a crime. It forbids an alien who once was deported to return to the United States without special permission, and it authorizes a prison term of up to, but no more than, two years. Subsection (b)(2) of the same section authorizes a prison term of up to, but no more than, 20 years for “any alien described” in subsection (a), if the initial “deportation was subsequent to a conviction for commission of an aggravated felony.” The question before us is whether this latter provision defines a separate crime or simply authorizes an enhanced penalty. If the former, i. e., if it constitutes a separate crime, then the Government must write an indictment that mentions the additional element, namely, a prior aggravated felony conviction. If the latter, i. e., if the provision simply authorizes an enhanced sentence when an offender also has an earlier conviction, then the indictment need not mention that fact, for the fact of an earlier conviction is not an element of the present crime.

We conclude that the subsection is a penalty provision, which simply authorizes a court to increase the sentence for a recidivist. It does not define a separate crime. Consequently, neither the statute nor the Constitution requires the *227Government to charge the factor that it mentions, an earlier conviction, in the indictment.

I

In September 1995, a federal grand jury returned an indictment charging petitioner, Hugo Almendarez-Torres, with having been “found in the United States ... after being deported” without the “permission and consent of the Attorney General” in “violation of... Section 1326.” App. 3. In December 1995, Almendarez-Torres entered a plea of guilty. At a hearing, before the District Court accepted his plea, Almendarez-Torres admitted that he had been deported, that he had later unlawfully returned to the United States, and that the earlier deportation had taken place “pursuant to” three earlier “convictions” for aggravated felonies. Id., at 10-14.

In March 1996, the District Court held a sentencing hearing. Almendarez-Torres pointed out that an indictment must set forth all the elements of a crime. See Hamling v. United States, 418 U. S. 87, 117 (1974). He added that his indictment had not mentioned his earlier aggravated felony convictions. And he argued that, consequently, the court could not sentence him to more than two years imprisonment, the maximum authorized for an offender without an earlier conviction. The District Court rejected this argument. It found applicable a Sentencing Guideline range of 77 to 96 months, see United States Sentencing Commission, Guidelines Manual §2L1.2; ch. 5, pt. A (sentencing table) (Nov. 1995) (USSG), and it imposed a sentence of 85 months’ imprisonment. App. 17.

On appeal the Fifth Circuit also rejected petitioner’s argument. 113 F. 3d 515 (1996). Like seven other Circuits, it has held that subsection (b)(2) is a penalty provision that simply permits a sentencing judge to impose a higher sentence when the unlawfully returning alien also has a record of prior convictions. United States v. Vasquez-Olvera, 999 *228F. 2d 943, 945-947 (CA5 1993); see United States v. Forbes, 16 F. 3d 1294, 1297-1300 (CA1 1994); United States v. DeLeon-Rodriguez, 70 F. 3d 764, 765-767 (CA3 1995); United States v. Crawford, 18 F. 3d 1173, 1176-1178 (CA4 1994); United States v. Munoz-Cerna, 47 F. 3d 207, 210, n. 6 (CA7 1995); United States v. Haggerty, 85 F. 3d 403, 404-405 (CA8 1996); United States v. Valdez, 103 F. 3d 95, 97-98 (CA10 1996); United States v. Palacios-Casquete, 55 F. 3d 557, 559-560 (CA11 1995); cf. United States v. Cole, 32 F. 3d 16, 18-19 (CA2 1994) (reaching same result with respect to 8 U. S. C. § 1326(b)(1)). The Ninth Circuit, however, has reached the opposite conclusion. United States v. Gonzalez-Medina, 976 F. 2d 570, 572 (1992) (subsection (b)(2) constitutes separate crime). We granted certiorari to resolve this difference among the Circuits.

II

An indictment must set forth each element of the crime that it charges. Hamling v. United States, supra, at 117. But it need not set forth factors relevant only to the sentencing of an offender found guilty of the charged crime. Within limits, see McMillan v. Pennsylvania, 477 U. S. 79, 84-91 (1986), the question of which factors are which is normally a matter for Congress. See Staples v. United States, 511 U. S. 600, 604 (1994) (definition of a criminal offense entrusted to the legislature, “ ‘particularly in the case of federal crimes, which are solely creatures of statute’ ”) (quoting Liparota v. United States, 471 U. S. 419, 424 (1985)). We therefore look to the statute before us and ask what Congress intended. Did it intend the factor that the statute mentions, the prior aggravated felony conviction, to help define a separate crime? Or did it intend the presence of an earlier conviction as a sentencing factor, a factor that a sentencing court might use to increase punishment? In answering this question, we look to the statute’s language, structure, subject matter, context, and history — factors that typically help courts determine a statute’s objectives and thereby illuminate its text. *229See, e. g., United States v. Wells, 519 U. S. 482, 490-492 (1997); Garrett v. United States, 471 U. S. 773, 779 (1985).

The directly relevant portions of the statute as it existed at the time of petitioner’s conviction included subsection (a), which Congress had enacted in 1952, and subsection (b), which Congress added in 1988. See 8 U. S. C. § 1326 (1952 ed.), as enacted June 27, 1952, §276, 66 Stat. 229; 8 U. S. C. §1326 (1988 ed.) (reflecting amendments made by § 7345(a), 102 Stat. 4471). We print those portions of text below:

Ҥ1326. Reentry of deported alien; criminal penalties for reentry of certain deported aliens.

“(a) Subject to subsection (b) of this section, any alien who—

“(1) has been ... deported ..., and thereafter “(2) enters ..., or is at any time found in, the United States [without the Attorney General’s consent or the legal equivalent],

“shall be fined under title 18, or imprisoned not more than 2 years, or both.

“(b) Notwithstanding subsection (a) of this section, in the case of any alien described in such subsection— “(1) whose deportation was subsequent to a conviction for commission of [certain misdemeanors], 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 shall be fined under such title, imprisoned not more than 20 years, or both.” 8 U. S. C. § 1326.

A

Although the statute’s language forces a close reading of the text, as well as consideration of other interpretive circumstances, see Wells, supra, we believe that the answer *230to the question presented — whether Congress intended subsection (b)(2) to set forth a sentencing factor or a separate crime — is reasonably clear.

At the outset, we note that the relevant statutory subject matter is recidivism. That subject matter — prior commission of a serious crime — is as typical a sentencing factor as one might imagine. See, e.g., USSG §§4A1.1, 4A1.2 (Nov. 1997) (requiring sentencing judge to consider an offender’s prior record in every case); 28 U. S. C. § 994(h) (instructing Commission to write Guidelines that increase sentences dramatically for serious recidivists); 18 U. S. C. § 924(e) (Armed Career Criminal Act of 1984) (imposing significantly higher sentence for felon-in-possession violation by serious recidivists); 21 U. S. C. §§841(b)(1)(A)-(D) (same for drug distribution); United States Sentencing Commission, 1996 Source-book of Federal Sentencing Statistics 35, 49 (for year ending Sept. 30, 1996, 20.3% of all federal cases involved offenders with substantial criminal records (criminal history categories IV-VI); 44.2% of drug eases involved offenders with prior convictions). Perhaps reflecting this fact, the lower courts have almost uniformly interpreted statutes (that authorize higher sentences for recidivists) as setting forth sentencing factors, not as creating new crimes (at least where the conduct, in the absence of the recidivism, is independently unlawful). E. g., United States v. McGatha, 891 F. 2d 1520, 1525 (CA11 1990) (18 U.S.C. § 924(e)); United States v. Arango-Montoya, 61 F. 3d 1331, 1339 (CA7 1995) (21 U. S. C. § 841(b)); United States v. Jackson, 824 F. 2d 21, 25, and n. 6 (CADC 1987). And we have found no statute that clearly makes recidivism an offense element in such circumstances. But cf. 18 U. S. C. § 922(g)(1) (prior felony conviction an element but conduct not otherwise unlawful).

With recidivism as the subject matter in mind, we turn to the statute’s language. In essence, subsection (a) says that “any alien” once “deported,” who.reappears in the United *231States'without appropriate permission, shall be fined or “imprisoned not more than 2 years.” Subsection (b) says that “any alien described in” subsection (a), “whose deportation was subsequent to a conviction” for a minor, or for a major, crime, may be subject to a much longer prison term.

The statute includes the words “subject to subsection (b)” at the beginning of subsection (a), and the words “Notwithstanding subsection (a)” at the beginning of subsection (b). If Congress intended subsection (b) to set forth substantive crimes, in respect to which subsection (a) would define a lesser included offense, see Blockburger v. United States, 284 U. S. 299, 304 (1932), what are those words doing there? The dissent believes that the words mean that the substantive crime defined by “subsection (a) is inapplicable to an alien covered by subsection (b),” post, at 264, hence the words represent an effort to say that a defendant cannot be punished for both substantive crimes. But that is not what the words say. Nor has Congress ever (to our knowledge) used these or similar words anywhere else in the federal criminal code for such a purpose. See, e. g., 18 U. S. C. § 113 (aggravated and simple assault); §§ 1111, 1112 (murder and manslaughter); § 2113 (bank robbery and incidental crimes); §§ 2241, 2242 (aggravated and simple sexual abuse). And this should come as no surprise since, for at least 60 years, the federal courts have presumed that Congress does not intend for a defendant to be cumulatively punished for two crimes where one crime is a lesser included offense of the other. See Whalen v. United States, 445 U. S. 684, 691-693 (1980); Blockburger, supra.

If, however, Congress intended subsection (b) to provide additional penalties, the mystery disappears. The words “subject to subsection (b)” and “Notwithstanding subsection (a)” then are neither obscure nor pointless. They say, without obscurity, that the crime set forth in subsection (a), which both defines a crime and sets forth a penalty, is “sub*232ject to” subsection (b)’s different penalties (where the alien is also a felon or aggravated felon). And (b)’s higher maximum penalties may apply to an offender who violates (a) “notwithstanding” the fact that (a) sets forth a lesser penalty for one who has committed the same substantive crime. Nor is it pointless to specify that (b)’s punishments, not (a)’s punishment, apply whenever an offender commits (a)’s offense in a manner set forth by (b).

Moreover, the circumstances of subsection (b)’s adoption support this reading of the statutory text. We have examined the language of the statute in 1988, when Congress added the provision here at issue. That original language does not help petitioner. In 1988, the statute read as follows (with the 1988 amendment underscored):

Ҥ 1326. Reentry of deported alien; criminal penalties for reentry of certain deported aliens.

“(a) Subject to subsection (b) of this section, any alien who—

“(1) has been ... deported ..., and thereafter

“(2) enters ..., or is at any time found in, the United States [without the Attorney General’s consent or the legal equivalent],

“shall be guilty of a felony, and upon conviction thereof, be punished by imprisonment of not more than two years, or by a fine of not more than $1,000, or both.

“(b) Notwithstanding subsection (a) of this section, in the case of any alien described in such subsection—

“(1) whose deportation was subsequent to a conviction for commission of a felony (other than an aggravated felony), such alien shall be fined under title 18, imprisoned not more than 5 years, or both; or

“(2) whose deportation was subsequent to a conviction for commission of an aggravated felony, such alien shall be fined under such title, imprisoned not more than 15years, or both” 8 U. S. C. § 1326 (1988 ed.) (emphasis added).

*233Thus, at the time of the amendment, the operative language of subsection (a)’s ordinary reentering-alien provision said that a reentering alien “shall he guilty of a felony, and upon conviction thereof, be punished by imprisonment of not more than two years, or by a fine of not more than $1,000.” The 1988 amendment, subsection (b), by way of contrast, referred only to punishment — an increased punishment for the felon, or the aggravated felon, whom subsection (a) has “described.” Although one could read the language, “any alien described in [subsection (a)],” standing alone, as importing subsection (a)’s elements into new offenses defined in subsection (b), that reading seems both unusual and awkward when taken in context, for the reasons just given. Linguistically speaking, it seems more likely that Congress simply meant to “describe” an alien who, in the words of the 1988 statute, was “guilty of a felony” defined in subsection (a) and “convict[ed] thereof.”

As the dissent points out, post, at 265, Congress later struck from subsection (a) the words just quoted, and added in their place the words, “shall be fined under title 18, or imprisoned not more than two years.” See Immigration Act of 1990 (1990 Act), § 543(b)(8), 104 Stat. 5059. But this amendment was one of a series in the 1990 Act that uniformly updated and simplified the phrasing of various, disparate civil and criminal penalty provisions in the Immigration and Naturalization Act. See, e. g., 1990 Act, § 548(b)(1) (amending 8 U. S. C. § 1282(c)); § 543(b)(2)(C) (amending 8 U. S. C. § 1325); § 543(b)(4) (amending 8 U. S. C. § 1327); § 543(b)(5) (amending 8 U. S. C. § 1328). The section of the Act that contained the amendment is titled “Increase in Fine Levels; Authority of the INS to Collect Fines,” and the relevant subsection, simply “Criminal Fine Levels.” 1990 Act, § 543(b), 104 Stat. 5057, 5059. Although the 1990 amendment did have the effect of making the penalty provision in subsection (a) (which had remained unchanged since 1952) parallel with its counterparts in later enacted subsection (b), *234neither the amendment’s language, nor the legislative history of the 1990 Act, suggests that in this housekeeping measure, Congress intended to change, or to clarify, the fundamental relationship between the two subsections.

We also note that “the title of a statute and the heading of a section” are “tools available for the resolution of a doubt” about the meaning of a statute. Trainmen v. Baltimore & Ohio R. Co., 331 U. S. 519, 528-529 (1947); see also INS v. National Center for Immigrants’ Rights, Inc., 502 U. S. 183, 189 (1991). The title of the 1988 amendment is “Criminal penalties for reentry of certain deported aliens.” § 7345, 102 Stat. 4471 (emphasis added). A title that contains the word “penalties” more often, but certainly not always, see post, at 266-267, signals a provision that deals with penalties for a substantive crime.

In this instance the amendment’s title does not reflect careless, or mistaken, drafting, for the title is reinforced by a legislative history that speaks about, and only about, the creation of new penalties. See S. 973, 100th Cong., 1st Sess. (1987), 133 Cong. Rec. 8771 (1987) (original bill titled, “A bill to provide for additional criminal penalties for deported aliens who reenter the United States, and for other purposes”); 134 Cong. Rec. 27429 (1988) (section-by-seetion analysis referring to Senate bill as increasing penalties for unlawful reentry); id., at 27445 (remarks of Sen. D’Amato) (law would “increas [e] current penalties for illegal reentry after deportation”); id., at 27462 (remarks of Sen. Chiles) (law would “impose stiff penalties” against deported aliens previously convicted of drug offenses); 133 Cong. Rec. 28840-28841 (1987) (remarks of Rep. Smith) (corresponding House bill creates three-tier penalty structure). The history, to our knowledge, contains no language at all that indicates Congress intended to create a new substantive crime.

Finally, the contrary interpretation — a substantive criminal offense — risks unfairness. If subsection (b)(2) sets forth a separate crime, the Government would be required to *235prove to the jury that the defendant was previously deported “subsequent to a conviction for commission of an aggravated felony.” As this Court has long recognized, the introduction of evidence of a defendant’s prior crimes risks significant prejudice. See, e.g., Spencer v. Texas, 385 U. S. 554, 560 (1967) (evidence of prior crimes “is generally recognized to have potentiality for prejudice”). Even if a defendant’s stipulation were to keep the name and details of the previous offense from the jury, see Old Chief v. United States, 519 U. S. 172, 178-179 (1997), jurors would still learn, from the indictment, the judge, or the prosecutor, that the defendant had committed an aggravated felony. And, as we said last Term, “there can be no question that evidence of the . . . nature of the prior offense,” here, that it was “aggravated” or serious, “carries a risk of unfair prejudice to the defendant.” Id., at 185 (emphasis added). Like several lower courts, we do not believe, other things being equal, that Congress would have wanted to create this kind of unfairness in respect to facts that are almost never contested. See, e. g., United States v. Forbes, 16 F. 3d, at 1298-1300; United States v. Rumney, 867 F. 2d 714, 718-719 (CA1 1989); United States v. Brewer, 853 F. 2d 1319, 1324-1325 (CA6 1988) (en banc); United States v. Jackson, 824 F. 2d, at 25-26; Government of Virgin Islands v. Castillo, 550 F. 2d 850, 854 (CA3 1977).

In sum, we believe that Congress intended to set forth a sentencing factor in subsection (b)(2) and not a separate criminal offense.

B

We must also consider several additional arguments that have been or might be made for a contrary interpretation of the statute. First, one might try to derive a congressional intent to establish a separate crime from the magnitude of the increase in the maximum authorized sentence. The magnitude of the change that Congress made in 1988, however, proves little. That change — from a 2-year maximum to 5- and 15-year máximums — is well within the range *236set forth in other statutes that the lower courts have generally interpreted as providing for sentencing enhancements. Compare 8 U. S. C. § 1326 (1988 ed.) with 21 U. S. C. §§ 841(b)(1)(B) and (D) (distributing less than 50 kilograms of marijuana, maximum 5 years; distributing 100 or more kilograms of marijuana, 5 to 40 years), §§ 841(b)(1)(A) and (C) (distributing less than 100 grams of heroin, maximum 20 years; distributing 1 kilogram or more of heroin, maximum of life imprisonment), § 841(b)(1)(B) (distributing 500 grams or more of cocaine, 5 to 40 years; same, with prior drug felony conviction, 10 years to life); §962 (doubling maximum term for second and subsequent violations of drug importation laws); 18 U. S. C. § 844 (using or carrying explosive device during commission of felony, maximum 10 years; subsequent offense, maximum 20 years); § 2241(c) (sexual abuse of children, maximum life; second offense, mandatory life); § 2320(a) (trafficking in counterfeit goods, maximum 10 years; subsequent offense, maximum 20 years). Congress later amended the statute, increasing the máximums to 10 and to 20 years, respectively. "Violent Crime Control and Law Enforcement Act of 1994, §§ 130001(b)(1)(B) and (b)(2), 108 Stat. 2023. But nothing suggests that, in doing so, Congress intended to transform that statute’s basic nature. And the later limits are close to the range suggested by other statutes regardless.

Second, petitioner and the dissent point, in part, to statutory language that did not exist when petitioner was convicted in 1995. Petitioner, for example, points out that in 1996, Congress added two new subsections, (b)(3) and (b)(4), which, petitioner says, created new substantive crimes. See Antiterrorism and Effective Death Penalty Act of 1996, § 401(c), 110 Stat. 1267 (adding subsection (b)(3)); Illegal Immigration Reform and Immigrant Responsibility Act of 1996 (IIRIRA), § 305(b), 110 Stat. 3009-606 to 3009-607 (adding subsection (b)(4)). Both petitioner and the dissent also refer to another 1996 statutory provision in which Congress used *237the word “offense” to refer to the subsection now before us. See IIRIRA, §834,110 Stat. 3009-635.

These later enacted laws, however, are beside the point. They do not declare the meaning of earlier law. Cf. Federal Housing Administration v. Darlington, Inc., 358 U. S. 84, 90 (1958). They do not seek to clarify an earlier enacted general term. Cf. Red Lion Broadcasting Co. v. FCC, 395 U. S. 367, 380-381 (1969). They do not depend for their effectiveness upon clarification, or a change in the meaning of an earlier statute. Cf. Seatrain Shipbuilding Corp. v. Shell Oil Co., 444 U. S. 572, 595-596 (1980). They do not reflect any direct focus by Congress upon the meaning of the earlier enacted provisions. Cf. ibid.; Darlington, supra, at 86. Consequently, we do not find in them any forward looking legislative mandate, guidance, or direct suggestion about how courts should interpret the earlier provisions.

Regardless, it is not obvious that the two new subsections to which petitioner points create new crimes (a matter on which we express no view) nor, in adding them, did Congress do more than leave the legal question here at issue where it found it. The fact that Congress used a technical, crime-suggesting word — “offense”—eight years later in a different, and minor, statutory provision proves nothing — not least because it is more than offset by different words in the same later statute that suggest with greater force the exact opposite, namely, the precise interpretation of the relation of subsection (b) to subsection (a) that we adopt. See IIRIRA, § 321(e), 110 Stat. 3009-628 (stating that a new definition of “aggravated felony” applies “under" subsection (b) “only to violations” of subsection (a)).

Finally, petitioner and the dissent argue that the doctrine of “constitutional doubt” requires us to interpret subsection (b)(2) as setting forth a separate crime. As Justice Holmes said long ago: “A statute must be construed, if fairly possible, so as to avoid not only the conclusion that it is unconstitutional but also grave doubts upon that score.” United *238 States v. Jin Fuey Moy, 241 U. S. 394, 401 (1916) (citing United States ex rel. Attorney General v. Delaware & Hudson Co., 213 U. S. 366, 408 (1909)); see also Ashwander v. TVA, 297 U. S. 288, 348 (1936) (Brandeis, J., concurring). “This canon is followed out of inspect for Congress, which we assume legislates in the light of constitutional limitations.” Rust v. Sullivan, 500 U. S. 173, 191 (1991); see also FTC v. American Tobacco Co., 264 U. S. 298, 305-307 (1924). The doctrine seeks in part to minimize disagreement between the branches by preserving congressional enactments that might otherwise founder on constitutional objections. It is not designed to aggravate that friction by creating (through the power of precedent) statutes foreign to those Congress intended, simply through fear of a constitutional difficulty that, upon analysis, will evaporate. Thus, those who invoke the doctrine must believe that the alternative is a serious likelihood that the statute will be held unconstitutional. Only then will the doctrine serve its basic democratic function of maintaining a set of statutes that reflect, rather than distort, the policy choices that elected representatives have made. For similar reasons, the statute must be genuinely susceptible to two constructions after, and not before, its complexities are unraveled. Only then is the statutory construction that avoids the constitutional question a “fair” one.

Unlike the dissent, we do not believe these conditions are met in the present case. The statutory language is somewhat complex. But after considering the matter in context, we believe the interpretative circumstances point significantly in one direction. More important, even if we were to assume that petitioner’s construction of the statute is “fairly possible,” Jin Fuey Moy, supra, at 401, the constitutional questions he raises, while requiring discussion, simply do not lead us to doubt gravely that Congress may authorize courts to impose longer sentences upon recidivists who commit a particular crime. The fact that we, unlike the dissent, do *239not gravely doubt the statute’s constitutionality in this respect is a crucial point. That is because the “constitutional doubt” doctrine does not apply mechanically whenever there arises a significant constitutional question the answer to which is not obvious. And precedent makes clear that the Court need not apply (for it has not always applied) 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. See, e. g., Rust, 500 U. S., at 190-191 (declining to apply doctrine although petitioner’s constitutional claims not “without some force”); id., at 204-207 (Blackmun, J., dissenting); United States v. Monsanto, 491 U. S. 600, 611 (1989); id., at 636 (Blackmun, J., dissenting); United States v. Locke, 471 U. S. 84, 95 (1985); id., at 120 (Stevens, J., dissenting).

HH HH b-H

Invoking several of the Court’s precedents, petitioner claims that the Constitution requires Congress to treat recidivism as an element of the offense — irrespective of Congress’ contrary intent. Moreover, petitioner says, that requirement carries with it three subsidiary requirements that the Constitution mandates in respect to ordinary, legislatively intended, elements of crimes. The indictment must state the “element.” See, e.g., Hamling v. United States, 418 U. S., at 117. The Government must prove that “element” to a jury. See, e. g., Duncan v. Louisiana, 391 U. S. 145, 149 (1968). And the Government must prove the “element” beyond a reasonable doubt. See, e.g., Patterson v. New York, 432 U. S. 197, 210 (1977). We cannot find sufficient support, however, in our precedents or elsewhere, for petitioner’s claim.

This Court has explicitly held that the Constitution’s 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.” *240 In re Winship, 397 U. S. 358, 364 (1970). But Winship, the ease in which the Court set forth this proposition of constitutional law, does not decide this case. It said that the Constitution entitles juveniles, like adults, to the benefit of proof beyond a reasonable doubt in respect to the elements of the crime. It did not consider whether, or when, the Constitution requires the Government to treat a particular fact as an element, i. e., as a “fact necessary to constitute the crime,” even where the crime-defining statute does not do so.

Mullaney v. Wilbur, 421 U. S. 684 (1975), provides petitioner with stronger support. The Court there struck down a state homicide statute under which the State presumed that all homicides were committed with “malice,” punishable by life imprisonment, unless the defendant proved that he had acted in the heat of passion. Id., at 688. The Court wrote that “if Winship 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” just by redefining “the elements that constituted] different crimes, characterising them as factors that bear solely on the extent of punishment.” Id., at 698. It simultaneously held that the prosecution must establish “beyond a reasonable doubt” the nonexistence of “heat of passion” — the fact that, under the State’s statutory scheme, distinguished a homicide punishable by a life sentence from a homicide punishable by a maximum of 20 years. Id., at 704. Read literally, this language, we concede, suggests that Congress cannot permit judges to increase a sentence in light of recidivism, or any other factor, not set forth in an indictment and proved to a jury beyond a reasonable doubt.

This Court’s later case, Patterson v. New York, supra, however, makes absolutely clear that such a reading of Mullaney is wrong. The Court, in Patterson, pointed out that the State in Mullaney made the critical fact — the absence of “heat of passion” — not simply a potential sentencing factor, but also a critical part of the definition of “malice afore*241thought,” which was itself in turn “part of” the statute’s definition of “homicide,” the crime in question. Patterson, 432 U. S., at 215-216. (The Maine Supreme Court, in defining the crime, had said that “malice” was “presumed” unless “rebutted” by the defendant’s showing of “heat of passion.” Id., at 216.) The Court found this circumstance extremely important. It said that Mullaney had considered (and held “impermissible”) the shifting of a burden of proof “with respect to a fact which the State deems so important that it must be either proved or presumed.” 432 U. S., at 215 (emphasis added). And the Court then held that similar burden shifting was permissible with respect to New York’s homicide-related sentencing factor “extreme emotional disturbance.” Id., at 205-206. That factor was not a factor that the state statute had deemed “so important” in relation to the crime that it must be either “proved or presumed.” Id., at 205-206, 215.

