This case turns on whether the federal carjacking statute, 18 U.S.C. § 2119 as it was when petitioner was charged, defined three distinct offenses or a single crime with a choice of three maximum penalties, two of them dependent on sentencing factors exempt from the requirements of charge and jury verdict. We think the better reading is of three distinct offenses, particularly in light of the rule that any interpretive uncertainty should be resolved to avoid serious questions about the statute's constitutionality.

* In December 1992, petitioner, Nathaniel Jones, and two others, Oliver and McMillan, held up two men, Mutanna and Mardaie. While Jones and McMillan went through the victims' pockets, Oliver stuck his gun in Mutanna's left ear, and later struck him on the head. Oliver and McMillan made their getaway in the Cadillac Jones had driven to the scene, while Jones forced Mardaie into Mutanna's Honda and drove off after them. After stopping to put Mardaie out, Jones sped away in the stolen car subject to police pursuit, which ended when Jones crashed into a telephone pole. United States v. Oliver, 60 F.3d 547, 549 (CA9 1995); Tr. 159, 387, 310 (July 27 28, 1993).

A grand jury in the Eastern District of California indicted Jones and his two accomplices on two counts: using or aiding and abetting the use of a firearm during and in relation to a crime of violence, in violation of 18 U.S.C. § 924(c), and carjacking or aiding and abetting carjacking, in violation of 18 U.S.C. § 2119 which then read as follows:

"Whoever, possessing a firearm as defined in section 921 of this title, takes a motor vehicle that has been transported, shipped, or received in interstate or foreign commerce from the person or presence of another by force and violence or by intimidation, or attempts to do so, shall

"(1) be fined under this title or imprisoned not more than 15 years, or both,

"(2) if serious bodily injury (as defined in section 1365 of this title) results, be fined under this title or imprisoned not more than 25 years, or both, and

"(3) if death results, be fined under this title or imprisoned for any number of years up to life, or both." 18 U.S.C. § 2119 (1988 ed., Supp. V).1

The indictment made no reference to the statute's numbered subsections and charged none of the facts mentioned in the latter two, and at the arraignment the Magistrate Judge told Jones that he faced a maximum sentence of 15 years on the carjacking charge. App. 4 5, 7. Consistently with this advice, the District Court's subsequent jury instructions defined the elements subject to the Government's burden of proof by reference solely to the first paragraph of §2119, with no mention of serious bodily injury. Id., at 10. The jury found Jones guilty on both counts.

The case took a new turn, however, with the arrival of the presentence report, which recommended that petitioner be sentenced to 25 years for the carjacking because one of the victims had suffered serious bodily injury. The report noted that Mutanna had testified that Oliver's gun caused profuse bleeding in Mutanna's ear, and that a physician had concluded that Mutanna had suffered a perforated eardrum, with some numbness and permanent hearing loss. Id., at 15 16; 60 F.3d, at 554. Jones objected that the 25-year recommendation was out of bounds, since serious bodily injury was an element of the offense defined in part by §2119(2), which had been neither pleaded in the indictment nor proven before the jury. App. 12 13. The District Court saw the matter differently and, based on its finding that the serious bodily injury allegation was supported by a preponderance of the evidence, imposed a 25-year sentence on the carjacking count, ibid., together with a consecutive 5-year sentence for the firearm offense, 60 F.3d, at 549.

Like the trial court, the Court of Appeals did not read §2119(2) as setting out an element of an independent offense.2 Id., at 551 554. The Ninth Circuit thus agreed with the Eleventh, see United States v. Williams, 51 F.3d 1004, 1009 1010 (1995), in reasoning that the structure of the statute, particularly the grammatical dependence of the numbered subsections on the first paragraph, demonstrated Congress's understanding that the subsections did not complete the definitions of separate crimes. 60 F.3d, at 552 553. For its view that the subsections provided sentencing factors, the court found additional support in the statute's legislative history. The heading on the subtitle of the bill creating §2119 was "Enhanced Penalties for Auto Theft," which the court took as indicating that the statute's numbered subsections merely defined sentencing enhancements. Id., at 553. The court also noted several references in the Committee Reports and floor debate on the bill to enhanced penalties for an apparently single carjacking offense. Ibid. Because of features arguably distinguishing this case from Almendarez-Torres v. United States, 523 U.S. 224 (1998), we granted certiorari, 523 U.S. 1045 (1998), and now reverse.

Much turns on the determination that a fact is an element of an offense rather than a sentencing consideration, given that elements must be charged in the indictment, submitted to a jury, and proven by the Government beyond a reasonable doubt. See, e.g., Hamling v. United States, 418 U.S. 87, 117 (1974); United States v. Gaudin, 515 U.S. 506, 509 510 (1995). Accordingly, some statutes come with the benefit of provisions straightforwardly addressing the distinction between elements and sentencing factors. See McMillan v. Pennsylvania, 477 U.S. 79, 85 86