The upshot is that Mullaney's language, if read literally, suggests that the Constitution requires that most, if not all, sentencing factors be treated as elements. But Patterson suggests the exact opposite, namely, that the Constitution requires scarcely any sentencing factors to be treated in that way. The cases, taken together, cannot significantly help petitioner, for the statute here involves a sentencing factor— the prior commission of an aggravated felony — that is neither “presumed” to be present, nor need be “proved” to be present, in order to prove the commission of the relevant crime. See 8 U. S. C. § 1326(a) (defining offense elements). Indeed, as we have said, it involves one of the most frequently found factors that affects sentencing — recidivism.

Nor does Specht v. Patterson, 386 U. S. 605 (1967), which petitioner cites, provide significant additional help, for Specht was decided before Patterson (indeed before Win-ship); it did not consider the kind of matter here at issue; and, as this Court later noted, the Colorado defendant in Specht was “confronted with ‘a radically different situation’ *242from the usual sentencing proceeding.” McMillan v. Pennsylvania, 477 U. S., at 89. At most, petitioner might read all these cases, taken together, for the broad proposition that sometimes the Constitution does require (though sometimes it does not require) the State to treat a sentencing factor as an element. But we do not see how they can help petitioner more than that.

We turn then to the case upon which petitioner must primarily rely, McMillan v. Pennsylvania. The Court there considered a Pennsylvania statute that set forth a sentencing factor — “visibly possessing a firearm” — the presence of which required the judge to impose a minimum prison term of five years. The Court held that the Constitution did not require the State to treat the factor as an element of the crime. In so holding, the Court said that the State’s “link-ting] the ‘severity of punishment’ to ‘the presence or absence of an identified fact’ ” did not automatically make of that fact an “element.” Id., at 84 (quoting Patterson v. New York, supra, at 214). It said, citing Patterson, that “the state legislature’s definition of the elements of the offense is usually dispositive.” 477 U. S., at 85. It said that it would not “define precisely the constitutional limits” of a legislature’s power to define the elements of an offense. Id., at 86. And it held that, whatever those limits might be, the State had not exceeded them. Ibid. Petitioner must therefore concede that “firearm possession” (in respect to a mandatory minimum sentence) does not violate those limits. And he must argue that, nonetheless, “recidivism” (in respect to an authorized maximum) does violate those limits.

In assessing petitioner’s claim, we have examined McMillan to determine the various features of the case upon which the Court’s conclusion arguably turned. The McMillan Court pointed out: (1) that the statute plainly “does not transgress the limits expressly set out in Patterson,” ibid.; (2) that the defendant (unlike Mullaney’s defendant) did not face “‘a differential in sentencing ranging from a nominal *243fine to a mandatory life sentence,' ” 477 U. S., at 87 (quoting Mullaney, 421 U. S., at 700); (3) that the statute did not “alte[r] the maximum penalty for the crime” but “operates solely to limit the sentencing court’s discretion in selecting a penalty within the range already available to it,” 477 U. S., at 87-88; (4) that the statute did not “ereat[e] a separate offense calling for a separate penalty,” id., at 88; and (5) that the statute gave “no impression of having been tailored to permit the visible possession finding to be a tail which wags the dog of the substantive offense,” but, to the contrary, “simply took one factor that has always been considered by sentencing courts to bear on punishment... and dictated the precise weight to be given that factor,” id., at 88, 89-90.

This case resembles McMillan in respect to most of these factors. But it is different in respect to the third factor, for it does “alte[r] the maximum penalty for the crime,” id., at 87; and it also creates a wider range of appropriate punishments than did the statute in McMillan. We nonetheless conclude that these differences do not change the constitutional outcome for several basic reasons.

First, the sentencing factor at issue here — recidivism—is a traditional, if not the most traditional, basis for a sentencing court’s increasing an offender’s sentence. See, e. g., Parke v. Raley, 506 U. S. 20, 26 (1992) (Recidivism laws “have a long tradition in this country that dates back to colonial times” and currently are in effect in all 50 States); U. S. Dept, of Justice, Office of Justice Programs, Statutes Requiring the Use of Criminal History Record Information 17-41 (June 1991) (50-state survey); USSG §§ 4A1.1, 4A1.2 (Nov. 1997) (requiring sentencing court to consider defendant’s prior record in every ease). Consistent with this tradition, the Court said long ago that a State need not allege a defendant’s prior conviction in the indictment or information that alleges the elements of an underlying crime, even though the conviction was “necessary to bring the ease within the statute.” Graham v. West Virginia, 224 U. S. 616, 624 (1912). That con-*244elusion followed, the Court said, from “the distinct nature of the issue,” and the fact that recidivism “does not relate to the commission of the offense, but goes to the punishment only, and therefore . . . may be subsequently decided.” Id., at 629 (emphasis added). The Court has not deviated from this view. See Oyler v. Boles, 368 U. S. 448, 452 (1962) (due process does not require advance notice that trial for substantive offense will be followed by accusation that the defendant is a habitual offender); Parke, supra, at 27 (“[A] charge -under a recidivism statute does not state a separate offense, but goes to punishment only”). And, as we said before, supra, at 230, Congress, reflecting this tradition, has never, to our knowledge, made a defendant’s recidivism an element of an offense where the conduct proscribed is otherwise unlawful. See United States v. Jackson, 824 F. 2d 21, 25, and n. 6 (CADC 1987) (opinion of R. Ginsburg, J.) (referring to fact that few, if any, federal statutes make “prior criminal convictions ... elements of another criminal offense to be proved before the jury”). Although these precedents do not foreclose petitioner’s claim (because, for example, the state statute at issue in Graham and Oyler provided for a jury determination of disputed prior convictions), to hold that the Constitution requires that recidivism be deemed an “element” of petitioner’s offense would mark an abrupt departure from a longstanding tradition of treating recidivism as “go[ing] to the punishment only.” Graham, supra, at 629.

Second, the major difference between this ease and McMillan consists of the circumstance that the sentencing factor at issue here (the prior conviction) triggers an increase in the maximum permissive sentence, while the sentencing factor at issue in McMillan triggered a mandatory minimum sentence. Yet that difference — between a permissive maximum and a mandatory minimum — does not systematically, or normally, work to the disadvantage of a criminal defendant. To the contrary, a statutory minimum binds a sentencing judge; a statutory maximum does not. A mandatory mini*245mum can, as Justice Stevens dissenting in McMillan pointed out, “mandate a minimum sentence of imprisonment more than twiee as severe as the maximum the trial judge would otherwise have imposed.” 477 U. S., at 95. It can eliminate a sentencing judge’s discretion in its entirety. See, e. g., 18 U. S. C. § 2241(c) (authorizing maximum term of life imprisonment for sexual abuse of children; mandating life imprisonment for second offense). And it can produce unfairly disproportionate impacts on certain kinds of offenders. See United States Sentencing Commission, Mandatory Minimum Penalties in the Federal Criminal Justice System 26-84 (Aug. 1991) (discussing “tariff” and “cliff” effects of mandatory mínimums). In sum, the risk of unfairness to a particular defendant is no less, and may well be greater, when a mandatory minimum sentence, rather than a permissive maximum sentence, is at issue.

Although McMillan pointed to a difference between mandatory mínimums and higher authorized máximums, it neither “rested its judgment” on that difference, nor “rejected” the above analysis, as the dissent contends, post, at 254. Rather, McMillan said that the petitioners’ argument in that ease would have had “more superficial appeal” if the sentencing fact “exposed them to greater or additional punishment.” 477 U. S., at 88 (emphasis added). For the reasons just given, and in light of the particular sentencing factor at issue in this ease — recidivism—we should take McMillan’s statement to mean no more than it said, and therefore not to make a determinative difference here.

Third, the statute’s broad permissive sentencing range does not itself create significantly greater unfairness. Judges (and parole boards) have typically exercised their discretion within broad statutory ranges. See, e. g., supra, at 232, 236 (statutory examples); National Institute of Justice, Sentencing Reform in the United States (Aug. 1985) (survey of sentencing laws in the 50 States); L. Friedman, Crime and Punishment in American History 159-163 (1993) *246(history of indeterminate sentencing). And the Sentencing Guidelines have recently sought to channel that discretion using “sentencing factors” which no one here claims that the Constitution thereby makes “elements” of a crime.

Finally, the remaining McMillan factors support the conclusion that Congress has the constitutional power to treat the feature before us — prior conviction of an aggravated felony' — as a sentencing factor for this particular offense (illegal entry after deportation). The relevant statutory provisions do not change a pre-existing definition of a well-established crime, nor is there any more reason here, than in McMillan, to think Congress intended to “evade” the Constitution, either by “presuming” guilt or “restructuring” the elements of an offense. Cf. McMillan, supra, at 86-87, 89-90.

For these reasons, we cannot find in McMillan (a case holding that the Constitution permits a legislature to require a longer sentence for gun possession) significant support for the proposition that the Constitution forbids a legislature to authorize a longer sentence for recidivism.

Petitioner makes two basic additional arguments in response. He points to what he calls a different “tradition”— that of courts having treated recidivism as an element of the related crime. See, e. g., Massey v. United States, 281 F. 293, 297-298 (CA8 1922); Singer v. United States, 278 F. 415, 420 (CA3 1922); People v. Sickles, 51 N. E. 288, 289 (N. Y. 1898); see also post, at 256-257 (citing authority). We do not find this claim convincing, however, for any such tradition is not uniform. See Spencer v. Texas, 385 U. S., at 566 (“The method for determining prior convictions varies . . . between jurisdictions affording a jury trial on this issue ... and those leaving that question to the court”); Note, Recidivist Procedures, 40 N. Y. U. L. Rev. 332, 347 (1965) (as of 1965, eight States’ recidivism statutes provide for determination of prior convictions by judge, not jury). Nor does it appear modern. Compare State v. Thorne, 129 Wash. 2d *247736, 776-784, 921 P. 2d 514, 533-538 (1996) (upholding state recidivism law against federal constitutional challenge), with State v. Furth, 5 Wash. 2d 1, 11-19, 104 P. 2d 925, 930-933 (1940). And it nowhere (to our knowledge) rested upon a federal constitutional guarantee. See, e. g., Massey v. United States, supra, at 297 (applying federal law, noting jury determination of prior offense applied “unless the statute designates a different mode of procedure”).

Petitioner also argues, in essence, that this Court, should simply adopt a rule that any significant increase in a statutory maximum sentence would trigger a constitutional “elements” requirement. We have explained why we believe the Constitution, as interpreted in McMillan and earlier eases, does not impose that requirement. We add that such a rule would seem anomalous in light of existing case law that permits a judge, father than a jury, to determine the existence of factors that can make a defendant eligible for the death penalty, a punishment far more severe than that faced by petitioner here. See Walton v. Arizona, 497 U. S. 639, 647 (1990) (rejecting capital defendant’s argument that every finding of fact underlying death sentence must be made by a jury); Hildwin v. Florida, 490 U. S. 638, 640-641 (1989) (per curiam) (judge may impose death penalty based on his finding of aggravating factor because such factor is not element of offense to be determined by jury); Spaziano v. Florida, 468 U. S. 447, 465 (1984) (same). And we would also find it difficult to reconcile any such rule with our precedent holding that the senteneing-related circumstances of recidivism are not part of the definition of the offense for double jeopardy purposes. Graham, 224 U. S., at 623-624.

For these reasons, we reject petitioner’s constitutional claim that his recidivism must be treated as an element of his offense.

IV

We mention one final point. Petitioner makes no separate, subsidiary, standard of proof claims with respect to his *248sentencing, perhaps because he admitted his recidivism at the time he pleaded guilty and would therefore find it difficult to show that the standard of proof could have made a difference to his ease. Accordingly, we express no view on whether some heightened standard of proof might apply to sentencing determinations that bear significantly on the severity of sentence. Cf. United States v. Watts, 519 U. S. 148, 156, and n. 2 (1997) (per curiam) (acknowledging, but not resolving, “divergence of opinion among the Circuits” as to proper standard for determining the existence of “relevant conduct” that would lead to an increase in sentence).

The judgment of the Court of Appeals is

Affirmed.

Justice Scalia,

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 *258 McMillan’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 *268 opment 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.

Justice SCALIAdelivered the opinion of the Court.

Under the Armed Career Criminal Act of 1984, a defendant convicted of being a felon in possession of a firearm faces more severe punishment if he has three or more previous convictions for a "violent felony," a term defined to include any felony that "involves conduct that presents a serious potential risk of physical injury to another." 18 U.S.C. § 924(e)(2)(B). We must decide whether this part of the definition of a violent felony survives the Constitution's prohibition of vague criminal laws.

I

Federal law forbids certain people-such as convicted felons, persons committed to mental institutions, and drug users-to ship, possess, and receive firearms. § 922(g). In general, the law punishes violation of this ban by up to 10 years' imprisonment. § 924(a)(2). But if the violator has three or more earlier convictions for a "serious drug offense" or a "violent felony," the Armed Career Criminal Act increases his prison term to a minimum of 15 years and a maximum of life. § 924(e)(1); Johnson v. United States,559 U.S. 133, 136, 130 S.Ct. 1265, 176 L.Ed.2d 1 (2010). The Act defines "violent felony" as follows:

"any crime punishable by imprisonment for a term exceeding one year ... that-

"(i) has as an element the use, attempted use, or threatened use of physical force against the person of another; or

"(ii) is burglary, arson, or extortion, involves use of explosives, or otherwise involves conduct that presents a serious *2556potential risk of physical injury to another." § 924(e)(2)(B)(emphasis added).

The closing words of this definition, italicized above, have come to be known as the Act's residual clause. Since 2007, this Court has decided four cases attempting to discern its meaning. We have held that the residual clause (1) covers Florida's offense of attempted burglary, James v. United States,550 U.S. 192, 127 S.Ct. 1586, 167 L.Ed.2d 532 (2007); (2) does notcover New Mexico's offense of driving under the influence, Begay v. United States,553 U.S. 137, 128 S.Ct. 1581, 170 L.Ed.2d 490 (2008); (3) does not cover Illinois' offense of failure to report to a penal institution, Chambers v. United States,555 U.S. 122, 129 S.Ct. 687, 172 L.Ed.2d 484 (2009); and (4) does cover Indiana's offense of vehicular flight from a law-enforcement officer, Sykes v. United States,564 U.S. 1, 131 S.Ct. 2267, 180 L.Ed.2d 60 (2011). In both Jamesand Sykes, the Court rejected suggestions by dissenting Justices that the residual clause violates the Constitution's prohibition of vague criminal laws. Compare James, 550 U.S., at 210, n. 6, 127 S.Ct. 1586, with id.,at 230, 127 S.Ct. 1586(SCALIA, J., dissenting); compare Sykes, 564 U.S., at ----, 131 S.Ct., at 2276-2277, with id.,at ----, 131 S.Ct., at 2286-2288(SCALIA, J., dissenting).

This case involves the application of the residual clause to another crime, Minnesota's offense of unlawful possession of a short-barreled shotgun. Petitioner Samuel Johnson is a felon with a long criminal record. In 2010, the Federal Bureau of Investigation began to monitor him because of his involvement in a white-supremacist organization that the Bureau suspected was planning to commit acts of terrorism. During the investigation, Johnson disclosed to undercover agents that he had manufactured explosives and that he planned to attack "the Mexican consulate" in Minnesota, "progressive bookstores," and " 'liberals.' " Revised Presentence Investigation in No. 0:12CR00104-001 (D. Minn.), p. 15, ¶ 16. Johnson showed the agents his AK-47 rifle, several semiautomatic firearms, and over 1,000 rounds of ammunition.

After his eventual arrest, Johnson pleaded guilty to being a felon in possession of a firearm in violation of § 922(g). The Government requested an enhanced sentence under the Armed Career Criminal Act. It argued that three of Johnson's previous offenses-including unlawful possession of a short-barreled shotgun, see Minn.Stat. § 609.67(2006)-qualified as violent felonies. The District Court agreed and sentenced Johnson to a 15-year prison term under the Act. The Court of Appeals affirmed. 526 Fed.Appx. 708 (C.A.8 2013)(per curiam). We granted certiorari to decide whether Minnesota's offense of unlawful possession of a short-barreled shotgun ranks as a violent felony under the residual clause. 572 U.S. ----, 134 S.Ct. 1871, 188 L.Ed.2d 910 (2014). We later asked the parties to present reargument addressing the compatibility of the residual clause with the Constitution's prohibition of vague criminal laws. 574 U.S. ----, 135 S.Ct. 939, 190 L.Ed.2d 718 (2015).

II

The Fifth Amendment provides that "[n]o person shall ... be deprived of life, liberty, or property, without due process of law." Our cases establish that the Government violates this guarantee by taking away someone's life, liberty, or property under a criminal law so vague that it fails to give ordinary people fair notice of the conduct it punishes, or so standardless that it invites arbitrary enforcement.Kolender v. Lawson,461 U.S. 352, 357-358, 103 S.Ct. 1855, 75 L.Ed.2d 903 (1983). The prohibition of vagueness *2557in criminal statutes "is a well-recognized requirement, consonant alike with ordinary notions of fair play and the settled rules of law," and a statute that flouts it "violates the first essential of due process." Connally v. General Constr. Co.,269 U.S. 385, 391, 46 S.Ct. 126, 70 L.Ed. 322 (1926). These principles apply not only to statutes defining elements of crimes, but also to statutes fixing sentences. United States v. Batchelder,442 U.S. 114, 123, 99 S.Ct. 2198, 60 L.Ed.2d 755 (1979).

In Taylor v. United States,495 U.S. 575, 600, 110 S.Ct. 2143, 109 L.Ed.2d 607 (1990), this Court held that the Armed Career Criminal Act requires courts to use a framework known as the categorical approach when deciding whether an offense "is burglary, arson, or extortion, involves use of explosives, or otherwise involves conduct that presents a serious potential risk of physical injury to another." Under the categorical approach, a court assesses whether a crime qualifies as a violent felony "in terms of how the law defines the offense and not in terms of how an individual offender might have committed it on a particular occasion." Begay, supra,at 141, 128 S.Ct. 1581.

Deciding whether the residual clause covers a crime thus requires a court to picture the kind of conduct that the crime involves in "the ordinary case," and to judge whether that abstraction presents a serious potential risk of physical injury. James, supra,at 208, 127 S.Ct. 1586. The court's task goes beyond deciding whether creation of risk is an element of the crime. That is so because, unlike the part of the definition of a violent felony that asks whether the crime "has as an elementthe use ... of physical force," the residual clause asks whether the crime "involves conduct" that presents too much risk of physical injury. What is more, the inclusion of burglary and extortion among the enumerated offenses preceding the residual clause confirms that the court's task also goes beyond evaluating the chances that the physical acts that make up the crime will injure someone. The act of making an extortionate demand or breaking and entering into someone's home does not, in and of itself, normally cause physical injury. Rather, risk of injury arises because the extortionist might engage in violence aftermaking his demand or because the burglar might confront a resident in the home afterbreaking and entering.

We are convinced that the indeterminacy of the wide-ranging inquiry required by the residual clause both denies fair notice to defendants and invites arbitrary enforcement by judges. Increasing a defendant's sentence under the clause denies due process of law.

A

Two features of the residual clause conspire to make it unconstitutionally vague. In the first place, the residual clause leaves grave uncertainty about how to estimate the risk posed by a crime. It ties the judicial assessment of risk to a judicially imagined "ordinary case" of a crime, not to real-world facts or statutory elements. How does one go about deciding what kind of conduct the "ordinary case" of a crime involves? "A statistical analysis of the state reporter? A survey? Expert evidence? Google? Gut instinct?" United States v. Mayer,560 F.3d 948, 952 (C.A.9 2009)(Kozinski, C.J., dissenting from denial of rehearing en banc). To take an example, does the ordinary instance of witness tampering involve offering a witness a bribe? Or threatening a witness with violence? Critically, picturing the criminal's behavior is not enough; as we have already discussed, assessing "potential risk" seemingly requires the judge to imagine how the idealized ordinary *2558case of the crime subsequently plays out. Jamesillustrates how speculative (and how detached from statutory elements) this enterprise can become. Explaining why attempted burglary poses a serious potential risk of physical injury, the Court said: "An armed would-be burglar may be spotted by a police officer, a private security guard, or a participant in a neighborhood watch program. Or a homeowner ... may give chase, and a violent encounter may ensue." 550 U.S., at 211, 127 S.Ct. 1586. The dissent, by contrast, asserted that any confrontation that occurs during an attempted burglary "is likely to consist of nothing more than the occupant's yelling 'Who's there?' from his window, and the burglar's running away." Id.,at 226, 127 S.Ct. 1586(opinion of SCALIA, J.). The residual clause offers no reliable way to choose between these competing accounts of what "ordinary" attempted burglary involves.

At the same time, the residual clause leaves uncertainty about how much risk it takes for a crime to qualify as a violent felony. It is one thing to apply an imprecise "serious potential risk" standard to real-world facts; it is quite another to apply it to a judge-imagined abstraction. By asking whether the crime "otherwiseinvolves conduct that presents a serious potential risk," moreover, the residual clause forces courts to interpret "serious potential risk" in light of the four enumerated crimes-burglary, arson, extortion, and crimes involving the use of explosives. These offenses are "far from clear in respect to the degree of risk each poses." Begay,553 U.S., at 143, 128 S.Ct. 1581. Does the ordinary burglar invade an occupied home by night or an unoccupied home by day? Does the typical extortionist threaten his victim in person with the use of force, or does he threaten his victim by mail with the revelation of embarrassing personal information? By combining indeterminacy about how to measure the risk posed by a crime with indeterminacy about how much risk it takes for the crime to qualify as a violent felony, the residual clause produces more unpredictability and arbitrariness than the Due Process Clause tolerates.

This Court has acknowledged that the failure of "persistent efforts ... to establish a standard" can provide evidence of vagueness. United States v. L. Cohen Grocery Co.,255 U.S. 81, 91, 41 S.Ct. 298, 65 L.Ed. 516 (1921). Here, this Court's repeated attempts and repeated failures to craft a principled and objective standard out of the residual clause confirm its hopeless indeterminacy. Three of the Court's previous four decisions about the clause concentrated on the level of risk posed by the crime in question, though in each case we found it necessary to resort to a different ad hoc test to guide our inquiry. In James,we asked whether "the risk posed by attempted burglary is comparable to that posed by its closest analog among the enumerated offenses," namely completed burglary; we concluded that it was. 550 U.S., at 203, 127 S.Ct. 1586. That rule takes care of attempted burglary, but offers no help at all with respect to the vast majority of offenses, which have no apparent analog among the enumerated crimes. "Is, for example, driving under the influence of alcohol more analogous to burglary, arson, extortion, or a crime involving use of explosives?"Id.,at 215, 127 S.Ct. 1586(SCALIA, J., dissenting).

Chambers,our next case to focus on risk, relied principally on a statistical report prepared by the Sentencing Commission to conclude that an offender who fails to report to prison is not "significantly more likely than others to attack, or physically to resist, an apprehender, thereby producing a 'serious potential risk of physical *2559injury.' " 555 U.S., at 128-129, 129 S.Ct. 687. So much for failure to report to prison, but what about the tens of thousands of federal and state crimes for which no comparable reports exist? And even those studies that are available might suffer from methodological flaws, be skewed toward rarer forms of the crime, or paint widely divergent pictures of the riskiness of the conduct that the crime involves. See Sykes,564 U.S., at ---- - ----, 131 S.Ct., at 2285-2287(SCALIA, J., dissenting); id.,at ----, n. 4, 131 S.Ct., at 2291, n. 4(KAGAN, J., dissenting).

Our most recent case, Sykes, also relied on statistics, though only to "confirm the commonsense conclusion that Indiana's vehicular flight crime is a violent felony." Id.,at ----, 131 S.Ct., at 2274(majority opinion). But common sense is a much less useful criterion than it sounds-as Sykesitself illustrates. The Indiana statute involved in that case covered everything from provoking a high-speed car chase to merely failing to stop immediately after seeing a police officer's signal. See id.,at ----, 131 S.Ct., at 2289-2290(KAGAN, J., dissenting). How does common sense help a federal court discern where the "ordinary case" of vehicular flight in Indiana lies along this spectrum? Common sense has not even produced a consistent conception of the degree of risk posed by each of the four enumerated crimes; there is no reason to expect it to fare any better with respect to thousands of unenumerated crimes. All in all, James,Chambers,and Sykesfailed to establish any generally applicable test that prevents the risk comparison required by the residual clause from devolving into guesswork and intuition.

The remaining case, Begay,which preceded Chambersand Sykes, took an entirely different approach. The Court held that in order to qualify as a violent felony under the residual clause, a crime must resemble the enumerated offenses "in kind as well as in degree of risk posed." 553 U.S., at 143, 128 S.Ct. 1581. The Court deemed drunk driving insufficiently similar to the listed crimes, because it typically does not involve "purposeful, violent, and aggressive conduct." Id.,at 144-145, 128 S.Ct. 1581(internal quotation marks omitted). Alas, Begaydid not succeed in bringing clarity to the meaning of the residual clause. It did not (and could not) eliminate the need to imagine the kind of conduct typically involved in a crime. In addition, the enumerated crimes are not much more similar to one another in kind than in degree of risk posed, and the concept of "aggressive conduct" is far from clear. Sykescriticized the "purposeful, violent, and aggressive" test as an "addition to the statutory text," explained that "levels of risk" would normally be dispositive, and confined Begayto "strict liability, negligence, and recklessness crimes." 564 U.S., at ---- - ----, 131 S.Ct., at 2275-2276.

The present case, our fifth about the meaning of the residual clause, opens a new front of uncertainty. When deciding whether unlawful possession of a short-barreled shotgun is a violent felony, do we confine our attention to the risk that the shotgun will go off by accident while in someone's possession? Or do we also consider the possibility that the person possessing the shotgun will later use it to commit a crime? The inclusion of burglary and extortion among the enumerated offenses suggests that a crime may qualify under the residual clause even if the physical injury is remote from the criminal act. But how remote is too remote? Once again, the residual clause yields no answers.

This Court is not the only one that has had trouble making sense of the residual *2560clause. The clause has "created numerous splits among the lower federal courts," where it has proved "nearly impossible to apply consistently." Chambers,555 U.S., at 133, 129 S.Ct. 687(ALITO, J., concurring in judgment). The most telling feature of the lower courts' decisions is not division about whether the residual clause covers this or that crime (even clear laws produce close cases); it is, rather, pervasive disagreement about the nature of the inquiry one is supposed to conduct and the kinds of factors one is supposed to consider. Some judges have concluded that deciding whether conspiracy is a violent felony requires evaluating only the dangers posed by the "simple act of agreeing [to commit a crime]," United States v. Whitson,597 F.3d 1218, 1222 (C.A.11 2010)(per curiam); others have also considered the probability that the agreement will be carried out, United States v. White,571 F.3d 365, 370-371 (C.A.4 2009). Some judges have assumed that the battery of a police officer (defined to include the slightest touching) could "explode into violence and result in physical injury," United States v. Williams,559 F.3d 1143, 1149 (C.A.10 2009); others have felt that it "do[es] a great disservice to law enforcement officers" to assume that they would "explod[e] into violence" rather than "rely on their training and experience to determine the best method of responding," United States v. Carthorne,726 F.3d 503, 514 (C.A.4 2013). Some judges considering whether statutory rape qualifies as a violent felony have concentrated on cases involving a perpetrator much older than the victim, United States v. Daye,571 F.3d 225, 230-231 (C.A.2 2009); others have tried to account for the possibility that "the perpetrator and the victim [might be] close in age," United States v. McDonald,592 F.3d 808, 815 (C.A.7 2010). Disagreements like these go well beyond disputes over matters of degree.

It has been said that the life of the law is experience. Nine years' experience trying to derive meaning from the residual clause convinces us that we have embarked upon a failed enterprise. Each of the uncertainties in the residual clause may be tolerable in isolation, but "their sum makes a task for us which at best could be only guesswork." United States v. Evans,333 U.S. 483, 495, 68 S.Ct. 634, 92 L.Ed. 823 (1948). Invoking so shapeless a provision to condemn someone to prison for 15 years to life does not comport with the Constitution's guarantee of due process.

B

The Government and the dissent claim that there will be straightforward cases under the residual clause, because some crimes clearly pose a serious potential risk of physical injury to another. See post, at 2562 - 2563 (opinion of ALITO, J.). True enough, though we think many of the cases the Government and the dissent deem easy turn out not to be so easy after all. Consider just one of the Government's examples, Connecticut's offense of "rioting at a correctional institution." See United States v. Johnson,616 F.3d 85 (C.A.2 2010). That certainly sounds like a violent felony-until one realizes that Connecticut defines this offense to include taking part in "any disorder, disturbance, strike, riot or other organized disobedience to the rules and regulations" of the prison. Conn. Gen.Stat. § 53a-179b(a)(2012). Who is to say which the ordinary "disorder" most closely resembles-a full-fledged prison riot, a food-fight in the prison cafeteria, or a "passive and nonviolent [act] such as disregarding an order to move," Johnson,616 F.3d, at 95(Parker, J., dissenting)?

In all events, although statements in some of our opinions could be read to *2561suggest otherwise, our holdingssquarely contradict the theory that a vague provision is constitutional merely because there is some conduct that clearly falls within the provision's grasp. For instance, we have deemed a law prohibiting grocers from charging an "unjust or unreasonable rate" void for vagueness-even though charging someone a thousand dollars for a pound of sugar would surely be unjust and unreasonable. L. Cohen Grocery Co.,255 U.S., at 89, 41 S.Ct. 298. We have similarly deemed void for vagueness a law prohibiting people on sidewalks from "conduct[ing] themselves in a manner annoying to persons passing by"-even though spitting in someone's face would surely be annoying. Coates v. Cincinnati,402 U.S. 611, 91 S.Ct. 1686, 29 L.Ed.2d 214 (1971). These decisions refute any suggestion that the existence of someobviously risky crimes establishes the residual clause's constitutionality.

Resisting the force of these decisions, the dissent insists that "a statute is void for vagueness only if it is vague in all its applications." Post, at 2574. It claims that the prohibition of unjust or unreasonable rates in L. Cohen Grocerywas "vague in all applications," even though one can easily envision rates so high that they are unreasonable by any measure. Post, at 2582. It seems to us that the dissent's supposed requirement of vagueness in all applications is not a requirement at all, but a tautology: If we hold a statute to be vague, it is vague in all its applications (and never mind the reality). If the existence of some clearly unreasonable rates would not save the law in L. Cohen Grocery,why should the existence of some clearly risky crimes save the residual clause?

The Government and the dissent next point out that dozens of federal and state criminal laws use terms like "substantial risk," "grave risk," and "unreasonable risk," suggesting that to hold the residual clause unconstitutional is to place these provisions in constitutional doubt. See post,at 2558 - 2559. Not at all. Almost none of the cited laws links a phrase such as "substantial risk" to a confusing list of examples. "The phrase 'shades of red,' standing alone, does not generate confusion or unpredictability; but the phrase 'fire-engine red, light pink, maroon, navy blue,or colors that otherwise involve shades of red' assuredly does so." James,550 U.S., at 230, n. 7, 127 S.Ct. 1586(SCALIA, J., dissenting). More importantly, almost all of the cited laws require gauging the riskiness of conduct in which an individual defendant engages on a particular occasion. As a general matter, we do not doubt the constitutionality of laws that call for the application of a qualitative standard such as "substantial risk" to real-world conduct; "the law is full of instances where a man's fate depends on his estimating rightly ... some matter of degree," Nash v. United States,229 U.S. 373, 377, 33 S.Ct. 780, 57 L.Ed. 1232 (1913). The residual clause, however, requires application of the "serious potential risk" standard to an idealized ordinary case of the crime. Because "the elements necessary to determine the imaginary ideal are uncertain both in nature and degree of effect," this abstract inquiry offers significantly less predictability than one "[t]hat deals with the actual, not with an imaginary condition other than the facts." International Harvester Co. of America v. Kentucky,234 U.S. 216, 223, 34 S.Ct. 853, 58 L.Ed. 1284 (1914).

Finally, the dissent urges us to save the residual clause from vagueness by interpreting it to refer to the risk posed by the particular conduct in which the defendant engaged, not the risk posed by the ordinary case of the defendant's crime.

*2562See post, at 2578 - 2580. In other words, the dissent suggests that we jettison for the residual clause (though not for the enumerated crimes) the categorical approach adopted in Taylor,see 495 U.S., at 599-602, 110 S.Ct. 2143, and reaffirmed in each of our four residual-clause cases, see James,550 U.S., at 202, 127 S.Ct. 1586; Begay,553 U.S., at 141, 128 S.Ct. 1581; Chambers,555 U.S., at 125, 129 S.Ct. 687; Sykes,564 U.S., ----, 131 S.Ct., at 2272-2273. We decline the dissent's invitation. In the first place, the Government has not asked us to abandon the categorical approach in residual-clause cases. In addition, Taylorhad good reasons to adopt the categorical approach, reasons that apply no less to the residual clause than to the enumerated crimes. Taylorexplained that the relevant part of the Armed Career Criminal Act "refers to 'a person who ... has three previous convictions' for-not a person who has committed-three previous violent felonies or drug offenses." 495 U.S., at 600, 110 S.Ct. 2143. This emphasis on convictions indicates that "Congress intended the sentencing court to look only to the fact that the defendant had been convicted of crimes falling within certain categories, and not to the facts underlying the prior convictions." Ibid. Tayloralso pointed out the utter impracticability of requiring a sentencing court to reconstruct, long after the original conviction, the conduct underlying that conviction. For example, if the original conviction rested on a guilty plea, no record of the underlying facts may be available. "[T]he only plausible interpretation" of the law, therefore, requires use of the categorical approach. Id.,at 602, 110 S.Ct. 2143.

C

That brings us to stare decisis. This is the first case in which the Court has received briefing and heard argument from the parties about whether the residual clause is void for vagueness. In James,however, the Court stated in a footnote that it was "not persuaded by [the principal dissent's] suggestion ... that the residual provision is unconstitutionally vague." 550 U.S., at 210, n. 6, 127 S.Ct. 1586. In Sykes, the Court again rejected a dissenting opinion's claim of vagueness. 564 U.S., at ---- - ----, 131 S.Ct., at 2276-2277.

The doctrine of stare decisisallows us to revisit an earlier decision where experience with its application reveals that it is unworkable. Payne v. Tennessee,501 U.S. 808, 827, 111 S.Ct. 2597, 115 L.Ed.2d 720 (1991). Experience is all the more instructive when the decision in question rejected a claim of unconstitutional vagueness. Unlike other judicial mistakes that need correction, the error of having rejected a vagueness challenge manifests itself precisely in subsequent judicial decisions: the inability of later opinions to impart the predictability that the earlier opinion forecast. Here, the experience of the federal courts leaves no doubt about the unavoidable uncertainty and arbitrariness of adjudication under the residual clause. Even after Sykestried to clarify the residual clause's meaning, the provision remains a "judicial morass that defies systemic solution," "a black hole of confusion and uncertainty" that frustrates any effort to impart "some sense of order and direction." United States v. Vann,660 F.3d 771, 787 (C.A.4 2011)(Agee, J., concurring).

This Court's cases make plain that even decisions rendered after full adversarial presentation may have to yield to the lessons of subsequent experience. See, e.g., United States v. Dixon,509 U.S. 688, 711, 113 S.Ct. 2849, 125 L.Ed.2d 556 (1993); Payne, 501 U.S., at 828-830, 111 S.Ct. 2597 (1991). But Jamesand Sykesopined about vagueness without full briefing *2563or argument on that issue-a circumstance that leaves us "less constrained to follow precedent," Hohn v. United States,524 U.S. 236, 251, 118 S.Ct. 1969, 141 L.Ed.2d 242 (1998). The brief discussions of vagueness in Jamesand Sykeshomed in on the imprecision of the phrase "serious potential risk"; neither opinion evaluated the uncertainty introduced by the need to evaluate the riskiness of an abstract ordinary case of a crime. 550 U.S., at 210, n. 6, 127 S.Ct. 1586, 564 U.S., at ----, 131 S.Ct., at 2276-2277. And departing from those decisions does not raise any concerns about upsetting private reliance interests.

Although it is a vital rule of judicial self-government, stare decisisdoes not matter for its own sake. It matters because it "promotes the evenhanded, predictable, and consistent development of legal principles." Payne, supra,at 827, 111 S.Ct. 2597. Decisions under the residual clause have proved to be anything but evenhanded, predictable, or consistent. Standing by Jamesand Sykeswould undermine, rather than promote, the goals that stare decisisis meant to serve.

* * *

We hold that imposing an increased sentence under the residual clause of the Armed Career Criminal Act violates the Constitution's guarantee of due process. Our contrary holdings in Jamesand Sykesare overruled. Today's decision does not call into question application of the Act to the four enumerated offenses, or the remainder of the Act's definition of a violent felony.

We reverse the judgment of the Court of Appeals for the Eighth Circuit and remand the case for further proceedings consistent with this opinion.

It is so ordered.

Justice KENNEDY, concurring in the judgment.

In my view, and for the reasons well stated by Justice ALITO in dissent, the residual clause of the Armed Career Criminal Act is not unconstitutionally vague under the categorical approach or a record-based approach. On the assumption that the categorical approach ought to still control, and for the reasons given by Justice THOMAS in Part I of his opinion concurring in the judgment, Johnson's conviction for possession of a short-barreled shotgun does not qualify as a violent felony.

For these reasons, I concur in the judgment.

Justice THOMAS, concurring in the judgment.

I agree with the Court that Johnson's sentence cannot stand. But rather than use the Fifth Amendment's Due Process Clause to ify an Act of Congress, I would resolve this case on more ordinary grounds. Under conventional principles of interpretation and our precedents, the offense of unlawfully possessing a short-barreled shotgun does not constitute a "violent felony" under the residual clause of the Armed Career Criminal Act (ACCA).

The majority wants more. Not content to engage in the usual business of interpreting statutes, it holds this clause to be unconstitutionally vague, notwithstanding the fact that on four previous occasions we found it determinate enough for judicial application. As Justice ALITO explains, that decision cannot be reconciled with our precedents concerning the vagueness doctrine. See post, at 2580 - 2581 (dissenting opinion). But even if it were a closer case under those decisions, I would be wary of holding the residual clause to be unconstitutionally vague. Although I have joined the Court in applying our modern vagueness *2564doctrine in the past, see FCC v. Fox Television Stations, Inc., 567 U.S. ----, ---- - ----, 132 S.Ct. 2307, 2319-2320, 183 L.Ed.2d 234 (2012), I have become increasingly concerned about its origins and application. Simply put, our vagueness doctrine shares an uncomfortably similar history with substantive due process, a judicially created doctrine lacking any basis in the Constitution.

I

We could have easily disposed of this case without ifying ACCA's residual clause. Under ordinary principles of statutory interpretation, the crime of unlawfully possessing a short-barreled shotgun does not constitute a "violent felony" under ACCA. In relevant part, that Act defines a "violent felony" as a "crime punishable by imprisonment for a term exceeding one year" that either

"(i) has as an element the use, attempted use, or threatened use of physical force against the person of another; or

"(ii) is burglary, arson, or extortion, involves use of explosives, or otherwise involves conduct that presents a serious potential risk of physical injury to another." 18 U.S.C. § 924(e)(2)(B).

The offense of unlawfully possessing a short-barreled shotgun neither satisfies the first clause of this definition nor falls within the enumerated offenses in the second. It therefore can constitute a violent felony only if it falls within ACCA's so-called "residual clause"-i.e.,if it "involves conduct that presents a serious potential risk of physical injury to another." § 924(e)(2)(B)(ii).

To determine whether an offense falls within the residual clause, we consider "whether the conduct encompassed by the elements of the offense, in the ordinary case, presents a serious potential risk of injury to another." James v. United States,550 U.S. 192, 208, 127 S.Ct. 1586, 167 L.Ed.2d 532 (2007). The specific crimes listed in § 924(e)(2)(B)(ii)-arson, extortion, burglary, and an offense involving the use of explosives-offer a "baseline against which to measure the degree of risk" a crime must present to fall within that clause. Id.,at 208, 127 S.Ct. 1586.Those offenses do not provide a high threshold, see id.,at 203, 207-208, 127 S.Ct. 1586, but the crime in question must still present a " 'serious' "-a " 'significant' or 'important' "-risk of physical injury to be deemed a violent felony, Begay v. United States,553 U.S. 137, 156, 128 S.Ct. 1581, 170 L.Ed.2d 490 (2008)(ALITO, J., dissenting); accord, Chambers v. United States,555 U.S. 122, 128, 129 S.Ct. 687, 172 L.Ed.2d 484 (2009).

To qualify as serious, the risk of injury generally must be closely related to the offense itself. Our precedents provide useful examples of the close relationship that must exist between the conduct of the offense and the risk presented. In Sykes v. United States,564 U.S. 1, 131 S.Ct. 2267, 180 L.Ed.2d 60 (2011), for instance, we held that the offense of intentional vehicular flight constitutes a violent felony because that conduct always triggers a dangerous confrontation, id.,at ----, 131 S.Ct., at 2274. As we explained, vehicular flights "by definitional necessity occur when police are present" and are done "in defiance of their instructions ... with a vehicle that can be used in a way to cause serious potential risk of physical injury to another." Ibid.In James,we likewise held that attempted burglary offenses "requir[ing] an overt act directed toward the entry of a structure" are violent felonies because the underlying conduct often results in a dangerous confrontation. 550 U.S., at 204, 206, 127 S.Ct. 1586. But we distinguished those crimes from "the more *2565attenuated conduct encompassed by" attempt offenses "that c[an] be satisfied by preparatory conduct that does not pose the same risk of violent confrontation," such as " 'possessing burglary tools.' " Id.,at 205, 206, and n. 4, 127 S.Ct. 1586. At some point, in other words, the risk of injury from the crime may be too attenuated for the conviction to fall within the residual clause, such as when an additional, voluntary act (e.g.,the useof burglary tools to enter a structure) is necessary to bring about the risk of physical injury to another.

In light of the elements of and reported convictions for the unlawful possession of a short-barreled shotgun, this crime does not "involv[e] conduct that presents a serious potential risk of physical injury to another," § 924(e)(2)(B)(ii). The acts that form the basis of this offense are simply too remote from a risk of physical injury to fall within the residual clause.

Standing alone, the elements of this offense-(1) unlawfully (2) possessing (3) a short-barreled shotgun-do not describe inherently dangerous conduct. As a conceptual matter, "simple possession [of a firearm], even by a felon, takes place in a variety of ways (e.g.,in a closet, in a storeroom, in a car, in a pocket) many, perhaps most, of which do not involve likely accompanying violence." United States v. Doe,960 F.2d 221, 225 (C.A.1 1992). These weapons also can be stored in a manner posing a danger to no one, such as unloaded, disassembled, or locked away. By themselves, the elements of this offense indicate that the ordinary commission of this crime is far less risky than ACCA's enumerated offenses.

Reported convictions support the conclusion that mere possession of a short-barreled shotgun does not, in the ordinary case, pose a serious risk of injury to others. A few examples suffice. In one case, officers found the sawed-off shotgun locked inside a gun cabinet in an empty home. State v. Salyers,858 N.W.2d 156, 157-158 (Minn.2015). In another, the firearm was retrieved from the trunk of the defendant's car. State v. Ellenberger,543 N.W.2d 673, 674 (Minn.App.1996). In still another, the weapon was found missing a firing pin. State v. Johnson,171 Wis.2d 175, 178, 491 N.W.2d 110, 111 (App.1992). In these instances and others, the offense threatened no one.

The Government's theory for why this crime should nonetheless qualify as a "violent felony" is unpersuasive. Although it does not dispute that the unlawful possession of a short-barreled shotgun can occur in a nondangerous manner, the Government contends that this offense poses a serious risk of physical injury due to the connection between short-barreled shotguns and other serious crimes. As the Government explains, these firearms are "weapons not typically possessed by law-abiding citizens for lawful purposes," District of Columbia v. Heller,554 U.S. 570, 625, 128 S.Ct. 2783, 171 L.Ed.2d 637 (2008), but are instead primarily intended for use in criminal activity. In light of that intended use, the Government reasons that the ordinary case of this possession offense will involve the useof a short-barreled shotgun in a serious crime, a scenario obviously posing a serious risk of physical injury.

But even assuming that those who unlawfully possess these weapons typically intend to use them in a serious crime, the risk that the Government identifies arises not from the act of possessing the weapon, but from the act of using it. Unlike attempted burglary (at least of the type at issue in James) or intentional vehicular flight-conduct that by itself often or always invites a dangerous confrontation-possession of a short-barreled shotgun poses a threat onlywhen an offender decides *2566to engage in additional, voluntary conduct that is not included in the elements of the crime. Until this weapon is assembled, loaded, or used, for example, it poses no risk of injury to others in and of itself. The risk of injury to others from mere possession of this firearm is too attenuated to treat this offense as a violent felony. I would reverse the Court of Appeals on that basis.

II

As the foregoing analysis demonstrates, ACCA's residual clause can be applied in a principled manner. One would have thought this proposition well established given that we have already decided four cases addressing this clause. The majority nonetheless concludes that the operation of this provision violates the Fifth Amendment's Due Process Clause.

Justice ALITO shows why that analysis is wrong under our precedents. See post, at 2580 - 2583 (dissenting opinion). But I have some concerns about our modern vagueness doctrine itself. Whether that doctrine is defensible under the original meaning of "due process of law" is a difficult question I leave for the another day, but the doctrine's history should prompt us at least to examine its constitutional underpinnings more closely before we use it to ify yet another duly enacted law.

A

We have become accustomed to using the Due Process Clauses to invalidate laws on the ground of "vagueness." The doctrine we have developed is quite sweeping: "A statute can be impermissibly vague ... if it fails to provide people of ordinary intelligence a reasonable opportunity to understand what conduct it prohibits" or "if it authorizes or even encourages arbitrary and discriminatory enforcement." Hill v. Colorado,530 U.S. 703, 732, 120 S.Ct. 2480, 147 L.Ed.2d 597 (2000). Using this framework, we have ified a wide range of enactments. We have struck down laws ranging from city ordinances, Papachristou v. Jacksonville,405 U.S. 156, 165-171, 92 S.Ct. 839, 31 L.Ed.2d 110 (1972), to Acts of Congress, United States v. L. Cohen Grocery Co.,255 U.S. 81, 89-93, 41 S.Ct. 298, 65 L.Ed. 516 (1921). We have struck down laws whether they are penal, Lanzetta v. New Jersey,306 U.S. 451, 452, 458, 59 S.Ct. 618, 83 L.Ed. 888 (1939), or not, Keyishian v. Board of Regents of Univ. of State of N. Y.,385 U.S. 589, 597-604, 87 S.Ct. 675, 17 L.Ed.2d 629 (1967).1We have struck down laws addressing subjects ranging from abortion, Colautti v. Franklin,439 U.S. 379, 390, 99 S.Ct. 675, 58 L.Ed.2d 596 (1979), and obscenity, Winters v. New York,333 U.S. 507, 517-520, 68 S.Ct. 665, 92 L.Ed. 840 (1948), to the minimum wage, Connally v. General Constr. Co.,269 U.S. 385, 390-395, 46 S.Ct. 126, 70 L.Ed. 322 (1926), and antitrust, Cline v. Frink Dairy Co.,274 U.S. 445, 453-465, 47 S.Ct. 681, 71 L.Ed. 1146 (1927). We have even struck down a *2567law using a term that has been used to describe criminal conduct in this country since before the Constitution was ratified.Chicago v. Morales,527 U.S. 41, 51, 119 S.Ct. 1849, 144 L.Ed.2d 67 (1999)(invalidating a "loitering" law); see id.,at 113, and n. 10, 119 S.Ct. 1849(THOMAS, J., dissenting) (discussing a 1764 Georgia law requiring the apprehension of "all able bodied persons ... who shall be found loitering").

That we have repeatedly used a doctrine to invalidate laws does not make it legitimate. Cf., e.g., Dred Scott v. Sandford,19 How. 393, 450-452, 15 L.Ed. 691 (1857)(stating that an Act of Congress prohibiting slavery in certain Federal Territories violated the substantive due process rights of slaveowners and was therefore void). This Court has a history of wielding doctrines purportedly rooted in "due process of law" to achieve its own policy goals, substantive due process being the poster child. See McDonald v. Chicago,561 U.S. 742, 811, 130 S.Ct. 3020, 177 L.Ed.2d 894 (2010)(THOMAS, J., concurring in part and concurring in judgment) ("The one theme that links the Court's substantive due process precedents together is their lack of a guiding principle to distinguish 'fundamental' rights that warrant protection from nonfundamental rights that do not"). Although our vagueness doctrine is distinct from substantive due process, their histories have disquieting parallels.

1

The problem of vague penal statutes is nothing new. The notion that such laws may be void under the Constitution's Due Process Clauses, however, is a more recent development.

Before the end of the 19th century, courts addressed vagueness through a rule of strict construction of penal statutes, not a rule of constitutional law. This rule of construction-better known today as the rule of lenity-first emerged in 16th-century England in reaction to Parliament's practice of making large swaths of crimes capital offenses, though it did not gain broad acceptance until the following century. See Hall, Strict or Liberal Construction of Penal Statutes, 48 Harv. L. Rev. 748, 749-751 (1935); see also 1 L. Radzinowicz, A History of English Criminal Law and Its Administration From 1750, pp. 10-11 (1948) (noting that some of the following crimes triggered the death penalty: "marking the edges of any current coin of the kingdom," "maliciously cutting any hop-binds growing on poles in any plantation of hops," and "being in the company of gypsies"). Courts relied on this rule of construction in refusing to apply vague capital-offense statutes to prosecutions before them. As an example of this rule, William Blackstone described a notable instance in which an English statute imposing the death penalty on anyone convicted of "stealing sheep, or other cattle" was "held to extend to nothing but mere sheep" as "th[e] general words, 'or other cattle,' [were] looked upon as much too loose to create a capital offence." 1 Commentaries on the Laws of England 88 (1765).2

Vague statutes surfaced on this side of the Atlantic as well. Shortly after the First Congress proposed the Bill of Rights, for instance, it passed a law providing *2568"[t]hat every person who shall attempt to trade with the Indian tribes, or be found in the Indian country with such merchandise in his possession as are usually vended to the Indians, without a license," must forfeit the offending goods. Act of July 22, 1790, ch. 33, § 3, 1 Stat. 137-138. At first glance, punishing the unlicensed possession of "merchandise ... usually vended to the Indians," ibid.,would seem far more likely to "invit [e] arbitrary enforcement," ante,at 2557, than does the residual clause.

But rather than strike down arguably vague laws under the Fifth Amendment Due Process Clause, antebellum American courts-like their English predecessors-simply refused to apply them in individual cases under the rule that penal statutes should be construed strictly. See, e.g.,United States v. Sharp,27 F.Cas. 1041 (No. 16,264) (C.C.Pa. 1815) (Washington, J.). In Sharp,for instance, several defendants charged with violating an Act rendering it a capital offense for "any seaman" to "make a revolt in [a] ship," Act of Apr. 30, 1790, § 8, 1 Stat. 114, objected that "the offence of making a revolt, [wa]s not sufficiently defined by this law, or by any other standard, to which reference could be safely made; to warrant the court in passing a sentence upon [them]." 27 F.Cas., at 1043. Justice Washington, riding circuit, apparently agreed, observing that the common definitions for the phrase "make a revolt" were "so multifarious, and so different" that he could not "avoid feeling a natural repugnance, to selecting from this mass of definitions, one, which may fix a crime upon these men, and that too of a capital nature." Ibid. Remarking that "[l]aws which create crimes, ought to be so explicit in themselves, or by reference to some other standard, that all men, subject to their penalties, may know what acts it is their duty to avoid," he refused to "recommend to the jury, to find the prisoners guilty of making, or endeavouring to make a revolt, however strong the evidence may be." Ibid.

Such analysis does not mean that federal courts believed they had the power to invalidate vague penal laws as unconstitutional. Indeed, there is good evidence that courts at the time understood judicial review to consist "of a refusal to give a statute effect as operative law in resolving a case," a notion quite distinct from our modern practice of " 'strik[ing] down' legislation." Walsh, Partial Unconstitutionality, 85 N.Y.U. L. Rev. 738, 756 (2010). The process of refusing to apply such laws appeared to occur on a case-by-case basis. For instance, notwithstanding his doubts expressed in Sharp,Justice Washington, writing for this Court, later rejected the argument that lower courts could arrest a judgment under the same ship-revolt statute because it "does not define the offence of endeavouring to make a revolt." United States v. Kelly,11 Wheat. 417, 418, 6 L.Ed. 508 (1826). The Court explained that "it is ... competent to the Court to give a judicial definition" of "the offence of endeavouring to make a revolt," and that such definition "consists in the endeavour of the crew of a vessel, or any one or more of them, to overthrow the legitimate authority of her commander, with intent to remove him from his command, or against his will to take possession of the vessel by assuming the government and navigation of her, or by transferring their obedience from the lawful commander to some other person." Id.,at 418-419. In dealing with statutory indeterminacy, federal courts saw themselves engaged in construction, not judicial review as it is now understood.3

*25692

Although vagueness concerns played a role in the strict construction of penal statutes from early on, there is little indication that anyone before the late 19th century believed that courts had the power under the Due Process Clauses to ify statutes on that ground. Instead, our modern vagueness doctrine materialized after the rise of substantive due process. Following the ratification of the Fourteenth Amendment, corporations began to use that Amendment's Due Process Clause to challenge state laws that attached penalties to unauthorized commercial conduct. In addition to claiming that these laws violated their substantive due process rights, these litigants began-with some success-to contend that such laws were unconstitutionally indefinite. In one case, a railroad company challenged a Tennessee law authorizing penalties against any railroad that demanded "more than a just and reasonable compensation" or engaged in "unjust and unreasonable discrimination" in setting its rates. Louisville & Nashville R. Co. v. Railroad Comm'n of Tenn.,19 F. 679, 690 (C.C.M.D.Tenn.1884)(internal quotation marks deleted). Without specifying the constitutional authority for its holding, the Circuit Court concluded that "[n]o citizen ... can be constitutionally subjected to penalties and despoiled of his property, in a criminal or quasi criminal proceeding, under and by force of such indefinite legislation." Id.,at 693(emphasis deleted).

Justice Brewer-widely recognized as "a leading spokesman for 'substantized' due process," Gamer, Justice Brewer and Substantive Due Process: A Conservative Court Revisited, 18 Vand. L. Rev. 615, 627 (1965)-employed similar reasoning while riding circuit, though he did not identify the constitutional source of judicial authority to ify vague laws. In reviewing an Iowa law authorizing fines against railroads for charging more than a "reasonable and just" rate, Justice Brewer mentioned in dictum that "no penal law can be sustained unless its mandates are so clearly expressed that any ordinary person can determine in advance what he may and what he may not do under it." Chicago & N.W.R. Co. v. Dey,35 F. 866, 876 (C.C.S.D.Iowa 1888).

Constitutional vagueness challenges in this Court initially met with some resistance. Although the Court appeared to acknowledge the possibility of unconstitutionally indefinite enactments, it repeatedly rejected vagueness challenges to penal laws addressing railroad rates, Railroad Comm'n Cases,116 U.S. 307, 336-337, 6 S.Ct. 1191, 29 L.Ed. 636 (1886), liquor sales, Ohio ex rel. Lloyd v. Dollison,194 U.S. 445, 450-451, 24 S.Ct. 703, 48 L.Ed. 1062 (1904), and anticompetitive conduct, Nash v. United States,229 U.S. 373, 376-378, 33 S.Ct. 780, 57 L.Ed. 1232 (1913); Waters-Pierce Oil Co. v. Texas (No. 1),212 U.S. 86, 108-111, 29 S.Ct. 220, 53 L.Ed. 417 (1909).

*2570In 1914, however, the Court ified a law on vagueness grounds under the Due Process Clause for the first time. In International Harvester Co. of America v. Kentucky,234 U.S. 216, 34 S.Ct. 853, 58 L.Ed. 1284 (1914), a tobacco company brought a Fourteenth Amendment challenge against several Kentucky antitrust laws that had been construed to render unlawful "any combination [made] ... for the purpose or with the effect of fixing a price that was greater or less than the real value of the article," id.,at 221, 34 S.Ct. 853. The company argued that by referring to "real value," the laws provided "no standard of conduct that it is possible to know." Ibid.The Court agreed. Id.,at 223-224, 34 S.Ct. 853. Although it did not specify in that case which portion of the Fourteenth Amendment served as the basis for its holding, ibid.,it explained in a related case that the lack of a knowable standard of conduct in the Kentucky statutes "violated the fundamental principles of justice embraced in the conception of due process of law." Collins v. Kentucky,234 U.S. 634, 638, 34 S.Ct. 924, 58 L.Ed. 1510 (1914).

3

Since that time, the Court's application of its vagueness doctrine has largely mirrored its application of substantive due process. During the Lochnerera, a period marked by the use of substantive due process to strike down economic regulations, e.g., Lochner v. New York,198 U.S. 45, 57, 25 S.Ct. 539, 49 L.Ed. 937 (1905), the Court frequently used the vagueness doctrine to invalidate economic regulations penalizing commercial activity.4Among the penal laws it found to be impermissibly vague were a state law regulating the production of crude oil, Champlin Refining Co. v. Corporation Comm'n of Okla.,286 U.S. 210, 242-243, 52 S.Ct. 559, 76 L.Ed. 1062 (1932), a state antitrust law, Cline,274 U.S., at 453-465, 47 S.Ct. 681, a state minimum-wage law, Connally,269 U.S., at 390-395, 46 S.Ct. 126, and a federal price-control statute, L. Cohen Grocery Co.,255 U.S., at 89-93, 41 S.Ct. 298.5

*2571Around the time the Court began shifting the focus of its substantive due process (and equal protection) jurisprudence from economic interests to "discrete and insular minorities," see United States v. Carolene Products Co.,304 U.S. 144, 153, n. 4, 58 S.Ct. 778, 82 L.Ed. 1234 (1938), the target of its vagueness doctrine changed as well. The Court began to use the vagueness doctrine to invalidate noneconomic regulations, such as state statutes penalizing obscenity, Winters,333 U.S., at 517-520, 68 S.Ct. 665, and membership in a gang, Lanzetta,306 U.S., at 458, 59 S.Ct. 618.

Successful vagueness challenges to regulations penalizing commercial conduct, by contrast, largely fell by the wayside. The Court, for instance, upheld a federal regulation punishing the knowing violation of an order instructing drivers transporting dangerous chemicals to " 'avoid, so far as practicable ... driving into or through congested thoroughfares, places where crowds are assembled, street car tracks, tunnels, viaducts, and dangerous crossings,' " Boyce Motor Lines, Inc. v. United States,342 U.S. 337, 338-339, 343, 72 S.Ct. 329, 96 L.Ed. 367 (1952). And notwithstanding its earlier conclusion that an Oklahoma law requiring state employees and contractors to be paid " 'not less than the current rate of per diem wages in the locality where the work is performed' " was unconstitutionally vague, Connally, supra,at 393, 46 S.Ct. 126, the Court found sufficiently definite a federal law forbidding radio broadcasting companies from attempting to compel by threat or duress a licensee to hire " 'persons in excess of the number of employees needed by such licensee to perform actual services,' " United States v. Petrillo,332 U.S. 1, 3, 6-7, 67 S.Ct. 1538, 91 L.Ed. 1877 (1947).

In more recent times, the Court's substantive due process jurisprudence has focused on abortions, and our vagueness doctrine has played a correspondingly significant role. In fact, our vagueness doctrine served as the basis for the first draft of the majority opinion in Roe v. Wade,410 U.S. 113, 93 S.Ct. 705, 35 L.Ed.2d 147 (1973), on the theory that laws prohibiting all abortions save for those done "for the purpose of saving the life of the mother" forced abortionists to guess when this exception would apply on penalty of conviction. See B. Schwartz, The Unpublished Opinions of the Burger Court 116-118 (1988) (reprinting first draft of Roe). Roe,of course, turned out as a substantive due process opinion. See 410 U.S., at 164, 93 S.Ct. 705. But since then, the Court has repeatedly deployed the vagueness doctrine to ify even mild regulations of the abortion industry. See Akron v. Akron Center for Reproductive Health, Inc.,462 U.S. 416, 451-452, 103 S.Ct. 2481, 76 L.Ed.2d 687 (1983)(ifying law requiring " 'that the remains of the unborn child [be] disposed of in a humane and sanitary manner' ");Colautti,439 U.S., at 381, 99 S.Ct. 675(ifying law mandating abortionists adhere to a prescribed standard of care if "there is 'sufficient reason to believe that the fetus may be viable' ").6

*2572In one of our most recent decisions ifying a law on vagueness grounds, substantive due process was again lurking in the background. In Morales,a plurality of this Court insisted that "the freedom to loiter for innocent purposes is part of the 'liberty' protected by the Due Process Clause of the Fourteenth Amendment," 527 U.S., at 53, 119 S.Ct. 1849, a conclusion that colored its analysis that an ordinance prohibiting loitering was unconstitutionally indeterminate, see id.,at 55, 119 S.Ct. 1849("When vagueness permeates the text of" a penal law "infring[ing] on constitutionally protected rights," "it is subject to facial attack").

I find this history unsettling. It has long been understood that one of the problems with holding a statute "void for 'indefiniteness' " is that " 'indefiniteness' ... is itself an indefinite concept," Winters, supra,at 524, 68 S.Ct. 665(Frankfurter, J., dissenting), and we as a Court have a bad habit of using indefinite concepts-especially ones rooted in "due process"-to invalidate democratically enacted laws.

B

It is also not clear that our vagueness doctrine can be reconciled with the original understanding of the term "due process of law." Our traditional justification for this doctrine has been the need for notice: "A conviction fails to comport with due process if the statute under which it is obtained fails to provide a person of ordinary intelligence fair notice of what is prohibited." United States v. Williams,553 U.S. 285, 304, 128 S.Ct. 1830, 170 L.Ed.2d 650 (2008); accord, ante,at 2564. Presumably, that justification rests on the view expressed in Murray's Lessee v. Hoboken Land & Improvement Co.,18 How. 272, 15 L.Ed. 372 (1856), that "due process of law" constrains the legislative branch by guaranteeing "usages and modes of proceeding existing in the common and statute law of England, before the emigration of our ancestors, and which are shown not to have been unsuited to their civil and political condition by having been acted on by them after the settlement of this country," id.,at 277. That justification assumes further that providing "a person of ordinary intelligence [with] fair notice of what is prohibited," Williams, supra,at 304, 128 S.Ct. 1830, is one such usage or mode.7

To accept the vagueness doctrine as founded in our Constitution, then, one must reject the possibility "that the Due Process Clause requires only that our Government must proceed according to the 'law of the land'-that is, according to *2573written constitutional and statutory provisions," which may be all that the original meaning of this provision demands. Hamdi v. Rumsfeld,542 U.S. 507, 589, 124 S.Ct. 2633, 159 L.Ed.2d 578 (2004)(THOMAS, J., dissenting) (some internal quotation marks omitted); accord, Turner v. Rogers,564 U.S. ----, ----, 131 S.Ct. 2507, 2521, 180 L.Ed.2d 452 (2011)(THOMAS, J., dissenting). Although Murray's Lesseestated the contrary, 18 How., at 276, a number of scholars and jurists have concluded that "considerable historical evidence supports the position that 'due process of law' was a separation-of-powers concept designed as a safeguard against unlicensed executive action, forbidding only deprivations not authorized by legislation or common law." D. Currie, The Constitution in the Supreme Court: The First Hundred Years 1789-1888, p. 272 (1985); see also, e.g., In re Winship,397 U.S. 358, 378-382, 90 S.Ct. 1068, 25 L.Ed.2d 368 (1970)(Black, J., dissenting). Others have disagreed. See, e.g., Chapman & McConnell, Due Process as Separation of Powers, 121 Yale L.J. 1672, 1679 (2012)(arguing that, as originally understood, "the principle of due process" required, among other things, that "statutes that purported to empower the other branches to deprive persons of rights without adequate procedural guarantees [be] subject to judicial review").

I need not choose between these two understandings of "due process of law" in this case. Justice ALITO explains why the majority's decision is wrong even under our precedents. See post, at 2580 - 2583 (dissenting opinion). And more generally, I adhere to the view that " '[i]f any fool would know that a particular category of conduct would be within the reach of the statute, if there is an unmistakable core that a reasonable person would know is forbidden by the law, the enactment is not unconstitutional on its face,' " Morales, supra,at 112, 119 S.Ct. 1849(THOMAS, J., dissenting), and there is no question that ACCA's residual clause meets that description, see ante,at 2568 (agreeing with the Government that "there will be straightforward cases under the residual clause").

* * *

I have no love for our residual clause jurisprudence: As I observed when we first got into this business, the Sixth Amendment problem with allowing district courts to conduct factfinding to determine whether an offense is a "violent felony" made our attempt to construe the residual clause " 'an unnecessary exercise.' " James,550 U.S., at 231, 127 S.Ct. 1586(THOMAS, J., dissenting). But the Court rejected my argument, choosing instead to begin that unnecessary exercise. I see no principled way that, four cases later, the Court can now declare that the residual clause has become too indeterminate to apply. Having damaged the residual clause through our misguided jurisprudence, we have no right to send this provision back to Congress and ask for a new one. I cannot join the Court in using the Due Process Clause to ify an Act of Congress that contains an unmistakable core of forbidden conduct, and I concur only in its judgment.

Justice ALITO, dissenting.

The Court is tired of the Armed Career Criminal Act of 1984 (ACCA) and in particular its residual clause. Anxious to rid our docket of bothersome residual clause cases, the Court is willing to do what it takes to get the job done. So brushing aside stare decisis, the Court holds that the residual clause is unconstitutionally vague even though we have twice rejected that very argument within the last eight years. The canons of interpretation get no greater respect. Inverting the canon that *2574a statute should be construed if possible to avoid unconstitutionality, the Court rejects a reasonable construction of the residual clause that would avoid any vagueness problems, preferring an alternative that the Court finds to be unconstitutionally vague. And the Court is not stopped by the well-established rule that a statute is void for vagueness only if it is vague in all its applications. While conceding that some applications of the residual clause are straightforward, the Court holds that the clause is now void in its entirety. The Court's determination to be done with residual clause cases, if not its fidelity to legal principles, is impressive.

I

A

Petitioner Samuel Johnson (unlike his famous namesake) has led a life of crime and violence. His presentence investigation report sets out a résumé of petty and serious crimes, beginning when he was 12 years old. Johnson's adult record includes convictions for, among other things, robbery, attempted robbery, illegal possession of a sawed-off shotgun, and a drug offense.

In 2010, the Federal Bureau of Investigation (FBI) began monitoring Johnson because of his involvement with the National Socialist Movement, a white-supremacist organization suspected of plotting acts of terrorism. In June of that year, Johnson left the group and formed his own radical organization, the Aryan Liberation Movement, which he planned to finance by counterfeiting United States currency. In the course of the Government's investigation, Johnson "disclosed to undercover FBI agents that he manufactured napalm, silencers, and other explosives for" his new organization. 526 Fed.Appx. 708, 709 (C.A.8 2013)(per curiam). He also showed the agents an AK-47 rifle, a semiautomatic rifle, a semiautomatic pistol, and a cache of approximately 1,100 rounds of ammunition. Later, Johnson told an undercover agent: "You know I'd love to assassinate some ... hoodrats as much as the next guy, but I think we really got to stick with high priority targets." Revised Presentence Investigation Report (PSR) ¶ 15. Among the top targets that he mentioned were "the Mexican consulate," "progressive bookstores," and individuals he viewed as "liberals." PSR ¶ 16.

In April 2012, Johnson was arrested, and he was subsequently indicted on four counts of possession of a firearm by a felon and two counts of possession of ammunition by a felon, in violation of 18 U.S.C. §§ 922(g)and § 924(e). He pleaded guilty to one of the firearms counts, and the District Court sentenced him to the statutory minimum of 15 years' imprisonment under ACCA, based on his prior felony convictions for robbery, attempted robbery, and illegal possession of a sawed-off shotgun.

B

ACCA provides a mandatory minimum sentence for certain violations of § 922(g), which prohibits the shipment, transportation, or possession of firearms or ammunition by convicted felons, persons previously committed to a mental institution, and certain others. Federal law normally provides a maximum sentence of 10 years' imprisonment for such crimes. See § 924(a)(2). Under ACCA, however, if a defendant convicted under § 922(g)has three prior convictions "for a violent felony or a serious drug offense," the sentencing court must impose a sentence of at least 15 years' imprisonment. § 924(e)(1).

ACCA's definition of a "violent felony" has three parts. First, a felony qualifies if it "has as an element the use, attempted use, or threatened use of physical force *2575against the person of another." § 924(e)(2)(B)(i). Second, the Act specifically names four categories of qualifying felonies: burglary, arson, extortion, and offenses involving the use of explosives. See § 924(e)(2)(B)(ii). Third, the Act contains what we have called a "residual clause," which reaches any felony that "otherwise involves conduct that presents a serious potential risk of physical injury to another." Ibid.

The present case concerns the residual clause. The sole question raised in Johnson's certiorari petition was whether possession of a sawed-off shotgun under Minnesota law qualifies as a violent felony under that clause. Although Johnson argued in the lower courts that the residual clause is unconstitutionally vague, he did not renew that argument here. Nevertheless, after oral argument, the Court raised the question of vagueness on its own. The Court now holds that the residual clause is unconstitutionally vague in all its applications. I cannot agree.

II

I begin with stare decisis. Eight years ago in James v. United States,550 U.S. 192, 127 S.Ct. 1586, 167 L.Ed.2d 532 (2007), Justice SCALIA, the author of today's opinion for the Court, fired an opening shot at the residual clause. In dissent, he suggested that the residual clause is void for vagueness. Id., at 230, 127 S.Ct. 1586. The Court held otherwise, explaining that the standard in the residual clause "is not so indefinite as to prevent an ordinary person from understanding" its scope. Id., at 210, n. 6, 127 S.Ct. 1586.

Four years later, in Sykes v. United States,564 U.S. 1, 131 S.Ct. 2267, 180 L.Ed.2d 60 (2011), Justice SCALIA fired another round. Dissenting once again, he argued that the residual clause is void for vagueness and rehearsed the same basic arguments that the Court now adopts. See id.,at ---- - ----, 131 S.Ct., at 2273-2274; see also Derby v. United States,564 U.S. ----, ---- - ----, 131 S.Ct. 2858, 2859-2860, 180 L.Ed.2d 904 (2011)(SCALIA, J., dissenting from denial of certiorari). As in James,the Court rejected his arguments. See Sykes,564 U.S., at ----, 131 S.Ct., at 2276-2277. In fact, Justice SCALIA was the onlyMember of the SykesCourt who took the position that the residual clause could not be intelligibly applied to the offense at issue. The opinion of the Court, which five Justices joined, expressly held that the residual clause "states an intelligible principle and provides guidance that allows a person to 'conform his or her conduct to the law.' " Id.,at ---- - ----, 131 S.Ct., at 2277(quoting Chicago v. Morales,527 U.S. 41, 58, 119 S.Ct. 1849, 144 L.Ed.2d 67 (1999)(plurality opinion)). Justice THOMAS's concurrence, while disagreeing in part with the Court's interpretation of the residual clause, did not question its constitutionality. See Sykes,564 U.S., at ----, 131 S.Ct., at ---- (opinion concurring in judgment). And Justice KAGAN's dissent, which Justice GINSBURG joined, argued that a proper application of the provision required a different result. See id.,at ----, 131 S.Ct., at ----. Thus, eight Members of the Court found the statute capable of principled application.

It is, of course, true that "[s]tare decisisis not an inexorable command." Payne v. Tennessee,501 U.S. 808, 828, 111 S.Ct. 2597, 115 L.Ed.2d 720 (1991). But neither is it an empty Latin phrase. There must be good reasons for overruling a precedent, and there is none here. Nothing has changed since our decisions in Jamesand Sykes-nothing, that is, except the Court's weariness with ACCA cases.

Reprising an argument that Justice SCALIA made to no avail in *2576Sykes, supra,at ----, 131 S.Ct., at 2287(dissenting opinion), the Court reasons that the residual clause must be unconstitutionally vague because we have had trouble settling on an interpretation. See ante,at 2558 - 2559. But disagreement about the meaning and application of the clause is not new. We were divided in Jamesand in Sykesand in our intervening decisions in Begay v. United States,553 U.S. 137, 128 S.Ct. 1581, 170 L.Ed.2d 490 (2008), and Chambers v. United States, 555 U.S. 122, 129 S.Ct. 687, 172 L.Ed.2d 484 (2009). And that pattern is not unique to ACCA; we have been unable to come to an agreement on many recurring legal questions. The Confrontation Clause is one example that comes readily to mind. See, e.g., Williams v. Illinois,567 U.S. ----, 132 S.Ct. 2221, 183 L.Ed.2d 89 (2012); Bullcoming v. New Mexico,564 U.S. ----, 131 S.Ct. 2705, 180 L.Ed.2d 610 (2011); Melendez-Diaz v. Massachusetts,557 U.S. 305, 129 S.Ct. 2527, 174 L.Ed.2d 314 (2009). Our disagreements about the meaning of that provision do not prove that the Confrontation Clause has no ascertainable meaning. Likewise, our disagreements on the residual clause do not prove that it is unconstitutionally vague.

The Court also points to conflicts in the decisions of the lower courts as proof that the statute is unconstitutional. See ante, at 2559 - 2560. The Court overstates the degree of disagreement below. For many crimes, there is no dispute that the residual clause applies. And our certiorari docket provides a skewed picture because the decisions that we are asked to review are usually those involving issues on which there is at least an arguable circuit conflict. But in any event, it has never been thought that conflicting interpretations of a statute justify judicial elimination of the statute. One of our chief responsibilities is to resolve those disagreements, see Supreme Court Rule 10, not to strike down the laws that create this work.

The Court may not relish the task of resolving residual clause questions on which the Circuits disagree, but the provision has not placed a crushing burden on our docket. In the eight years since James,we have decided all of three cases involving the residual clause. See Begay, supra; Chambers, supra; Sykes, supra. Nevertheless, faced with the unappealing prospect of resolving more circuit splits on various residual clause issues, see ante, at 2559, six Members of the Court have thrown in the towel. That is not responsible.

III

Even if we put stare decisisaside, the Court's decision remains indefensible. The residual clause is not unconstitutionally vague.

A

The Fifth Amendment prohibits the enforcement of vague criminal laws, but the threshold for declaring a law void for vagueness is high. "The strong presumptive validity that attaches to an Act of Congress has led this Court to hold many times that statutes are not automatically invalidated as vague simply because difficulty is found in determining whether certain marginal offenses fall within their language." United States v. National Dairy Products Corp.,372 U.S. 29, 32, 83 S.Ct. 594, 9 L.Ed.2d 561 (1963). Rather, it is sufficient if a statute sets out an "ascertainable standard." United States v. L. Cohen Grocery Co.,255 U.S. 81, 89, 41 S.Ct. 298, 65 L.Ed. 516 (1921). A statute is thus void for vagueness only if it wholly "fails to provide a person of ordinary intelligence fair notice of what is prohibited, or is so standardless that it authorizes or encourages seriously discriminatory enforcement."

*2577United States v. Williams,553 U.S. 285, 304, 128 S.Ct. 1830, 170 L.Ed.2d 650 (2008).

The bar is even higher for sentencing provisions. The fair notice concerns that inform our vagueness doctrine are aimed at ensuring that a " 'person of ordinary intelligence [has] a reasonable opportunity to know what is prohibited, so that he may act accordingly.' " Hoffman Estates v. Flipside, Hoffman Estates, Inc.,455 U.S. 489, 498, 102 S.Ct. 1186, 71 L.Ed.2d 362 (1982)(quoting Grayned v. City of Rockford,408 U.S. 104, 108, 92 S.Ct. 2294, 33 L.Ed.2d 222 (1972)). The fear is that vague laws will " 'trap the innocent.' " 455 U.S., at 498, 102 S.Ct. 1186. These concerns have less force when it comes to sentencing provisions, which come into play only after the defendant has been found guilty of the crime in question. Due process does not require, as Johnson oddly suggests, that a "prospective criminal" be able to calculate the precise penalty that a conviction would bring. Supp. Brief for Petitioner 5; see Chapman v. United States,500 U.S. 453, 467-468, 111 S.Ct. 1919, 114 L.Ed.2d 524 (1991)(concluding that a vagueness challenge was "particularly" weak "since whatever debate there is would center around the appropriate sentence and not the criminality of the conduct").

B

ACCA's residual clause unquestionably provides an ascertainable standard. It defines "violent felony" to include any offense that "involves conduct that presents a serious potential risk of physical injury to another." 18 U.S.C. § 924(e)(2)(B)(ii). That language is by no means incomprehensible. Nor is it unusual. There are scores of federal and state laws that employ similar standards. The Solicitor General's brief contains a 99-page appendix setting out some of these laws. See App. to Supp. Brief for United States; see also James, supra,at 210, n. 6, 127 S.Ct. 1586. If all these laws are unconstitutionally vague, today's decision is not a blast from a sawed-off shotgun; it is a nuclear explosion.

Attempting to avoid such devastation, the Court distinguishes these laws primarily on the ground that almost all of them "require gauging the riskiness of conduct in which an individual defendant engages on a particular occasion." Ante,at 2561 (emphasis in original). The Court thus admits that, "[a]s a general matter, we do not doubt the constitutionality of laws that call for the application of a qualitative standard such as 'substantial risk' to real-world conduct." Ibid.Its complaint is that the residual clause "requires application of the 'serious potential risk' standard to an idealized ordinary case of the crime." Ibid. (emphasis added). Thus, according to the Court, ACCA's residual clause is unconstitutionally vague because its standard must be applied to "an idealized ordinary case of the crime" and not, like the vast majority of the laws in the Solicitor General's appendix, to "real-world conduct."

ACCA, however, makes no reference to "an idealized ordinary case of the crime." That requirement was the handiwork of this Court in Taylor v. United States,495 U.S. 575, 110 S.Ct. 2143, 109 L.Ed.2d 607 (1990). And as I will show, the residual clause can reasonably be interpreted to refer to "real-world conduct."1

*2578C

When a statute's constitutionality is in doubt, we have an obligation to interpret the law, if possible, to avoid the constitutional problem. See, e.g., Edward J. DeBartolo Corp. v. Florida Gulf Coast Building & Constr. Trades Council,485 U.S. 568, 575, 108 S.Ct. 1392, 99 L.Ed.2d 645 (1988). As one treatise puts it, "[a] statute should be interpreted in a way that avoids placing its constitutionality in doubt." A. Scalia & B. Garner, Reading Law: The Interpretation of Legal Texts § 38, p. 247 (2012). This canon applies fully when considering vagueness challenges. In cases like this one, "our task is not to destroy the Act if we can, but to construe it, if consistent with the will of Congress, so as to comport with constitutional limitations." Civil Service Comm'n v. Letter Carriers, 413 U.S. 548, 571, 93 S.Ct. 2880, 37 L.Ed.2d 796 (1973); see also Skilling v. United States,561 U.S. 358, 403, 130 S.Ct. 2896, 177 L.Ed.2d 619 (2010). Indeed, " '[t]he elementary rule is that every reasonable construction must be resorted to, in order to save a statute from unconstitutionality.' " Id.,at 406, 130 S.Ct. 2896(quoting Hooper v. California,155 U.S. 648, 657, 15 S.Ct. 207, 39 L.Ed. 297 (1895); emphasis deleted); see also Ex parte Randolph,20 F.Cas. 242, 254 (No. 11,558) (C.C.Va.1833) (Marshall, C.J.).

The Court all but concedes that the residual clause would be constitutional if it applied to "real-world conduct." Whether that is the bestinterpretation of the residual clause is beside the point. What matters is whether it is a reasonable interpretation of the statute. And it surely is that.

First, this interpretation heeds the pointed distinction that ACCA draws between the "element[s]" of an offense and "conduct." Under § 924(e)(2)(B)(i), a crime qualifies as a "violent felony" if one of its "element [s]" involves "the use, attempted use, or threatened use of physical force against the person of another." But the residual clause, which appears in the very next subsection, § 924(e)(2)(B)(ii), focuses on "conduct"-specifically, "conduct that presents a serious potential risk of physical injury to another." The use of these two different terms in § 924(e)indicates that "conduct" refers to things done during the commission of an offense that are not part of the elements needed for conviction. Because those extra actions vary from case to case, it is natural to interpret "conduct" to mean real-world conduct, not the conduct involved in some Platonic ideal of the offense.

Second, as the Court points out, standards like the one in the residual clause almost always appear in laws that call for application by a trier of fact. This strongly suggests that the residual clause calls for the same sort of application.

Third, if the Court is correct that the residual clause is nearly incomprehensible when interpreted as applying to an "idealized ordinary case of the crime," then that is telling evidence that this is not what Congress intended. When another interpretation is ready at hand, why should we assume that Congress gave the clause a meaning that is impossible-or even, exceedingly difficult-to apply?

D

Not only does the "real-world conduct" interpretation fit the terms of the residual *2579clause, but the reasons that persuaded the Court to adopt the categorical approach in Tayloreither do not apply or have much less force in residual clause cases.

In Taylor,the question before the Court concerned the meaning of "burglary," one of ACCA's enumerated offenses. The Court gave three reasons for holding that a judge making an ACCA determination should generally look only at the elements of the offense of conviction and not to other things that the defendant did during the commission of the offense. First, the Court thought that ACCA's use of the term "convictions" pointed to the categorical approach. The Court wrote: "Section 924(e)(1)refers to 'a person who ... has three previous convictions' for-not a person who has committed-three previous violent felonies or drug offenses." 495 U.S., at 600, 110 S.Ct. 2143. Second, the Court relied on legislative history, noting that ACCA had previously contained a generic definition of burglary and that "the deletion of [this] definition ... may have been an inadvertent casualty of a complex drafting process." Id.,at 589-590, 601, 110 S.Ct. 2143. Third, the Court felt that "the practical difficulties and potential unfairness of a factual approach [were] daunting." Id., at 601, 110 S.Ct. 2143.

None of these three grounds dictates that the categorical approach must be used in residual clause cases. The second ground, which concerned the deletion of a generic definition of burglary, obviously has no application to the residual clause. And the first ground has much less force in residual clause cases. In Taylor,the Court reasoned that a defendant has a "conviction" for burglary only if burglary is the offense set out in the judgment of conviction. For instance, if a defendant commits a burglary but pleads guilty, under a plea bargain, to possession of burglar's tools, the TaylorCourt thought that it would be unnatural to say that the defendant had a convictionfor burglary. Now consider a case in which a gang member is convicted of illegal possession of a sawed-off shotgun and the evidence shows that he concealed the weapon under his coat, while searching for a rival gang member who had just killed his brother. In that situation, it is not at all unnatural to say that the defendant had a conviction for a crime that "involve[d] conductthat present[ed] a serious potential risk of physical injury to another." § 924(e)(2)(B)(ii)(emphasis added). At the very least, it would be a reasonable way to describe the defendant's conviction.

The TaylorCourt's remaining reasons for adopting the categorical approach cannot justify an interpretation that renders the residual clause unconstitutional. While the TaylorCourt feared that a conduct-specific approach would unduly burden the courts, experience has shown that application of the categorical approach has not always been easy. Indeed, the Court's main argument for overturning the statute is that this approach is unmanageable in residual clause cases.

As for the notion that the categorical approach is more forgiving to defendants, there is a strong argument that the opposite is true, at least with respect to the residual clause. Consider two criminal laws: Injury occurs in 10% of cases involving the violation of statute A, but in 90% of cases involving the violation of statute B. Under the categorical approach, a truly dangerous crime under statute A might not qualify as a violent felony, while a crime with no measurable risk of harm under statute B would count against the defendant. Under a conduct-specific inquiry, on the other hand, a defendant's actual conduct would determine whether ACCA's mandatory penalty applies.

*2580It is also significant that the allocation of the burden of proof protects defendants. The prosecution bears the burden of proving that a defendant has convictions that qualify for sentencing under ACCA. If evidentiary deficiencies, poor recordkeeping, or anything else prevents the prosecution from discharging that burden under the conduct-specific approach, a defendant would not receive an ACCA sentence.

Nor would a conduct-specific inquiry raise constitutional problems of its own. It is questionable whether the Sixth Amendment creates a right to a jury trial in this situation. See Almendarez-Torres v. United States,523 U.S. 224, 118 S.Ct. 1219, 140 L.Ed.2d 350 (1998). But if it does, the issue could be tried to a jury, and the prosecution could bear the burden of proving beyond a reasonable doubt that a defendant's prior crimes involved conduct that presented a serious potential risk of injury to another. I would adopt this alternative interpretation and hold that the residual clause requires an examination of real-world conduct.

The Court's only reason for refusing to consider this interpretation is that "the Government has not asked us to abandon the categorical approach in residual-clause cases." Ante,at 2562. But the Court cites no case in which we have suggested that a saving interpretation may be adopted only if it is proposed by one of the parties. Nor does the Court cite any secondary authorities advocating this rule. Cf. Scalia, Reading Law § 38 (stating the canon with no such limitation). On the contrary, we have long recognized that it is "our plain duty to adopt that construction which will save [a] statute from constitutional infirmity," where fairly possible. United States ex rel. Attorney General v. Delaware & Hudson Co.,213 U.S. 366, 407, 29 S.Ct. 527, 53 L.Ed. 836 (1909). It would be strange if we could fulfill that "plain duty" only when a party asks us to do so. And the Court's refusal to consider a saving interpretation not advocated by the Government is hard to square with the Court's adoption of an argument that petitioner chose not to raise. As noted, Johnson did not ask us to hold that the residual clause is unconstitutionally vague, but the Court interjected that issue into the case, requested supplemental briefing on the question, and heard reargument. The Court's refusal to look beyond the arguments of the parties apparently applies only to arguments that the Court does not want to hear.

E

Even if the categorical approach is used in residual clause cases, however, the clause is still not void for vagueness. "It is well established that vagueness challenges to statutes which do not involve First Amendment freedoms must be examined" on an as-applied basis. United States v. Mazurie,419 U.S. 544, 550, 95 S.Ct. 710, 42 L.Ed.2d 706 (1975). "Objections to vagueness under the Due Process Clause rest on the lack of notice, and hence may be overcome in any specific case where reasonable persons would know that their conduct is at risk." Maynard v. Cartwright,486 U.S. 356, 361, 108 S.Ct. 1853, 100 L.Ed.2d 372 (1988). Thus, in a due process vagueness case, we will hold that a law is facially invalid "only if the enactment is impermissibly vague in allof its applications." Hoffman Estates,455 U.S., at 494-495, 102 S.Ct. 1186(emphasis added); see also Chapman,500 U.S., at 467, 111 S.Ct. 1919.2

In concluding that the residual clause is facially void for vagueness, the Court flatly *2581contravenes this rule. The Court admits "that there will be straightforward cases under the residual clause." Ante,at 2560. But rather than exercising the restraint that our vagueness cases prescribe, the Court holds that the residual clause is unconstitutionally vague even when its application is clear.

The Court's treatment of this issue is startling. Its facial invalidation precludes a sentencing court that is applying ACCA from counting convictions for even those specific offenses that this Court previously found to fall within the residual clause. See James,550 U.S., at 203-209, 127 S.Ct. 1586(attempted burglary); Sykes,564 U.S., at ---- - ----, 131 S.Ct., at 2272-2275(flight from law enforcement in a vehicle). Still worse, the Court holds that vagueness bars the use of the residual clause in other cases in which its applicability can hardly be questioned. Attempted rape is an example. See, e.g.,Dawson v. United States,702 F.3d 347, 351-352 (C.A.6 2012). Can there be any doubt that "an idealized ordinary case of th[is] crime" "involves conduct that presents a serious potential risk of physical injury to another"? How about attempted arson,3attempted kidnapping,4solicitation to commit aggravated assault,5possession of a loaded weapon with the intent to use it unlawfully against another person,6possession of a weapon in prison,7or compelling a person to act as a prostitute?8Is there much doubt that those offenses "involve conduct that presents a serious potential risk of physical injury to another"?

Transforming vagueness doctrine, the Court claims that we have never actually held that a statute may be voided for vagueness only when it is vague in all its applications. But that is simply wrong. In Hoffman Estates,we reversed a Seventh Circuit decision that voided an ordinance prohibiting the sale of certain items. See 455 U.S., at 491, 102 S.Ct. 1186. The Seventh Circuit struck down the ordinance because it was "unclear in someof its applications," but we reversed and emphasized that a law is void for vagueness "only if [it] is impermissibly vague in all of its applications." Id.,at 494-495, 102 S.Ct. 1186; see also id.,at 495, n. 7, 102 S.Ct. 1186(collecting cases). Applying that principle, we held that the "facial challenge [wa]s unavailing" because "at least some of the items sold ... [we]re covered" by the *2582ordinance. Id.,at 500, 102 S.Ct. 1186. These statements were not dicta. They were the holding of the case. Yet the Court does not even mention this binding precedent.

Instead, the Court says that the facts of two earliercases support a broader application of the vagueness doctrine. See ante, at 2560 - 2561. That, too, is incorrect. Neither case remotely suggested that mere overbreadth is enough for facial invalidation under the Fifth Amendment.

In Coates v. Cincinnati,402 U.S. 611, 612, 91 S.Ct. 1686, 29 L.Ed.2d 214 (1971), we addressed an ordinance that restricted free assembly and association rights by prohibiting "annoying" conduct. Our analysis turned in large part on those First Amendment concerns. In fact, we specifically explained that the "vice of the ordinance lies not alone in its violation of the due process standard of vagueness." Id.,at 615, 91 S.Ct. 1686. In the present case, by contrast, no First Amendment rights are at issue. Thus, Coatescannot support the Court's rejection of our repeated statements that "vagueness challenges to statutes which do not involve First Amendment freedomsmust be examined in light of the facts ... at hand." Mazurie, supra,at 550, 95 S.Ct. 710(emphasis added).

Likewise, L. Cohen Grocery Co.,255 U.S. 81, 41 S.Ct. 298, 65 L.Ed. 516, proves precisely the opposite of what the Court claims. In that case, we struck down a statute prohibiting " 'unjust or unreasonable rate[s]' " because it provided no "ascertainable standard of guilt" and left open "the widest conceivable inquiry, the scope of which no one can foresee and the result of which no one can foreshadow or adequately guard against." Id.,at 89, 41 S.Ct. 298. The clear import of this language is that the law at issue was impermissibly vague in all applications. And in the years since, we have never adopted the majority's contradictory interpretation. On the contrary, we have characterized the case as involving a statute that could "not constitutionally be applied to any set of facts." United States v. Powell,423 U.S. 87, 92, 96 S.Ct. 316, 46 L.Ed.2d 228 (1975). Thus, our holdings and our dicta prohibit the Court's expansion of the vagueness doctrine. The Constitution does not allow us to hold a statute void for vagueness unless it is vague in all its applications.

IV

Because I would not strike down ACCA's residual clause, it is necessary for me to address whether Johnson's conviction for possessing a sawed-off shotgun qualifies as a violent felony. Under either the categorical approach or a conduct-specific inquiry, it does.

A

The categorical approach requires us to determine whether "the conduct encompassed by the elements of the offense, in the ordinary case, presents a serious potential risk of injury to another." James,550 U.S., at 208, 127 S.Ct. 1586. This is an "inherently probabilistic" determination that considers the circumstances and conduct that ordinarily attend the offense. Id.,at 207, 127 S.Ct. 1586. The mere fact that a crime couldbe committed without a risk of physical harm does not exclude it from the statute's reach. See id.,at 207-208, 127 S.Ct. 1586. Instead, the residual clause speaks of "potential risk[s]," § 924(e)(2)(B)(ii), a term suggesting "that Congress intended to encompass possibilities even more contingent or remote than a simple 'risk,' much less a certainty." James, supra, at 207-208, 127 S.Ct. 1586.

Under these principles, unlawful possession of a sawed-off shotgun qualifies as a violent felony. As we recognized in District of Columbia v. Heller,554 U.S. 570, 625, 128 S.Ct. 2783, 171 L.Ed.2d 637 (2008), sawed-off shotguns are "not typically possessed by law-abiding citizens for *2583lawful purposes." Instead, they are uniquely attractive to violent criminals. Much easier to conceal than long-barreled shotguns used for hunting and other lawful purposes, short-barreled shotguns can be hidden under a coat, tucked into a bag, or stowed under a car seat. And like a handgun, they can be fired with one hand-except to more lethal effect. These weapons thus combine the deadly characteristics of conventional shotguns with the more convenient handling of handguns. Unlike those common firearms, however, they are not typically possessed for lawful purposes. And when a person illegally possesses a sawed-off shotgun during the commission of a crime, the risk of violence is seriously increased. The ordinary case of unlawful possession of a sawed-off shotgun therefore "presents a serious potential risk of physical injury to another." § 922(e)(2)(B)(ii).

Congress' treatment of sawed-off shotguns confirms this judgment. As the Government's initial brief colorfully recounts, sawed-off shotguns were a weapon of choice for gangsters and bank robbers during the Prohibition Era. See Brief for United States 4.9In response, Congress enacted the National Firearms Act of 1934, which required individuals possessing certain especially dangerous weapons-including sawed-off shotguns-to register with the Federal Government and pay a special tax. 26 U.S.C. §§ 5845(a)(1)-(2). The Act was passed on the understanding that "while there is justification for permitting the citizen to keep a pistol or revolver for his own protection without any restriction, there is no reason why anyone except a law officer should have a ... sawed-off shotgun." H.R.Rep. No. 1780, 73d Cong., 2d Sess., 1 (1934). As amended, the Act imposes strict registration requirements for any individual wishing to possess a covered shotgun, see, e.g.,§§ 5822, 5841(b), and illegal possession of such a weapon is punishable by imprisonment for up to 10 years. See §§ 5861(b)-(d), 5871. It is telling that this penalty exceeds that prescribed by federal law for quintessential violent felonies.10It thus seems perfectly clear that Congress has long regarded the illegal possession of a sawed-off shotgun as a crime that poses a serious risk of harm to others.

The majority of States agree. The Government informs the Court, and Johnson does not dispute, that 28 States have followed Congress' lead by making it a crime to possess an unregistered sawed-off shotgun, and 11 other States and the District of Columbia prohibit private possession of sawed-off shotguns entirely. See Brief for United States 8-9 (collecting statutes). Minnesota, where petitioner was convicted, *2584has adopted a blanket ban, based on its judgment that "[t]he sawed-off shotgun has no legitimate use in the society whatsoever." State v. Ellenberger,543 N.W.2d 673, 676 (Minn.App.1996)(internal quotation marks and citation omitted). Possession of a sawed-off shotgun in Minnesota is thus an inherently criminal act. It is fanciful to assume that a person who chooses to break the law and risk the heavy criminal penalty incurred by possessing a notoriously dangerous weapon is unlikely to use that weapon in violent ways.

B

If we were to abandon the categorical approach, the facts of Johnson's offense would satisfy the residual clause as well. According to the record in this case, Johnson possessed his sawed-off shotgun while dealing drugs. When police responded to reports of drug activity in a parking lot, they were told by two people that "Johnson and another individual had approached them and offered to sell drugs." PSR ¶ 45. The police then searched the vehicle where Johnson was seated as a passenger, and they found a sawed-off shotgun and five bags of marijuana. Johnson admitted that the gun was his.

Understood in this context, Johnson's conduct posed an acute risk of physical injury to another. Drugs and guns are never a safe combination. If one of his drug deals had gone bad or if a rival dealer had arrived on the scene, Johnson's deadly weapon was close at hand. The sawed-off nature of the gun elevated the risk of collateral damage beyond any intended targets. And the location of the crime-a public parking lot-significantly increased the chance that innocent bystanders might be caught up in the carnage. This is not a case of "mere possession" as Johnson suggests. Brief for Petitioner i. He was not storing the gun in a safe, nor was it a family heirloom or collector's item. He illegally possessed the weapon in case he needed to use it during another crime. A judge or jury could thus conclude that Johnson's offense qualified as a violent felony.

There should be no doubt that Samuel Johnson was an armed career criminal. His record includes a number of serious felonies. And he has been caught with dangerous weapons on numerous occasions. That this case has led to the residual clause's demise is confounding. I only hope that Congress can take the Court at its word that either amending the list of enumerated offenses or abandoning the categorical approach would solve the problem that the Court perceives.

*14Justice O’Connor

announced the judgment of the Court and delivered an opinion, in which The Chief Justice and Justice Kennedy join.

In this case, we decide whether the Eighth Amendment prohibits the State of California from sentencing a repeat felon to a prison term of 25 years to life under the State’s “Three Strikes and You’re Out” law.

r — <

A

California’s three strikes law reflects a shift in the State’s sentencing policies toward incapacitating and deterring repeat offenders who threaten the public safety. The law was designed “to ensure longer prison sentences and greater punishment for those who commit a felony and have been previously convicted of serious and/or violent felony offenses.” Cal. Penal Code Ann. § 667(b) (West 1999). On March 3, 1993, California Assemblymen Bill Jones and Jim Costa introduced Assembly Bill 971, the legislative version of what would later become the three strikes law. The Assembly Committee on Public Safety defeated the bill only weeks later. Public outrage over the defeat sparked a voter initiative to add Proposition 184, based loosely on the bill, to the ballot in the November 1994 general election.

On October 1, 1993, while Proposition 184 was circulating, 12-year-old Polly Klaas was kidnaped from her home in Petaluma, California. Her admitted killer, Richard Allen Davis, had a long criminal history that included two prior kidnaping convictions. Davis had served only half of his *15most recent sentence (16 years for kidnaping, assault, and burglary). Had Davis served his entire sentence, he would still have been in prison on the day that Polly Klaas was kidnaped.

Polly Klaas’ murder galvanized support for the three strikes initiative. Within days, Proposition 184 was on its way to becoming the fastest qualifying initiative in California history. On January 3, 1994, the sponsors of Assembly Bill 971 resubmitted an amended version of the bill that conformed to Proposition 184. On January 31,1994, Assembly Bill 971 passed the Assembly by a 63 to 9 margin. The Senate passed it by a 29 to 7 margin on March 3, 1994. Governor Pete Wilson signed the bill into law on March 7, 1994. California voters approved Proposition 184 by a margin of 72 to 28 percent on November 8, 1994.

California thus became the second State to enact a three strikes law. In November 1993, the voters of Washington State approved their own three strikes law, Initiative 593, by a margin of 3 to 1. U. S. Dept. of Justice, National Institute of Justice, J. Clark, J. Austin, & D. Henry, “Three Strikes and You’re Out”: A Review of State Legislation 1 (Sept. 1997) (hereinafter Review of State Legislation). Between 1993 and 1995, 24 States and the Federal Government enacted three strikes laws. Ibid. Though the three strikes laws vary from State to State, they share a common goal of protecting the public safety by providing lengthy prison terms for habitual felons.

B

California’s current three strikes law consists of two virtually identical statutory schemes “designed to increase the prison terms of repeat felons.” People v. Superior Court of San Diego Cty. ex rel. Romero, 13 Cal. 4th 497, 504, 917 P. 2d 628, 630 (1996) (Romero). When a defendant is convicted of a felony, and he has previously been convicted of one or more prior felonies defined as “serious” or “violent” in Cal. Penal Code Ann. §§ 667.5 and 1192.7 (West Supp. 2002), sentencing *16is conducted pursuant to the three strikes law. Prior convictions must be alleged in the charging document, and the defendant has a right to a jury determination that the prosecution has proved the prior convictions beyond a reasonable doubt. § 1025; § 1158 (West 1985).

If the defendant has one prior “serious” or “violent” felony conviction, he must be sentenced to “twice the term otherwise provided as punishment for the current felony conviction.” § 667(e)(1) (West 1999); § 1170.12(c)(1) (West Supp. 2002). If the defendant has two or more prior “serious” or “violent” felony convictions, he must receive “an indeterminate term of life imprisonment.” § 667(e)(2)(A) (West 1999); § 1170.12(c)(2)(A) (West Supp. 2002). Defendants sentenced to life under the three strikes law become eligible for parole on a date calculated by reference to a “minimum term,” which is the greater of (a) three times the term otherwise provided for the current conviction, (b) 25 years, or (c) the term determined by the court pursuant to §1170 for the underlying conviction, including any enhancements. §§667(e)(2)(A)(i)-(iii) (West 1999); §§ 1170.12(c)(2)(A)(i)-(iii) (West Supp. 2002).

Under California law, certain offenses may be classified as either felonies or misdemeanors. These crimes are known as “wobblers.” Some crimes that would otherwise be misdemeanors become “wobblers” because of the defendant’s prior record. For example, petty theft, a misdemeanor, becomes a “wobbler” when the defendant has previously served a prison term for committing specified theft-related crimes. §490 (West 1999); §666 (West Supp. 2002). Other crimes, such as grand theft, are “wobblers” regardless of the defendant’s prior record. See § 489(b) (West 1999). Both types of “wobblers” are triggering offenses under the three strikes law only when they are treated as felonies. Under California law, a “wobbler” is presumptively a felony and “remains a felony except when the discretion is actually exercised” to make the crime a misdemeanor. People v. Wil*17liams, 27 Cal. 2d 220, 229, 163 P. 2d 692, 696 (1945) (emphasis deleted and internal quotation marks omitted).

In California, prosecutors may exercise their discretion to charge a “wobbler” as either a felony or a misdemeanor. Likewise, California trial courts have discretion to reduce a “wobbler” charged as a felony to a misdemeanor either before preliminary examination or at sentencing to avoid imposing a three strikes sentence. Cal. Penal Code Ann. §§ 17(b)(5), 17(b)(1) (West 1999); People v. Superior Court of Los Angeles Cty. ex rel. Alvarez, 14 Cal. 4th 968, 978, 928 P. 2d 1171, 1177-1178 (1997). In exercising this discretion, the court may consider “those factors that direct similar sentencing decisions,” such as “the nature and circumstances of the offense, the defendant’s appreciation of and attitude toward the offense,... [and] the general objectives of sentencing.” Ibid, (internal quotation marks and citations omitted).

California trial courts can also vacate allegations of prior “serious” or “violent” felony convictions, either on motion by the prosecution or sua sponte. Romero, supra, at 529-530, 917 P. 2d, at 647-648. In ruling whether to vacate allegations of prior felony convictions, courts consider whether, “in light of the nature and circumstances of [the defendant’s] present felonies and prior serious and/or violent felony convictions, and the particulars of his background, character, and prospects, the defendant may be deemed outside the [three strikes’] scheme’s spirit, in whole or in part.” People v. Williams, 17 Cal. 4th 148, 161, 948 P. 2d 429, 437 (1998). Thus, trial courts may avoid imposing a three strikes sentence in two ways: first, by reducing “wobblers” to misdemeanors (which do not qualify as triggering offenses), and second, by vacating allegations of prior “serious” or “violent” felony convictions.

C

On parole from a 9-year prison term, petitioner Gary Ewing walked into the pro shop of the El Segundo Golf *18Course in Los Angeles County on March 12, 2000. He walked out with three golf clubs, priced at $399 apiece, concealed in his pants leg. A shop employee, whose suspicions were aroused when he observed Ewing limp out of the pro shop, telephoned the police. The police apprehended Ewing in the parking lot.

Ewing is no stranger to the criminal justice system. In 1984, at the age of 22, he pleaded guilty to theft. The court sentenced him to six months in jail (suspended), three years’ probation, and a $300 fine. In 1988, he was convicted of felony grand theft auto and sentenced to one year in jail and three years’ probation. After Ewing completed probation, however, the sentencing court reduced the crime to a misdemeanor, permitted Ewing to withdraw his guilty plea, and dismissed the case. In 1990, he was convicted of petty theft with a prior and sentenced to 60 days in the county jail and three years’ probation. In 1992, Ewing was convicted of battery and sentenced to 30 days in the county jail and two years’ summary probation. One month later, he was convicted of theft and sentenced to 10 days in the county jail and 12 months’ probation. In January 1993, Ewing was convicted of burglary and sentenced to 60 days in the county jail and one year’s summary probation. In February 1993, he was convicted of possessing drug paraphernalia and sentenced to six months in the county jail and three years’ probation. In July 1993, he was convicted of appropriating lost property and sentenced to 10 days in the county jail and two years’ summary probation. In September 1993, he was convicted of unlawfully possessing a firearm and trespassing and sentenced to 30 days in the county jail and one year’s probation.

In October and November 1993, Ewing committed three burglaries and one robbery at a Long Beach, California, apartment complex over a 5-week period. He awakened one of his victims, asleep on her living room sofa, as he tried to disconnect her video cassette recorder from the television in *19that room. When she screamed, Ewing ran out the front door. On another occasion, Ewing accosted a victim in the mailroom of the apartment complex. Ewing claimed to have a gun and ordered the victim to hand over his wallet. When the victim resisted, Ewing produced a knife and forced the victim back to the apartment itself. While Ewing rifled through the bedroom, the victim fled the apartment screaming for help. Ewing absconded with the victim’s money and credit cards.

On December 9,1993, Ewing was arrested on the premises of the apartment complex for trespassing and lying to a police officer. The knife used in the robbery and a glass cocaine pipe were later found in the back seat of the patrol car used to transport Ewing to the police station. A jury convicted Ewing of first-degree robbery and three counts of residential burglary. Sentenced to nine years and eight months in prison, Ewing was paroled in 1999.

Only 10 months later, Ewing stole the golf clubs at issue in this case. He was charged with, and ultimately convicted of, one count of felony grand theft of personal property in excess of $400. See Cal. Penal Code Ann. § 484 (West Supp. 2002); §489 (West 1999). As required by the three strikes law, the prosecutor formally alleged, and the trial court later found, that Ewing had been convicted previously of four serious or violent felonies for the three burglaries and the robbery in the Long Beach apartment complex. See § 667(g) (West 1999); § 1170.12(e) (West Supp. 2002).

At the sentencing hearing, Ewing asked the court to reduce the conviction for grand theft, a “wobbler” under California law, to a misdemeanor so as to avoid a three strikes sentence. See §§ 17(b), 667(d)(1) (West 1999); § 1170.12(b)(1) (West Supp. 2002). Ewing also asked the trial court to exercise its discretion to dismiss the allegations of some or all of his prior serious or violent felony convictions, again for purposes of avoiding a three strikes sentence. See Romero, 13 Cal. 4th, at 529-531, 917 P. 2d, at 647-648. Before sen-*20fencing Ewing, the trial court took note of his entire criminal history, including the fact that he was on parole when he committed his latest offense. The court also heard arguments from defense counsel and a plea from Ewing himself.

In the end, the trial judge determined that the grand theft should remain a felony. The court also ruled that the four prior strikes for the three burglaries and the robbery in Long Beach should stand. As a newly convicted felon with two or more “serious” or “violent” felony convictions in his past, Ewing was sentenced under the three strikes law to 25 years to life.

The California Court of Appeal affirmed in an unpublished opinion. No. B143745 (Apr. 25, 2001). Relying on our decision in Rummel v. Estelle, 445 U. S. 263 (1980), the court rejected Ewing’s claim that his sentence was grossly disproportionate under the Eighth Amendment. Enhanced sentences under recidivist statutes like the three strikes law, the court reasoned, serve the “legitimate goal” of deterring and incapacitating repeat offenders. The Supreme Court of California denied Ewing’s petition for review, and we granted certiorari, 535 U. S. 969 (2002). We now affirm.

II

A

The Eighth Amendment, which forbids cruel and unusual punishments, contains a “narrow proportionality principle” that “applies to noncapital sentences.” Harmelin v. Michigan, 501 U. S. 957, 996-997 (1991) (Kennedy, J., concurring in part and concurring in judgment); cf. Weems v. United States, 217 U. S. 349, 371 (1910); Robinson v. California, 370 U. S. 660, 667 (1962) (applying the Eighth Amendment to the States via the Fourteenth Amendment). We have most recently addressed the proportionality principle as applied to terms of years in a series of cases beginning with Rummel v. Estelle, supra.

*21In Rummel, we held that it did not violate the Eighth Amendment for a State to sentence a three-time offender to life in prison with the possibility of parole. Id., at 284-285. Like Ewing, Rummel was sentenced to a lengthy prison term under a recidivism statute. Rummel’s two prior offenses were a 1964 felony for “fraudulent use of a credit card to obtain $80 worth of goods or services,” and a 1969 felony conviction for “passing a forged check in the amount of $28.36.” Id., at 265. His triggering offense was a conviction for felony theft — “obtaining $120.75 by false pretenses.” Id., at 266.

This Court ruled that “[hjaving twice imprisoned him for felonies, Texas was entitled to place upon Rummel the onus of one who is simply unable to bring his conduct within the social norms prescribed by the criminal law of the State.” Id., at 284. The recidivism statute “is nothing more than a societal decision that when such a person commits yet another felony, he should be subjected to the admittedly serious penalty of incarceration for life, subject only to the State’s judgment as to whether to grant him parole.” Id., at 278. We noted that this Court “has on occasion stated that the Eighth Amendment prohibits imposition of a sentence that is grossly disproportionate to the severity of the crime.” Id., at 271. But “[ojutside the context of capital punishment, successful challenges to the proportionality of particular sentences have been exceedingly rare.” Id., at 272. Although we stated that the proportionality principle “would . . . come into play in the extreme example ... if a legislature made overtime parking a felony punishable by life imprisonment,” id., at 274, n. 11, we held that “the mandatory life sentence imposed upon this petitioner does not constitute cruel and unusual punishment under the Eighth and Fourteenth Amendments,” id., at 285.

In Hutto v. Davis, 454 U. S. 370 (1982) (per curiam), the defendant was sentenced to two consecutive terms of 20 years in prison for possession with intent to distribute nine *22ounces of marijuana and distribution of marijuana. We held that such a sentence was constitutional: “In short, Rummel stands for the proposition that federal courts should be reluctant to review legislatively mandated terms of imprisonment, and that successful challenges to the proportionality of particular sentences should be exceedingly rare.” Id., at 374 (citations and internal quotation marks omitted).

Three years after Rummel, in Solem v. Helm, 463 U. S. 277, 279 (1983), we held that the Eighth Amendment prohibited “a life sentence without possibility of parole for a seventh nonviolent felony.” The triggering offense in Solem was “uttering a ‘no account’ check for $100.” Id., at 281. We specifically stated that the Eighth Amendment’s ban on cruel and unusual punishments “prohibits .. . sentences that are disproportionate to the crime committed,” and that the “constitutional principle of proportionality has been recognized explicitly in this Court for almost a century.” Id., at 284, 286. The Solem Court then explained that three factors may be relevant to a determination of whether a sentence is so disproportionate that it violates the Eighth Amendment: “(i) the gravity of the offense and the harshness of the penalty; (ii) the sentences imposed on other criminals in the same jurisdiction; and (iii) the sentences imposed for commission of the same crime in other jurisdictions.” Id., at 292.

Applying these factors in Solem, we struck down the defendant’s sentence of life without parole. We specifically noted the contrast between that sentence and the sentence in Rummel, pursuant to which the defendant was eligible for parole. 463 U. S., at 297; see also id., at 300 (“[T]he South Dakota commutation system is fundamentally different from the parole system that was before us in Rummel”). Indeed, we explicitly declined to overrule Rummel: “[Q]ur conclusion today is not inconsistent with Rummel v. Estelle.” 463 U. S., at 303, n. 32; see also id., at 288, n. 13 (“[O]ur decision *23is entirely consistent with this Court’s prior cases — including Rummel v. Estelle”).

Eight years after Solem, we grappled with the proportionality issue again in Harmelin. Harmelin was not a recidivism case, but rather involved a first-time offender convicted of possessing 672 grams of cocaine. He was sentenced to life in prison without possibility of parole. A majority of the Court rejected Harmelin’s claim that his sentence was so grossly disproportionate that it violated the Eighth Amendment. The Court, however, could not agree on why his proportionality argument failed. Justice Scalia, joined by The Chief Justice, wrote that the proportionality principle was “an aspect of our death penalty jurisprudence, rather than a generalizable aspect of Eighth Amendment law.” 501 U. S. at 994. He would thus have declined to apply gross disproportionality principles except in reviewing capital sentences. Ibid.

Justice Kennedy,

joined by two other Members of the Court, concurred in part and concurred in the judgment. Justice Kennedy specifically recognized that “[t]he Eighth Amendment proportionality principle also applies to noncap-ital sentences.” Id., at 997. He then identified four principles of proportionality review — “the primacy of the legislature, the variety of legitimate penological schemes, the nature of our federal system, and the requirement that proportionality review be guided by objective factors” — that “inform the final one: The Eighth Amendment does not require strict proportionality between crime and sentence. Rather, it forbids only extreme sentences that are ‘grossly disproportionate’ to the crime.” Id., at 1001 (citing Solem, supra, at 288). Justice Kennedy’s concurrence also stated that Solem “did not mandate” comparative analysis “within and between jurisdictions.” 501 U. S., at 1004-1005.

The proportionality principles in our cases distilled in Justice Kennedy’s concurrence guide our application of the *24Eighth Amendment in the new context that we are called upon to consider.

B

For many years, most States have had laws providing for enhanced sentencing of repeat offenders. See, e. g., U. S. Dept. of Justice, Bureau of Justice Assistance, National Assessment of Structured Sentencing (1996). Yet between 1993 and 1995, three strikes laws effected a sea change in criminal sentencing throughout the Nation.1 These laws responded to widespread public concerns about crime by targeting the class of offenders who pose the greatest threat to public safety: career criminals. As one of the chief architects of California’s three strikes law has explained: “Three Strikes was intended to go beyond simply making sentences tougher. It was intended to be a focused effort to create a sentencing policy that would use the judicial system to reduce serious and violent crime.” Ardaiz, California’s Three Strikes Law: History, Expectations, Consequences, 32 McGeorge L. Rev. 1, 12 (2000) (hereinafter Ardaiz).

Throughout the States, legislatures enacting three strikes laws made a deliberate policy choice that individuals who have repeatedly engaged in serious or violent criminal behavior, and whose conduct has not been deterred by more conventional approaches to punishment, must be isolated from society in order to protect the public safety. Though three strikes laws may be relatively new, our tradition of deferring to state legislatures in making and implementing such important policy decisions is longstanding. Weems, 217 U. S., at 379; Gore v. United States, 357 U. S. 386, 393 *25(1958); Payne v. Tennessee, 501 U. S. 808, 824 (1991); Rummel, 445 U. S., at 274; Solem, 463 U. S., at 290; Harmelin, 501 U. S., at 998 (Kennedy, J., concurring in part and concurring in judgment).

Our traditional deference to legislative policy choices finds a corollary in the principle that the Constitution “does not mandate adoption of any one penological theory.” Id., at 999 (Kennedy, J., concurring in part and concurring in judgment). A sentence can have a variety of justifications, such as incapacitation, deterrence, retribution, or rehabilitation. See 1 W. LaFave & A. Scott, Substantive Criminal Law § 1.5, pp. 30-36 (1986) (explaining theories of punishment). Some or all of these justifications may play a role in a State’s sentencing scheme. Selecting the sentencing rationales is generally a policy choice to be made by state legislatures, not federal courts.

When the California Legislature enacted the three strikes law, it made a judgment that protecting the public safety requires incapacitating criminals who have already been convicted of at least one serious or violent crime. Nothing in the Eighth Amendment prohibits California from making that choice. To the contrary, our cases establish that “States have a valid interest in deterring and segregating habitual criminals.” Parke v. Raley, 506 U. S. 20, 27 (1992); Oyler v. Boles, 368 U. S. 448, 451 (1962) (“[T]he constitutionality of the practice of inflicting severer criminal penalties . upon habitual offenders is no longer open to serious challenge”). Recidivism has long been recognized as a legitimate basis for increased punishment. See Almendarez-Torres v. United States, 523 U. S. 224, 230 (1998) (recidivism “is as typical a sentencing factor as one might imagine”); Witte v. United States, 515 U. S. 389, 400 (1995) (“In repeatedly upholding such recidivism statutes, we have rejected double jeopardy challenges because the enhanced punishment imposed for the later offense ... [is] ‘a stiffened penalty for the latest crime, which is considered to be an aggravated *26offense because a repetitive one’ ” (quoting Gryger v. Burke, 334 U. S. 728, 732 (1948))).

California’s justification is no pretext. Recidivism is a serious public safety concern in California and throughout the Nation. According to a recent report, approximately 67 percent of former inmates released from state prisons were charged with at least one “serious” new crime within three years of their release. See U. S. Dept. of Justice, Bureau of Justice Statistics, P. Langan & D. Levin, Special Report: Recidivism of Prisoners Released in 1994, p. 1 (June 2002). In particular, released property offenders like Ewing had higher recidivism rates than those released after committing violent, drug, or public-order offenses. Id., at 8. Approximately 73 percent of the property offenders released in 1994 were arrested again within three years, compared to approximately 61 percent of the violent offenders, 62 percent of the public-order offenders, and 66 percent of the drug offenders. Ibid.

In 1996, when the Sacramento Bee studied 233 three strikes offenders in California, it found that they had an aggregate of 1,165 prior felony convictions, an average of 5 apiece. See Furillo, Three Strikes — The Verdict: Most Offenders Have Long Criminal Histories, Sacramento Bee, Mar. 31, 1996, p. A1. The prior convictions included 322 robberies and 262 burglaries. Ibid. About 84 percent of the 233 three strikes offenders had been convicted of at least one violent crime. Ibid. In all, they were responsible for 17 homicides, 7 attempted slayings, and 91 sexual assaults and child molestations. Ibid. The Sacramento Bee concluded, based on its investigation, that “[i]n the vast majority of the cases, regardless of the third strike, the [three strikes] law is snaring [the] long-term habitual offenders with multiple felony convictions . . ..” Ibid.

The State’s interest in deterring crime also lends some support to the three strikes law. We have long viewed both incapacitation and deterrence as rationales for recidivism *27statutes: “[A] recidivist statute[’s] . . . primary goals are to deter repeat offenders and, at some point in the life of one who repeatedly commits criminal offenses serious enough to be punished as felonies, to segregate that person from the rest of society for an extended period of time.” Rummel, supra, at 284. Four years after the passage of California’s three strikes law, the recidivism rate of parolees returned to prison for the commission of a new crime dropped by nearly 25 percent. California Dept. of Justice, Office of the Attorney General, “Three Strikes and You’re Out” — Its Impact on the California Criminal Justice System After Four Years, p. 10 (1998). Even more dramatically:

“An unintended but positive consequence of ‘Three Strikes’ has been the impact on parolees leaving the state. More California parolees are now leaving the state than parolees from other jurisdictions entering California. This striking turnaround started in 1994. It was the first time more parolees left the state than entered since 1976. This trend has continued and in 1997 more than 1,000 net parolees left California.” Ibid.

See also Janiskee & Erler, Crime, Punishment, and Romero: An Analysis of the Case Against California’s Three Strikes Law, 39 Duquesne L. Rev. 43, 45-46 (2000) (“Prosecutors in Los Angeles routinely report that ‘felons tell them they are moving out of the state because they fear getting a second or third strike for a nonviolent offense’ ” (quoting Sanchez, A Movement Builds Against “Three Strikes” Law, Washington Post, Feb. 18, 2000, p. A3)).

To be sure, California’s three strikes law has sparked controversy. Critics have doubted the law’s wisdom, cost-efficiency, and effectiveness in reaching its goals. See, e. g., Zimring, Hawkins, & Kamin, Punishment and Democracy: Three Strikes and You’re Out in California (2001); Vitiello, Three Strikes: Can We Return to Rationality? 87 J. Crim. *28L. & C. 395, 423 (1997). This criticism is appropriately directed at the legislature, which has primary responsibility for making the difficult policy choices that underlie any criminal sentencing scheme. We do not sit as a “superlegisla-ture” to second-guess these policy choices. It is enough that the State of California has a reasonable basis for believing that dramatically enhanced sentences for habitual felons “advance[s] the goals of [its] criminal justice system in any substantial way.” See Solem, 463 U. S., at 297, n. 22.

III

Against this backdrop, we consider Ewing’s claim that his three strikes sentence of 25 years to life is unconstitutionally disproportionate to his offense of “shoplifting three golf clubs.” Brief for Petitioner 6. We first address the gravity of the offense compared to the harshness of the penalty. At the threshold, we note that Ewing incorrectly frames the issue. The gravity of his offense was not merely “shoplifting three golf clubs.” Rather, Ewing was convicted of felony grand theft for- stealing nearly $1,200 worth of merchandise after previously having been convicted of at least two “violent” or “serious” felonies. Even standing alone, Ewing’s theft should not be taken lightly. His crime was certainly not “one of the most passive felonies a person could commit.” Solem, supra, at 296 (internal quotation marks omitted). To the contrary, the Supreme Court of California has noted the “seriousness” of grand theft in the context of proportionality review. See In re Lynch, 8 Cal. 3d 410, 432, n. 20, 503 P. 2d 921, 936, n. 20 (1972). Theft of $1,200 in property is a felony under federal law, 18 U. S. C. § 641, and in the vast majority of States. See App. B to Brief for Petitioner 21a.

That grand theft is a “wobbler” under California law is of no moment. Though California courts have discretion to reduce a felony grand theft charge to a misdemeanor, it remains a felony for all purposes “unless and until the trial *29court imposes a misdemeanor sentence.” In re Anderson, 69 Cal. 2d 618, 626, 447 P. 2d 117, 126 (1968) (Tobriner, J., concurring); see generally 1 B. Witkin & N. Epstein, California Criminal Law §73 (3d ed. 2000). “The purpose of the trial judge’s sentencing discretion” to downgrade certain felonies is to “impose a misdemeanor sentence in those cases in which the rehabilitation of the convicted defendant either does not require, or would be adversely affected by, incarceration in a state prison as a felon.” Anderson, supra, at 664-665, 447 P. 2d, at 152 (Tobriner, J., concurring). Under California law, the reduction is not based on the notion that a “wobbler” is “conceptually a misdemeanor.” Necochea v. Superior Court, 23 Cal. App. 3d 1012, 1016, 100 Cal. Rptr. 693, 695 (1972). Rather, it is “intended to extend misde-meanant treatment to a potential felon.” Ibid. In Ewing’s case, however, the trial judge justifiably exercised her discretion not to extend such lenient treatment given Ewing’s long criminal history.

In weighing the gravity of Ewing’s offense, we must place on the scales not only his current felony, but also his long history of felony recidivism. Any other approach would fail to accord proper deference to the policy judgments that find expression in the legislature’s choice of sanctions. In imposing a three strikes sentence, the State’s interest is not merely punishing the offense of conviction, or the “triggering” offense: “[I]t is in addition the interest... in dealing in a harsher manner with those who by repeated criminal acts have shown that they are simply incapable of conforming to the norms of society as established by its criminal law.” Rummel, 445 U. S., at 276; Solem, supra, at 296. To give full effect to the State’s choice of this legitimate penological goal, our proportionality review of Ewing’s sentence must take that goal into account.

Ewing’s sentence is justified by the State’s public-safety interest in incapacitating and deterring recidivist felons, and *30amply supported by his own long, serious criminal record.2 Ewing has been convicted of numerous misdemeanor and felony offenses, served nine separate terms of incarceration, and committed most of his crimes while on probation or parole. His prior “strikes” were serious felonies including robbery and three residential burglaries. To be sure, Ewing’s sentence is a long one. But it reflects a rational legislative judgment, entitled to deference, that offenders who have committed serious or violent felonies and who continue to commit felonies must be incapacitated. The State of California “was entitled to place upon [Ewing] the onus of one who is simply unable to bring his conduct within the social norms prescribed by the criminal law of the State.” Rummel, supra, at 284. Ewing’s is not “the rare case in which a threshold comparison of the crime committed and the sentence imposed leads to an inference of gross disproportion-ality.” Harmelin, 501 U. S., at 1005 (Kennedy, J., concurring in part and concurring in judgment).

We hold that Ewing’s sentence of 25 years to life in prison, imposed for the offense of felony grand theft under the three strikes law, is not grossly disproportionate and therefore does not violate the Eighth Amendment’s prohibition on *31cruel and unusual punishments. The judgment of the California Court of Appeal is affirmed.

It is so ordered.

Justice Scalia,

concurring in the judgment.

In my opinion in Harmelin v. Michigan, 501 U. S. 957, 985 (1991), I concluded that the Eighth Amendment’s prohibition of “cruel and unusual punishments” was aimed at excluding only certain modes of punishment, and was hot a “guarantee against disproportionate sentences.” Out of respect for the principle of stare decisis, I might nonetheless accept the contrary holding of Solem v. Helm, 463 U. S. 277 (1983) — that the Eighth Amendment contains a narrow proportionality principle — if I felt I could intelligently apply it. This case demonstrates why I cannot.

Proportionality — the notion that the punishment should fit the crime — is inherently a concept tied to the penological goal of retribution. “[I]t becomes difficult even to speak intelligently of ‘proportionality,’ once deterrence and rehabilitation are given significant weight,” Harmelin, supra, at 989 — not to mention giving weight to the purpose of California’s three strikes law: incapacitation. In the present case, the game is up once the plurality has acknowledged that “the Constitution does not mandate adoption of any one penological theory,” and that a “sentence can have a variety of justifications, such as incapacitation, deterrence, retribution, or rehabilitation.” Ante, at 25 (internal quotation marks omitted). That acknowledgment having been made, it no longer suffices merely to assess “the gravity of the offense compared to the harshness of the penalty,” ante, at 28; that classic description of the proportionality principle (alone and in itself quite resistant to policy-free, legal analysis) now becomes merely the “first” step of the inquiry, ibid. Having completed that step (by a discussion which, in all fairness, does not convincingly establish that 25-years-to-life is a “proportionate” punishment for stealing three golf clubs), the *32plurality must then add an analysis to show that “Ewing’s sentence is justified by the State’s public-safety interest in incapacitating and deterring recidivist felons.” Ante, at 29.

Which indeed it is — though why that has anything to do with the principle of proportionality is a mystery. Perhaps the plurality should revise its terminology, so that what it reads into the Eighth Amendment is not the unstated proposition that all punishment should be reasonably proportionate to the gravity of the offense, but rather the unstated proposition that all punishment should reasonably pursue the multiple purposes of the criminal law. That formulation would make it clearer than ever, of course, that the plurality is not applying law but evaluating policy.

Because I agree that petitioner’s sentence does not violate the Eighth Amendment’s prohibition against cruel and unusual punishments, I concur in the judgment.

Justice Thomas,

concurring in the judgment.

I agree with Justice Scalia’s view that the proportionality test announced in Solem v. Helm, 463 U. S. 277 (1983), is incapable of judicial application. Even were Solem’,s test perfectly clear, however, I would not feel compelled by stare decisis to apply it. In my view, the Cruel and Unusual Punishments Clause of the Eighth Amendment contains no proportionality principle. See Harmelin v. Michigan, 501 U. S. 957, 966-985 (1991) (opinion of Scalia, J.).

Because the plurality concludes that petitioner’s sentence does not violate the Eighth Amendment’s prohibition on cruel and unusual punishments, I concur in the judgment.

Justice Stevens,

with whom Justice Souter, Justice Ginsburg, and Justice Breyer join, dissenting.

Justice Breyer has cogently explained why the sentence imposed in this case is both cruel and unusual.1 The eoncur-*33rences prompt this separate writing to emphasize that proportionality review is not only capable of judicial application but also required by the Eighth Amendment.

“The Eighth Amendment succinctly prohibits ‘excessive’ sanctions.” Atkins v. Virginia, 536 U. S. 304, 311 (2002); see also U. S. Const., Arndt. 8 (“Excessive bail shall not be required, nor excessive fines imposed, nor cruel and unusual punishments inflicted”). Faithful to the Amendment’s text, this Court has held that the Constitution directs judges to apply their best judgment in determining the proportionality of fines, see, e. g., United States v. Bajakajian, 524 U. S. 321, 334-336 (1998), bail, see, e. g., Stack v. Boyle, 342 U. S. 1, 5 (1951), and other forms of punishment, including the imposition of a death sentence, see, e. g., Coker v. Georgia, 433 U. S. 584, 592 (1977). It “would be anomalous indeed” to suggest that the Eighth Amendment makes proportionality review applicable in the context of bail and fines but not in the context of other forms of punishment, such as imprisonment. Solem v. Helm, 463 U. S. 277, 289 (1983). Rather, by broadly prohibiting excessive sanctions, the Eighth Amendment directs judges to exercise their wise judgment in assessing the proportionality of all forms of punishment.

The absence of a black-letter rule does not disable judges from exercising their discretion in construing the outer limits on sentencing authority that the Eighth Amendment imposes. After all, judges are “constantly called upon to draw . . . lines in a variety of contexts,” id., at 294, and to exercise their judgment to give meaning to the Constitution’s broadly phrased protections. For example, the Due Process Clause directs judges to employ proportionality re*34view in assessing the constitutionality of punitive damages awards on a case-by-case basis. See, e. g., BMW of North America, Inc. v. Gore, 517 U. S. 559, 562 (1996). Also, although the Sixth Amendment guarantees criminal defendants the right to a speedy trial, the courts often are asked to determine on a case-by-case basis whether a particular delay is constitutionally permissible or not. See, e. g., Doggett v. United States, 505 U. S. 647 (1992).2

Throughout most of the Nation’s history — before guideline sentencing became so prevalent — federal and state trial judges imposed specific sentences pursuant to grants of authority that gave them uncabined discretion within broad ranges. See K. Stith & J. Cabranes, Fear of Judging: Sentencing Guidelines in the Federal Courts 9 (1998) (hereinafter Stith & Cabranes) (“From the beginning of the Republic, federal judges were entrusted with wide sentencing discretion”); see also Mistretta v. United States, 488 U. S. 361, 364 (1989). It was not unheard of for a statute to authorize a sentence ranging from one year to life, for example. See, e. g., State v. Perley, 86 Me. 427, 30 A. 74, 75 (1894) (citing Maine statute that made robbery punishable by imprisonment for life or any term of years); In re Southard, 298 Mich. 75, 77, 298 N. W. 457 (1941) (“The offense of ‘robbery armed’ is punishable by imprisonment for life or any term *35of years”). In exercising their discretion, sentencing judges wisely employed a proportionality principle that took into account all of the justifications for punishment — namely, deterrence, incapacitation, retribution, and rehabilitation. See Stith & Cabranes 14. Likewise, I think it clear that the Eighth Amendment’s prohibition of “cruel and unusual punishments” expresses a broad and basic proportionality principle that takes into account all of the justifications for penal sanctions. It is this broad proportionality principle that would preclude reliance on any of the justifications for punishment to support, for example, a life sentence for overtime parking. See Rummel v. Estelle, 445 U. S. 263, 274, n. 11 (1980).

Accordingly, I respectfully dissent.

Justice Breyer,

with whom Justice Stevens, Justice Souter, and Justice Ginsburg join, dissenting.

The constitutional question is whether the “three strikes” sentence imposed by California upon repeat-offender Gary Ewing is “grossly disproportionate” to his crime. Ante, at 14, 30-31 (plurality opinion). The sentence amounts to a real prison term of at least 25 years. The sentence-triggering criminal conduct consists of the theft of three golf clubs priced at a total of $1,197. See ante, at 18. The offender has a criminal history that includes four felony convictions arising out of three separate burglaries (one armed). Ante, at 18-19. In Solem v. Helm, 463 U. S. 277 (1983), the Court found grossly disproportionate a somewhat longer sentence imposed on a recidivist offender for triggering criminal conduct that was somewhat less severe. In my view, the differences are not determinative, and the Court should reach the same ultimate conclusion here.

I

This Court’s precedent sets forth a framework for analyzing Ewing’s Eighth Amendment claim. The Eighth Amendment forbids, as “cruel and unusual punishments,” prison *36terms (including terms of years) that are “grossly disproportionate.” Solem, supra, at 303; see Lockyer v. Andrade, post, at 71. In applying the “gross disproportionality” principle, courts must keep in mind that “legislative policy” will primarily determine the appropriateness of a punishment’s “severity,” and hence defer to such legislative policy judgments. Gore v. United States, 357 U. S. 386, 393 (1958); see Harmelin v. Michigan, 501 U. S. 957, 998 (1991) (Kennedy, J., concurring in part and concurring in judgment); Solem, supra, at 289-290; Rummel v. Estelle, 445 U. S. 263, 274-276 (1980); Weems v. United States, 217 U. S. 349, 373 (1910). If courts properly respect those judgments, they will find that the sentence fails the test only in rare instances. Solem, supra, at 290, n. 16; Harmelin, supra, at 1004 (Kennedy, J., concurring in part and concurring in judgment); Rummel, supra, at 272 (“[Successful challenges to the proportionality of particular sentences have been exceedingly rare”). And they will only “‘rarely’” find it necessary to “‘engage in extended analysis’” before rejecting a claim that a sentence is “grossly disproportionate.” Harmelin, supra, at 1004 (Kennedy, J., concurring in part and concurring in judgment) (quoting Solem, supra, at 290, n. 16).

The plurality applies Justice Kennedy’s analytical framework in Harmelin, supra, at 1004-1005 (opinion concurring in part and concurring in judgment). Ante, at 23-24. And, for present purposes, I will consider Ewing’s Eighth Amendment claim on those terms. But see ante, at 32-33, n. 1 (Stevens, J., dissenting). To implement this approach, courts faced with a “gross disproportionality” claim must first make “a threshold comparison of the crime committed and the sentence imposed.” Harmelin, supra, at 1005 (Kennedy, J., concurring in part and concurring in judgment). If a claim crosses that threshold — itself a rare occurrence — then the court should compare the sentence at issue to other sentences “imposed on other criminals” in the same, or in other, jurisdictions. Solem, supra, at 290-291; *37Harmelin, 501 U. S., at 1005 (Kennedy, J., concurring in part and concurring in judgment). The comparative analysis will “validate” or invalidate “an initial judgment that a sentence is grossly disproportionate to a crime.” Ibid.

I recognize the warnings implicit in the Court’s frequent repetition of words such as “rare.” Nonetheless I believe that the case before us is a “rare” case — one in which a court can say with reasonable confidence that the punishment is “grossly disproportionate” to the crime.

II

Ewing’s claim crosses the gross disproportionality “threshold.” First, precedent makes clear that Ewing’s sentence raises a serious disproportionality question. Ewing is a recidivist. Hence the two cases most directly in point are those in which the Court considered the constitutionality of recidivist sentencing: Rummel and Solem. Ewing’s claim falls between these two cases. It is stronger than the claim presented in Rummel, where the Court upheld a recidivist’s sentence as constitutional. It is weaker than the claim presented in Solem, where the Court struck down a recidivist sentence as unconstitutional.

Three kinds of sentence-related characteristics define the relevant comparative spectrum: (a) the length of the prison term in real time, i. e., the time that the offender is likely actually to spend in prison; (b) the sentence-triggering criminal conduct, i. e., the offender’s actual behavior or other offense-related circumstances; and (c) the offender’s criminal history. See Rummel, supra, at 265-266, 269, 276, 278, 280-281 (using these factors); Solem, supra, at 290-303 (same). Cf. United States Sentencing Commission, Guidelines Manual ch. 1, pt. A, intro., n. 5 (Nov. 1987) (USSG) (empirical study of “summary reports of some 40,000 convictions [and] a sample of 10,000 augmented presentence reports” leads to sentences based primarily upon (a) offense characteristics and (b) offender’s criminal record); see id., p. s. 3.

*38In Rummel, the Court held constitutional (a) a sentence of life imprisonment with parole available within 10 to 12 years, (b) for the offense of obtaining $120 by false pretenses, (c) committed by an offender with two prior felony convictions (involving small amounts of money). 445 U. S. 263; ante, at 21. In Solem, the Court held unconstitutional (a) a sentence of life imprisonment without parole, (b) for the crime of writing a $100 check on a nonexistent bank account, (c) committed by an offender with six prior felony convictions (including three for burglary). 463 U. S. 277; ante, at 22-23. Which of the three pertinent comparative factors made the constitutional difference?

The third factor, prior record, cannot explain the difference. The offender’s prior record was worse in Solem, where the Court found the sentence too long, than in Rum-mel, where the Court upheld the sentence. The second factor, offense conduct, cannot explain the difference. The nature of the triggering offense — viewed in terms of the actual monetary loss — in the two cases was about the same. The one critical factor that explains the difference in the outcome is the length of the likely prison term measured in real time. In Rummel, where the Court upheld the sentence, the state sentencing statute authorized parole for the offender, Rum-mel, after 10 or 12 years. 445 U. S., at 280; id., at 293 (Powell, J., dissenting). In Solem, where the Court struck down the sentence, the sentence required the offender, Helm, to spend the rest of his life in prison.

Now consider the present case. The third factor, offender characteristics — i. e., prior record — does not differ significantly here from that in Solem. Ewing’s prior record consists of four prior felony convictions (involving three burglaries, one with a knife) contrasted with Helm’s six prior felony convictions (including three burglaries, though none with weapons). The second factor, offense behavior, is worse than that in Solem, but only to a degree. It would be difficult to say that the actual behavior itself here (shop*39lifting) differs significantly from that at issue in Solem (passing a bad check) or in Rummel (obtaining money through false pretenses). Rather the difference lies in the value of the goods obtained. That difference, measured in terms of the most relevant feature (loss to the victim, i. e., wholesale value) and adjusted for the irrelevant feature of inflation, comes down (in 1979 values) to about $379 here compared with $100 in Solem, or (in 1973 values) to $232 here compared with $120.75 in Rummel. See USSG §2B1.1, comment., n. 2(A)(i) (Nov. 2002) (loss to victim properly measures value of goods unlawfully taken); U. S. Dept. of Labor, Bureau of Labor Statistics, Inflation and Consumer Spending, Inflation Calculator (Jan. 23, 2003), http://www.bls.gov (hereinafter Inflation Calculator). Alternatively, if one measures the inflation-adjusted value difference in terms of the golf clubs’ sticker price, it comes down to $505 here compared to $100 in Solem, or $309 here compared to $120.75 in Rummel. See Inflation Calculator.

The difference in length of the real prison term — the first, and critical, factor in Solem and Rummel — is considerably more important. Ewing’s sentence here amounts, in real terms, to at least 25 years without parole or good-time credits. That sentence is considerably shorter than Helm’s sentence in Solem, which amounted, in real terms, to life in prison. Nonetheless Ewing’s real prison term is more than twice as long as the term at issue in Rummel, which amounted, in real terms, to at least 10 or 12 years. And, Ewing’s sentence, unlike Rummel’s (but like Helm’s sentence in Solem), is long enough to consume the productive remainder of almost any offender’s life. (It means that Ewing himself, seriously ill when sentenced at age 38, will likely die in prison.)

The upshot is that the length of the real prison term — the factor that explains the Solem/Rummel difference in outcome — places Ewing closer to Solem than to Rummel, though the greater value of the golf clubs that Ewing stole *40moves Ewing’s case back slightly in Rummel’s direction. Overall, the comparison places Ewing’s sentence well within the twilight zone between Solem and Rummel — a zone where the argument for unconstitutionality is substantial, where the cases themselves cannot determine the constitutional outcome.

Second, Ewing’s sentence on its face imposes one of the most severe punishments available upon a recidivist who subsequently engaged in one of the less serious forms of criminal conduct. See infra, at 44-45. I do not deny the seriousness of shoplifting, which an amicus curiae tells us costs retailers in the range of $30 billion annually. Brief for California District Attorneys Association as Amicus Curiae 27. But consider that conduct in terms of the factors that this Court mentioned in Solem — the “harm caused or threatened to the victim or society,” the “absolute magnitude of the crime,” and the offender’s “culpability.” 463 U. S., at 292-293. In respect to all three criteria, the sentence-triggering behavior here ranks well toward the bottom of the criminal conduct scale.

The Solicitor General has urged us to consider three other criteria: the “frequency” of the crime’s commission, the “ease or difficulty of detection,” and “the degree to which the crime may be deterred by differing amounts of punishment.” Brief for United States as Amicus Curiae 24-25. When considered in terms of these criteria — or at least the latter two — the triggering conduct also ranks toward the bottom of the scale. Unlike, say, drug crimes, shoplifting often takes place in stores open to other customers whose presence, along with that of store employees or cameras, can help to detect the crime. Nor is there evidence presented here that the law enforcement community believes lengthy prison terms necessary adequately to deter shoplifting. To the contrary, well-publicized instances of shoplifting suggest that the offense is often punished without any prison sentence at all. On the other hand, shoplifting is a frequently com*41mitted crime; but “frequency,” standing alone, cannot make a critical difference. Otherwise traffic offenses would warrant even more serious punishment.

This ease, of course, involves shoplifting engaged in by a recidivist. One might argue that any crime committed by a recidivist is a serious crime potentially warranting a 25-year sentence. But this Court rejected that view in Solem, and in Harmelin, with the recognition that “no penalty is per se constitutional.” Solem, supra, at 290; Harmelin, 501 U. S., at 1001 (Kennedy, J., concurring in part and concurring in judgment). Our cases make clear that, in cases involving recidivist offenders, we must focus upon “the [offense] that triggers the life sentence,” with recidivism playing a “relevant,” but not necessarily determinative, role. Solem, supra, at 296, n. 21; see Witte v. United States, 515 U. S. 389, 402, 403 (1995) (the recidivist defendant is “punished only for the offense of conviction,” which “‘is considered to be an aggravated offense because a repetitive one’” (quoting Gryger v. Burke, 334 U. S. 728, 732 (1948))). And here, as I have said, that offense is among the less serious, while the punishment is among the most serious. Cf. Rummel, 445 U. S., at 288 (Powell, J., dissenting) (overtime parking violation cannot trigger a life sentence even for a serious recidivist).

Third, some objective evidence suggests that many experienced judges would consider Ewing’s sentence disproportionately harsh. The United States Sentencing Commission (having based the federal Sentencing Guidelines primarily upon its review of how judges had actually sentenced offenders) does not include shoplifting (or similar theft-related offenses) among the crimes that might trigger especially long sentences for recidivists, see USSG §4B1.1 (Nov. 2002) (Guideline for sentencing “career offenders”); id., ch. 1, pt. A, intro., n. 5 (sentences based in part upon Commission’s review of “summary reports of some 40,000 convictions [and] a sample of 10,000 augmented presentence reports”); see also *42infra, at 45, nor did Congress include such offenses among triggering crimes when it sought sentences “at or near the statutory maximum” for certain recidivists, S. Rep. No. 98-225, p. 175 (1983); 28 U. S. C. § 994(h) (requiring sentence “at or near the maximum” where triggering crime is crime of “violence” or drug related); 18 U. S. C. § 3559(c) (grand theft not among triggering or “strike” offenses under federal “three strikes” law); see infra, at 45-46. But see 28 U. S. C. § 994(i)(l) (requiring “a substantial term of imprisonment” for those who have “a history of two or more prior ... felony convictions”).

Taken together, these three circumstances make clear that Ewing’s “gross disproportionality” argument is a strong one. That being so, his claim must pass the “threshold” test. If it did not, what would be the function of the test? A threshold test must permit arguably unconstitutional sentences, not only actually unconstitutional sentences, to pass the threshold — at least where the arguments for unconstitutionality are unusually strong ones. A threshold test that blocked every ultimately invalid constitutional claim — even strong ones — would not be a threshold test but a determinative test. And, it would be a determinative test that failed to take account of highly pertinent sentencing information, namely, comparison with other sentences, Solem, supra, at 291-292, 298-300. Sentencing comparisons are particularly important because they provide proportionality review with objective content. By way of contrast, a threshold test makes the assessment of constitutionality highly subjective. And, of course, so to transform that threshold test would violate this Court’s earlier precedent. See 463 U. S., at 290, 291-292; Harmelin, supra, at 1000, 1005 (Kennedy, J., concurring in part and concurring in judgment).

> — l ► — <

Believing Ewing’s argument a strong one, sufficient to pass the threshold, I turn to the comparative analysis. A *43comparison of Ewing’s sentence with other sentences requires answers to two questions. First, how would other jurisdictions (or California at other times, i. e,, without the three strikes penalty) punish the same offense conduct? Second, upon what other conduct would other jurisdictions (or California) impose the same prison term? Moreover, since hypothetical punishment is beside the point, the relevant prison time, for comparative purposes, is real prison time, i. e., the time that an offender must actually serve.

Sentencing statutes often shed little light upon real prison time. That is because sentencing laws normally set maximum sentences, giving the sentencing judge discretion to choose an actual sentence within a broad range, and because many States provide good-time credits and parole, often permitting release after, say, one-third of the sentence has been served, see, e.g., Alaska Stat. § 33.20.010(a) (2000); Conn. Gen. Stat. § 18-7a (1998). Thus, the statutory maximum is rarely the sentence imposed, and the sentence imposed is rarely the sentence that is served. For the most part, the parties’ briefs discuss sentencing statutes. Nonetheless, that discussion, along with other readily available information, validates my initial belief that Ewing’s sentence, comparatively speaking, is extreme.

As to California itself, we know the following: First, between the end of World War II and 1994 (when California enacted the three strikes law, ante, at 15), no one like Ewing could have served more than 10 years in prison. We know that for certain because the maximum sentence for Ewing’s crime of conviction, grand theft, was for most of that period 10 years. Cal. Penal Code Ann. §§ 484, 489 (West 1970); see Cal. Dept. of Corrections, Offender Information Services, Administrative Services Division, Historical Data for Time Served by Male Felons Paroled from Institutions: 1945 Through 1981, p. 11 (1982) (Table 10) (hereinafter Historical Data for Time Served by California Felons), Lodging of Petitioner. From 1976 to 1994 (and currently, absent application *44of the three strikes penalty), a Ewing-type offender would have received a maximum sentence of four years. Cal. Penal Code Ann. §489 (West 1999), § 667.5(b) (West Supp. 2002). And we know that California’s “habitual offender” laws did not apply to grand theft. §§ 644(a), (b) (West 1970) (repealed 1977). We also know that the time that any offender actually served was likely far less than 10 years. This is because statistical data show that the median time actually served for grand theft (other than auto theft) was about two years, and 90 percent of all those convicted of that crime served less than three or four years. Historical Data for Time Served by California Felons 11 (Table 10).

Second, statistics suggest that recidivists of all sorts convicted during that same time period in California served a small fraction of Ewing’s real-time sentence. On average, recidivists served three to four additional (recidivist-related) years in prison, with 90 percent serving less than an additional real seven to eight years. Id., at 22 (Table 21).

Third, we know that California has reserved, and still reserves, Ewing-type prison time, i. e., at least 25 real years in prison, for criminals convicted of crimes far worse than was Ewing’s. Statistics for the years 1945 to 1981, for example, indicate that typical.(nonrecidivist) male first-degree murderers served between 10 and 15 real years in prison, with 90 percent of all such murderers serving less than 20 real years. Id., at 3 (Table 2). Moreover, California, which has moved toward a real-time sentencing system (where the statutory punishment approximates the time served), still punishes far less harshly those who have engaged in far more serious conduct. It imposes, for example, upon nonrecidi-vists guilty of arson causing great bodily injury a maximum sentence of nine years in prison, Cal. Penal Code Ann. § 451(a) (West 1999) (prison term of 5, 7, or 9 years for arson that causes great bodily injury); it imposes upon those guilty of voluntary manslaughter a maximum sentence of 11 years, § 193 (prison term of 3, 6, or 11 years for voluntary man*45slaughter). It reserves the sentence that it here imposes upon (former-burglar-now-golf-club-thief) Ewing for non-recidivist, first-degree murderers. See § 190(a) (West Supp. 2003) (sentence of 25 years to life for first-degree murder).

As to other jurisdictions, we know the following: The United States, bound by the federal Sentencing Guidelines, would impose upon a recidivist, such as Ewing, a sentence that, in any ordinary case, would not exceed 18 months in prison. USSG §2Bl.l(a) (Nov. 1999) (assuming a base offense level of 6, a criminal history of VI, and no mitigating or aggravating adjustments); id., ch. 5, pt. A, Sentencing Table. The Guidelines, based in part upon a study of some 40,000 actual federal sentences, see supra, at 37, 41, reserve a Ewing-type sentence for Ewing-type recidivists who currently commit such crimes as murder, §2A1.2; air piracy, §2A5.1; robbery (involving the discharge of a firearm, serious bodily injury, and about $1 million), §2B3.1; drug offenses involving more than, for example, 20 pounds of heroin, §2D1.1; aggravated theft of more than $100 million, § 2B1.1; and other .similar offenses. The Guidelines reserve 10 years of real prison time (with good time) — less than 40 percent of Ewing’s sentence — for Ewing-type recidivists who go on to commit, for instance, voluntary manslaughter, §2A1.3; aggravated assault with a firearm (causing serious bodily injury and motivated by money), §2A2.2; kidnaping, §2A4.1; residential burglary involving more than $5 million, §2B2.1; drug offenses involving at least one pound of cocaine, § 2D 1.1; and other similar offenses. Ewing also would not have been subject to the federal “three strikes” law, 18 U. S. C. § 3559(c), for which grand theft is not a triggering offense.

With three exceptions, see infra, at 46-47, we do not have before us information about actual time served by Ewing-type offenders in other States. We do know, however, that the law would make it legally impossible for a Ewing-type offender to serve more than 10 years in prison in 33 jurisdictions, as well as the federal courts, see Appendix, *46Part A, infra, more than 15 years in 4 other States, see Appendix, Part B, infra, and more than 20 years in 4 additional States, see Appendix, Part C, infra. In nine other States, the law might make it legally possible to impose a sentence of 25 years or more, see Appendix, Part D, infra — though that fact by itself, of course, does not mean that judges have actually done so. But see infra this page. I say “might” because the law in five of the nine last mentioned States restricts the sentencing judge’s ability to impose a term so long that, with parole, it would amount to at least 25 years of actual imprisonment. See Appendix, Part D, infra.

We also know that California, the United States, and other States supporting California in this case, despite every incentive to find someone else like Ewing who will have to serve, or who has actually served, a real prison term anywhere approaching that imposed upon Ewing, have come up with precisely three examples. Brief for United States as Amicus Curiae 28-29, n. 13. The Government points to Ex parte Howington, 622 So. 2d 896 (Ala. 1993), where an Alabama court sentenced an offender with three prior burglary convictions and two prior grand theft convictions to “life” for the theft of a tractor-trailer. The Government also points to State v. Heftel, 513 N. W. 2d 397 (S. D. 1994), where a South Dakota court sentenced an offender with seven prior felony convictions to 50 years’ imprisonment for theft. And the Government cites Sims v. State, 107 Nev. 438, 814 P. 2d 63 (1991), where a Nevada court sentenced a defendant with three prior felony convictions (including armed robbery) and nine misdemeanor convictions to life without parole for the theft of a purse and wallet containing $476.

The first of these cases, Howington, is beside the point, for the offender was eligible for parole after 10 years (as in Rummel), not 25 years (as here). Ala. Code §15-22-28(e) (West 1982). The second case, Heftel, is factually on point, but it is not legally on point, for the South Dakota courts did not consider the constitutionality of the sentence. 513 N. W. *472d, at 401. The third case, Sims, is on point both factually and legally, for the Nevada Supreme Court (by a vote of 3 to 2) found the sentence constitutional. I concede that example — a single instance of a similar sentence imposed outside the context of California’s three strikes law, out of a prison population now approaching two million individuals. U. S. Dept. of Justice, Office of Justice Programs, Bureau of Justice Statistics, Prison Statistics (Jan. 8, 2003), http://www.ojp.usdoj.gov/bjs/prisons.htm (available in Clerk of Court’s case file).

The upshot is that comparison of other sentencing practices, both in other jurisdictions and in California at other times (or in respect to other crimes), validates what an initial threshold examination suggested. Given the information available, given the state and federal parties’ ability to provide additional contrary data, and given their failure to do so, we can assume for constitutional purposes that the following statement is true: Outside the California three strikes context, Ewing’s recidivist sentence is virtually unique in its harshness for his offense of conviction, and by a considerable degree.

IV

This is not the end of the matter. California sentenced Ewing pursuant to its “three strikes” law. That law represents a deliberate effort to provide stricter punishments for recidivists. Cal. Penal Code Ann. § 667(b) (West 1999) (“It is the intent of the Legislature ... to ensure longer prison sentences and greater punishment for those who commit a felony and have been previously convicted of serious and/or violent felony offenses”); ante, at 24. And, it is important to consider whether special criminal justice concerns related to California’s three strikes policy might justify including Ewing’s theft within the class of triggering criminal conduct (thereby imposing a severe punishment), even if Ewing’s sentence would otherwise seem disproportionately harsh. *48Cf. Harmelin, 501 U. S., at 998-999, 1001 (noting “the primacy of the legislature” in making sentencing policy).

I can find no such special criminal justice concerns that might justify this sentence. The most obvious potential justification for bringing Ewing’s theft within the ambit of the statute is administrative. California must draw some kind of workable line between conduct that will trigger, and con-' duct that will not trigger, a “three strikes” sentence. “But the fact that a line has to be drawn somewhere does not justify its being drawn anywhere.” Pearce v. Commissioner, 315 U. S. 543, 558 (1942) (Frankfurter, J., dissenting). The statute’s administrative objective would seem to be one of separating more serious, from less serious, triggering criminal conduct. Yet the statute does not do that job particularly well.

The administrative line that the statute draws separates “felonies” from “misdemeanors.” See Brief for Respondent 6 (“The California statute relies, fundamentally, on traditional classifications of certain crimes as felonies”). Those words suggest a graduated difference in degree. But an examination of how California applies these labels in practice to criminal conduct suggests that the offenses do not necessarily reflect those differences. See United States v. Watson, 423 U. S. 411, 438-441 (1976) (Marshall, J., dissenting) (felony/misdemeanor distinction often reflects history, not logic); Rummel, 445 U. S., at 284 (“The most casual review of the various criminal justice systems now in force in the 50 States of the Union shows that the line dividing felony theft from petty larceny, a line usually based on the value of the property taken, varies markedly from one State to another”). Indeed, California uses those words in a way unrelated to the seriousness of offense conduct in a set of criminal statutes called “ ‘wobblers,’ ” see ante, at 16, one of which is at issue in this case.

Most “wobbler” statutes classify the same criminal conduct either as a felony or as a misdemeanor, depending upon *49the actual punishment imposed, Cal. Penal Code Ann. §§ 17(a), (b) (West 1999); ante, at 16-17, which in turn depends primarily upon whether “the rehabilitation of the convicted defendant” either does or does not “require” (or would or would not “be adversely affected by”) “incarceration in a state prison as a felon.” In re Anderson, 69 Cal. 2d 613, 664-665, 447 P. 2d 117, 152 (1968) (Tobriner, J., concurring in part and dissenting in part); ante, at 29. In such cases, the felony/misdemeanor classification turns primarily upon the nature of the offender, not the comparative seriousness of the offender’s conduct.

A subset of “wobbler” statutes, including the “petty theft with a prior” statute, Cal. Penal Code Ann. § 666 (West Supp. 2002), defining the crime in the companion case, Lockyer v. Andrade, post, p. 63, authorizes the treatment of otherwise misdemeanor conduct, see Cal. Penal Code Ann. § 490 (West 1999), as a felony only when the offender has previously committed a property crime. Again, the distinction turns upon characteristics of the offender, not the specific offense conduct at issue.

The result of importing this kind of distinction into California’s three strikes statute is a series of anomalies. One anomaly concerns the seriousness of the triggering behavior. “Wobbler” statutes cover a wide variety of criminal behavior, ranging from assault with a deadly weapon, §245, vehicular manslaughter, § 193(c)(1), and money laundering, § 186.10(a), to the defacement of property with graffiti, § 594(b)(2)(A) (West Supp. 2002), or stealing more than $100 worth of chickens, nuts, or avocados, § 487(b)(1)(A) (West Supp. 2003); § 489 (West 1999). Some of this behavior is obviously less serious, even if engaged in twice, than other criminal conduct that California statutes classify as pure misdemeanors, such as reckless driving, Cal. Veh. Code Ann. §23103 (West Supp. 2003); § 23104(a) (West 2000) (reckless driving causing bodily injury), the use of force or threat of force to interfere with another’s civil rights, Cal. Penal Code *50Ann. §422.6 (West 1999), selling poisoned alcohol, §347b, child neglect, § 270, and manufacturing or selling false government documents with the intent to conceal true citizenship, § 112(a) (West Supp. 2002).

Another anomaly concerns temporal order. An offender whose triggering crime is his third crime likely will not fall within the ambit of the three strikes statute provided that (a) his first crime was chicken theft worth more than $100, and (b) he subsequently graduated to more serious crimes, say, crimes of violence. That is because such chicken theft, when a first offense, will likely be considered a misdemeanor. A similar offender likely will fall within the scope of the three strikes statute, however, if such chicken theft was his third crime. That is because such chicken theft, as a third offense, will likely be treated as a felony.

A further anomaly concerns the offender’s criminal record. California’s “wobbler” “petty theft with a prior” statute, at issue in Lockyer v. Andrade, post, p. 63, classifies a petty theft as a “felony” if, but only if, the offender has a prior record that includes at least one conviction for certain theft-related offenses. Cal. Penal Code Ann. §666 (West Supp. 2002). Thus a violent criminal who has committed two violent offenses and then steals $200 will not fall within the ambit of the three strikes statute, for his prior record reveals no similar property crimes. A similar offender will fall within the scope of the three strikes statute, however, if that offender, instead of having committed two previous violent crimes, has committed one previous violent crime and one previous petty theft. (Ewing’s conduct would have brought him within the realm of the petty theft statute prior to 1976 but for inflation.)

At the same time, it is difficult to find any strong need to define the lower boundary as the State has done. The three strikes statute itself, when defining prior “strikes,” simply lists the kinds of serious criminal conduct that falls within the definition of a “strike.” § 667.5(c) (listing “violent” felon*51ies); § 1192.7(e) (West Supp. 2003) (listing “serious” felonies). There is no obvious reason why the statute could not enumerate, consistent with its purposes, the relevant triggering crimes. Given that possibility and given the anomalies that result from California’s chosen approach, I do not see how California can justify on administrative grounds a sentence as seriously disproportionate as Ewing’s. See Parts II and III, supra.

Neither do I see any other way in which inclusion of Ewing’s conduct (as a “triggering crime”) would further a significant criminal justice objective. One might argue that those who commit several property crimes should receive long terms of imprisonment in order to “incapacitate” them, i. e., to prevent them from committing further crimes in the future. But that is not the object of this particular three strikes statute. Rather, as the plurality says, California seeks “‘to reduce serious and violent crime.’” Ante, at 24 (quoting Ardaiz, California’s Three Strikes Law: History, Expectations, Consequences, 32 McGeorge L. Rev. 1 (2000) (emphasis added)). The statute’s definitions of both kinds of crime include crimes against the person, crimes that create danger of physical harm, and drug crimes. See, e. g., Cal. Penal Code Ann. § 667.5(c)(1) (West Supp. 2002), § 1192.7(c)(1) (West Supp. 2003) (murder or voluntary manslaughter); §667.5(c)(21) (West Supp. 2002), § 1192.7(c)(18) (West Supp. 2003) (first-degree burglary); § 1192.7(c)(24) (selling or giving or offering to sell or give heroin or cocaine to a minor). They do not include even serious crimes against property, such as obtaining large amounts of money, say, through theft, embezzlement, or fraud. Given the omission of vast categories of property crimes — including grand theft (unarmed)— from the “strike” definition, one cannot argue, on property-crime-related incapacitation grounds, for inclusion of Ewing’s crime among the triggers.

Nor do the remaining criminal law objectives seem relevant. No one argues for Ewing’s inclusion within the ambit *52of the three strikes statute on grounds of “retribution.” Cf. Vitiello, Three Strikes: Can We Return to Rationality? 87 J. Crim. L. & C. 395, 427 (1997) (California’s three strikes law, like other “[hjabitual offender statutes[, is] not retributive” because the term of imprisonment is “imposed without regard to the culpability of the offender or [the] degree of social harm caused by the offender’s behavior,” and “has little to do with the gravity of the offens[e]”). For reasons previously discussed, in terms of “deterrence,” Ewing’s 25-year term amounts to overkill. See Parts II and III, supra. And “rehabilitation” is obviously beside the point. The upshot is that, in my view, the State cannot find in its three strikes law a special criminal justice need sufficient to rescue a sentence that other relevant considerations indicate is unconstitutional.

V

Justice Scalia and Justice Thomas argue that we should not review for gross disproportionality a sentence to a term of years. Ante, at 81 (Scalia, J., concurring in judgment); ante, at 32 (Thomas, J., concurring in judgment). Otherwise, we make it too difficult for legislators and sentencing judges to. determine just when their sentencing laws and practices pass constitutional muster.

I concede that a bright-line rule would give legislators and sentencing judges more guidance. But application of the Eighth Amendment to a sentence of a term of years requires a case-by-ease approach. And, in my view, like that of the plurality, meaningful enforcement of the Eighth Amendment demands that application — even if only at sentencing’s outer bounds.

A case-by-case approach can nonetheless offer guidance through example. Ewing’s sentence is, at a minimum, 2 to 3 times the length of sentences that other jurisdictions would impose in similar circumstances. That sentence itself is sufficiently long to require a typical offender to spend virtually all the remainder of his active life in prison. These and the *53other factors that I have discussed, along with the questions that I have asked along the way, should help to identify “gross disproportionality” in a fairly objective way — at the outer bounds of sentencing.

In stun, even if I accept for present purposes the plurality’s analytical framework, Ewing’s sentence (life imprisonment with a minimum term of 25 years) is grossly disproportionate to the triggering offense conduct — stealing three golf clubs — Ewing’s recidivism notwithstanding.

For these reasons, I dissent.

APPENDIX TO OPINION OF BREYER, J.

A

Thirty-three jurisdictions, as well as the federal courts, have laws that would make it impossible to sentence a Ewing-type offender to more than 10 years in prison:1

Federal: 12 to 18 months. USSG §2B1.1 (Nov. 1999); id., ch. 5, pt. A, Sentencing Table.

Alaska: three to five years; presumptive term of three years. Alaska Stat. §§ 11.46.130(a)(1), (c), 12.55.125(e) (2000).

Arizona: four to six years; presumptive sentence of five years. Ariz. Rev. Stat. Ann. §§ 13-604(C), 13-1802(E) (West 2001).

Connecticut: 1 to 10 years. Conn. Gen. Stat. §§53a-35a(6), 53a-40(j), 53a-124(a)(2) (2001).

Delaware: not more than two years. Del. Code Ann., Tit. 11, § 840(d) (Supp. 2000); § 4205(b)(7) (1995). Recidivist offender penalty not applicable. See §4214; Buckingham v. State, 482 A. 2d 327 (Del. 1984).

District of Columbia: not more than 10 years. D. C. Code Ann. §22-3212(a) (West 2001). Recidivist offender penalty *54not applicable. See § 22-1804a(c)(2) (West 2001) (amended 2001).

Florida: not more than 10 years. Fla. Stat. Ann. §§775.084(l)(a), (4)(a)(3) (West 2000) (amended 2002); § 812.014(c)(1) (West 2000).

Georgia: 10 years. Ga. Code Ann. § 16-8-12(a)(1) (1996); § 17-10-7(a) (Supp. 1996).

Hawaii: 20 months. Haw. Rev. Stat. §§ 708-831(1)(b), 706-606.5(1)(a)(iv), (7)(a) (Supp. 2001).

Idaho: 1 to 14 years. Idaho Code §§ 18-2403, 18-2407(b)(1), 18-2408(2)(a) (1948-1997). Recidivist/habitual' offender penalty of five years to life in prison, §19-2514, likely not applicable. Idaho has a general rule that “ ‘convictions entered the same day or charged in the same information should count as a single conviction for purposes of establishing habitual offender status.’ ” State v. Harrington, 133 Idaho 563, 565, 990 P. 2d 144, 146 (App. 1999) (quoting State v. Brandt, 110 Idaho 341, 344, 715 P. 2d 1011, 1014 (App. 1986)). However, “the nature of the convictions in any given situation must be examined to make certain that [this] general rule is appropriate.” Ibid. In this case, Ewing’s prior felony convictions stemmed from acts committed at the same apartment complex, and three of the four felonies were committed within a day of each other; the fourth offense was committed five weeks earlier. See App. 6; Tr. 45-46 (Information, Case No. NA018343-01 (Cal. Super. Ct.) (available in Clerk of Court’s ease file)). A review of Idaho ease law suggests that this case is factually distinguishable from cases in which the Idaho courts have declined to adhere to the general rule. See, e. g., Brandt, supra, at 348, 344, 715 P. 2d, at 1013, 1014 (three separately charged property offenses involving three separate homes and different victims committed “during a two-month period”); State v. Mace, 133 Idaho 903, 907, 994 P. 2d 1066, 1070 (App. 2000) (unrelated crimes (grand theft and DU I) committed on different dates in different counties); State v. Smith, 116 Idaho 553, 560, 777 *55P. 2d 1226, 1233 (App. 1989) (separate and distinguishable crimes committed on different victims in different counties).

Illinois: two to five years. Ill. Comp. Stat., ch. 730, §5/ 5-8-1(a)(6) (Supp. 2001); ch. 720, § 5/16-1(b)(4). Recidivist offender penalty not applicable. § 5/33B-1(a) (2000).

Indiana: 18 months (with not more than 18 months added for aggravating circumstances). Ind. Code § 35-43-4-2(a) (1993); §35-50-2-7(a). Recidivist offender penalty not applicable. See § 35-50-2-8 (amended 2001).

Iowa: three to five years. Iowa Code Ann. §§714.2(2), 902.9(5) (West Supp. 2002); §902.8 (West 1994).

Kansas: 9 to 11 months. Kan. Stat. Ann. §§21-3701(b)(2), 21-4704(a) (1995). Recidivist offender penalty not applicable. See §21-4504(e)(3).

Kentucky: 5 to 10 years. Ky. Rev. Stat. Ann. § 514.030(2) (Lexis Supp. 2002); §§ 532.060(2)(c), (d), 532.080(2), (5) (Lexis 1999).

Maine: less than one year. Me. Rev. Stat. Ann., Tit. 17-A, §353 (West 1983); §362(4)(B) (West Supp. 2000) (amended 2001); §1252(2)(D) (West 1983 and Supp. 2002). Recidivist offender penalty not applicable. See § 1252(4-A) (West Supp. 2000) (amended 2001).

Massachusetts: not more than five years. Mass. Gen. Laws, ch. 266, §30(1) (West 2000). Recidivist offender penalty not applicable. See ch. 279, §25 (West 1998); Commonwealth v. Hall, 397 Mass. 466, 468, 492 N. E. 2d 84, 85 (1986).

Minnesota: not more than five years. Minn. Stat. § 609.52, subd. 3(3)(a) (2002). Recidivist offender penalty not applicable. See §609.1095, subd. 2.

Mississippi: not more than five years. Miss. Code Ann. § 97-17-41(1)(a) (Lexis 1973-2000). Recidivist offender penalty not applicable. See §99-19-81.

Nebraska: not more than five years. Neb. Rev. Stat. § 28-105(1) (2000 Cum. Supp.); §28-518(2) (1995). Recidivist offender penalty not applicable. See §29-2221(1).

*56New Jersey: Extended term of between 5 to 10 years (instead of three to five years, N. J. Stat. Ann. § 2C:43-6 (1995)), § 2C:43-7(a)(4) (Supp. 2002), whether offense is treated as theft, § 2C:20-2(b)(2)(a), or shoplifting, §§2C:20-ll(b), (c)(2), because, even if Ewing’s felonies are regarded as one predicate crime, Ewing has been separately convicted and sentenced for at least one other crime for which at least a 6-month sentence was authorized, §2C:44-3(a); §2C:44-4(c) (1995).

New Mexico: 30 months. N. M. Stat. Ann. §30-16-20(B)(3) (1994); §31-18-15(A)(6) (2000); §31-18-17(B) (2000) (amended 2002).

New York: three to four years. N. Y. Penal Law § 70.06(3)(e) (West 1998); §155.30 (West 1999).

North Carolina: 4 to 25 months (with exact sentencing range dependent on details of offender’s criminal history). N. C. Gen. Stat. §§ 15A-1340.14, 15A-1340.17(c), (d), 14-72(a) (2001). Recidivist offender penalty not applicable. See §§ 14-7.1, 14-7.6.

North Dakota: not more than 10 years. N. D. Cent. Code § 12.1-23-05(2)(a) (1997); §§ 12.1-32-09(1), (2)(c) (1997) (amended 2001).

Ohio: 6 to 12 months. Ohio Rev. Code Ann. §§ 2913.02(B)(2), 2929.14(A)(5) (West Supp. 2002). No general recidivist statute.

Oregon: not more than five years. Ore. Rev. Stat. §161.605 (1997); Ore. Rev. Stat. Ann. §§ 164.055(l)(a), (3) (Supp. 1998). No general recidivist statute.

Pennsylvania: not more than five years (if no more than one prior theft was “retail theft”); otherwise, not more than seven years. Pa. Stat. Ann., Tit. 18, §§ 1103(3), 1104(1) (Pur-don 1998); §§ 3903(b), 3929(b)(l)(iii)-(iv) (Purdon Supp. 2002); § 3921 (Purdon 1983). Recidivist offender penalty not applicable. See 42 Pa. Cons. Stat. § 9714(a)(1) (1998).

*57Rhode Island: not more than 10 years. R. I. Gen. Laws § 11-41-5(a) (2002). Recidivist offender penalty not applicable. See § 12-19-21(a).

South Carolina: not more than five years. S. C. Code Ann. §§ 16-13-30, 16-13-110(B)(2) (West 2001 Cum. Supp.). Recidivist offender penalty not applicable. See § 17-25-45.

Tennessee: four to eight years. Tenn. Code Ann. §§39-14-105(3), 40-35-106(a)(1), (c), 40-35-112(b)(4) (1997).

Utah: not more than five years. Utah Code Ann. § 76-3-203(3) (1999) (amended 2000); § 76-6-412(l)(b)(i) (1999). Recidivist offender penalty not applicable. See §76-3-203.5 (Supp. 2002).

Washington: not more than 14 months (with exact sentencing range dependent on details of offender score), Wash. Rev. Code §§ 9A.56.040(l)(a), (2) (2000); §§9.94A.510(1), 9.94A.515, 9.94A.525 (2003 Supp. Pamphlet); maximum sentence of five years, §§9A.56.040(l)(a), (2), 9A.20.021(l)(c) (2000). Recidivist offender penalty not applicable. See §§9.94A.030(27), (31) (2000); § 9.94A.570 (2003 Supp. Pamphlet).

Wyoming: not more than 10 years. Wyo. Stat. Ann. § 6-3-404(a)(i) (Michie 2001). Recidivist offender penalty not applicable. See § 6-10-201(a).

B

In four other States, a Ewing-type offender could not have received a sentence of more than 15 years in prison:

Colorado: 4 to 12 years for “extraordinary aggravating circumstances” (e. g., defendant on parole for another felony at the time of commission of the triggering offense). Colo. Rev. Stat. §§ 18-l-105(l)(a)(V)(A), 18-l-105(9)(a)(II), 18-4-401(2)(c) (2002). Recidivist offender penalty not applicable. See §§ 16-13-101(f)(1.5), (2) (2001).

Maryland: not more than 15 years. Md. Ann. Code, Art. 27, § 342(f)(1) (1996) (repealed 2002). Recidivist offender penalty not applicable. See § 643B.

*58New Hampshire: not more than 15 years. N. H. Stat. Ann. §§637:ll(I)(a), 651:2(II)(a) (West Supp. 2002). Recidivist offender penalty not applicable. See §651:6(I)(c).

Wisconsin: not more than 11 years (at the time of Ewing’s offense). Wis. Stat. Ann. § 939.50(3)(e) (West Supp. 2002); §§939.62(l)(b), (2), 943.20(3)(b) (West 1996) (amended 2001). Wisconsin subsequently amended the relevant statutes so that a Ewing-type offender would only be eligible for a sentence of up to three years. See §§ 939.51(3)(a), 943.20(3)(a), 939.62(l)(a) (West Supp. 2003). And effective February 1, 2003, such an offender is eligible for a sentence of only up to two years. See §§ 939.51(3)(a), 943.20(3)(a), 939.62(l)(a).

C

In four additional States, a Ewing-type offender could not have been sentenced to more than’20 years in prison:

Arkansas: 3 to 20 years. Ark. Code Ann. § 5-36-103(b)(2)(A) (1997); §§ 5-4-501(a)(2)(D), (e)(1) (1997) (amended 2001). Eligible for parole after serving one-third of the sentence. § 5-4-501 (1997); § 16-93-608 (1987).

Missouri: not more than 20 years. Mo. Rev. Stat. §558.016(7X3) (2000); § 570.030(3)(1) (2000) (amended 2002). Eligible for parole after 15 years at the latest. §558.011(4)(l)(c).

Texas: 2 to 20 years. Tex. Penal Code Ann. §§ 12.33(a), 12.35(c)(2)(A) (1994); §§ 12.42(a)(3), 31.03(e)(4)(D) (Supp. 2003). Eligible for parole after serving one-fourth of sentence. Tex. Govt. Code Ann. §508.145(f) (Supp. 2003).

Virginia: statutory range of 1 to 20 years (or less than 12 months at the discretion of the jury or court following bench trial), Va. Code Ann. § 18.2-95 (Supp. 2002), but discretionary sentencing guideline ranges established by the Virginia Sentencing Commission, §§ 17.1-805, 19.2-298.01 (2000), with a maximum of 6 years, 3 months, to 15 years, 7 months, see Virginia Criminal Sentencing Commission, Virginia Sentencing Guidelines Manual, Larceny — Section C Recommenda*59tion Table (6th ed. 2002) (with petitioner likely falling within the discretionary guideline range of 2 years, 1 month, to 5 years, 3 months, see Brief for Petitioner 33, n. 25). Recidivist offender penalty not applicable. See § 19.2-297.1 (2000).

D

In nine other States, the law might make it legally possible to impose a sentence of 25 years or more upon a Ewing-type offender. But in five of those nine States,2 the offender would be parole-eligible before 25 years:

Alabama: “life or any term of not less than 20 years.” Ala. Code § 13A-5-9(c)(2) (Lexis Supp. 2002); §§ 13A-8-3(a), (c) (1994). Eligible for parole after the lesser of one-third of the sentence or 10 years. § 15-22-28(e) (1995).

Louisiana: Louisiana courts could have imposed a sentence of life without the possibility of parole at the time of Ewing’s offense. La. Stat. Ann. §§ 14:67.10(B)(1), 14:2(4), (13)(y) (West Supp. 2003); §§ 15:529.1(A)(l)(b)(ii) and (c)(iMii) (West 1992) (amended 2001). Petitioner argues that, despite the statutory authority to impose such a sentence, Louisiana courts would have carefully scrutinized his life sentence, as they had in other cases involving recidivists charged with a nonviolent crime. Brief for Petitioner 35-36, n. 29; see Brief for Families Against Mandatory Minimums as Amicus Curiae 24-25, and n. 21; State v. Hayes, 98-1526, p. 4 (La. App. 6/25/99), 739 So. 2d 301, 303-304 (holding that a life sentence was impermissibly excessive for a defendant convicted of theft of over $1,000, who had a prior robbery conviction). But see Brief for Respondent 45-46, n. 12 (contesting petitioner’s argument). Louisiana has amended its recidivist statute to require that the triggering offense be a violent felony, and that the offender have at least two prior violent felony convictions to be eligible for a life sentence. La. Stat. *60Ann. § 15:529.l(A)(l)(b)(ii) (West Supp. 2003). Under current law, a Ewing-type offender would face a sentence of 6% to 20 years. §§ 14:67.10(B)(1), 15:529.1(A)(b)(i).

Michigan: “imprisonment for life or for a lesser term,” Mich. Comp. Laws Ann. §769.12(l)(a) (West 2000) (instead of “not more than 15 years,” §769.12(l)(b), as petitioner contends, see Brief for Petitioner 34, n. 26; Brief for Families Against Mandatory Mínimums as Amicus Curiae 16-17, n. 15, 22-23, n. 20), because the triggering offense is “punishable upon a first conviction by imprisonment for a maximum term of 5 years or more,” § 769.12(l)(a) (West 2000). The larceny for which Ewing was convicted was, under Michigan law, “a felony punishable by imprisonment for not more than 5 years.” § 750.356(3)(a) (West Supp. 2002). Eligible for parole following minimum term set by sentencing judge. §769.12(4) (West 2000).

Montana: 5 to 100 years. Mont. Code Ann. §45-6-301(7)(b) (1999); §§46-18-501, 46-18-502(1) (2001). A Ewing-type offender would not have been subject to a minimum term of 10 years in prison (as the State suggests, Brief for Respondent 44) because Ewing does not meet the requirements of §46-18-502(2) (must be a “persistent felony offender,” as defined in §46-18-501, at the time of the offender’s previous felony conviction). See Reply Brief for Petitioner 18, n. 14. Eligible for parole after one-fourth of the term. §46-23-201(2).

Nevada: “life without the possibility of parole,” or “life with the possibility of parole [after serving] 10 years,” or “a definite term of 25 years, with eligibility for parole [after serving] 10 years.” Nev. Rev. Stat. §§207.010(l)(b)(l)-{3) (1995).

Oklahoma: not less than 20 years (at the time of Ewing’s offense). Okla. Stat., Tit. 21, § 51.1(B) (West Supp. 2000) (amended in 2001 to four years to life, § 51.1(C) (West 2001)); § 1704 (West 1991) (amended 2001). Eligible for parole after serving one-third of sentence. Tit. 57, § 332.7(B) (West *612001). Thus, assuming a sentence to a term of years of up to 100 years (as in Montana, see supra, at 60), parole eligibility could arise as late as after 33 years.

South Dakota: maximum penalty of life imprisonment, with no minimum term. S. D. Codified Laws §22-7-8 (1998); §22-30A-17(l) (Supp. 2002). Eligible for parole after serving one-half of sentence. §24-15-5(3) (1998). Thus, assuming a sentence to a term of years of up to 100 years (as in Montana, see supra, at 60), parole eligibility could arise as late as after 50 years.

Vermont: “up to and including life,” Vt. Stat. Ann., Tit. 13, § 11 (1998), or not more than 10 years, §2501; State v. Angelucci, 137 Vt. 272, 289-290, 405 A. 2d 33, 42 (1979) (court has discretion to sentence habitual offender to the sentence that is specified for grand larceny alone). Eligible for parole after six months. Tit. 28, § 501 (2000) (amended 2001).

West Virginia: Petitioner contends that he would only have been subject to a misdemeanor sentence of not more than 60 days for shoplifting, W. Va. Code §§61-3A-1, 61-3A-3(a)(2) (2000); Brief for Petitioner 31, n. 19, 33-34, n. 25. However, a Ewing-type offender could have been charged with grand larceny, see State ex rel. Chadwell v. Duncil, 196 W. Va. 643, 647-648, 474 S. E. 2d 573, 577-578 (1996) (prosecutor has discretion to charge defendant with either shoplifting or grand larceny), a felony punishable by imprisonment in the state penitentiary for 1 to 10 years (or, at the discretion of the trial court, not more than 1 year in jail). §61-3-13(a). Under West Virginia’s habitual offender statute, a felon “twice before convicted ... of a crime punishable by confinement in a penitentiary.. . . shall be sentenced to . . . life [imprisonment],” § 61 — 11—18(c), with parole eligibility after 15 years, § 62-12-13(c). Amicus curiae on behalf of petitioner notes that, in light of existing state-law precedents, West Virginia courts “would not countenance a sentence of life without the possibility of parole for 25 years for shoplifting golf clubs.” Brief for Families Against Mandatory Mínimums as Amicus *62Curiae 25-26 (citing State v. Barker, 186 W. Va. 73, 74-75, 410 S. E. 2d 712, 713-714 (1991) (per curiam); and State v. Deal, 178 W. Va. 142, 146-147, 358 S. E. 2d 226, 230-231 (1987)). But see Brief for Respondent 45, n. 11 (contesting that argument).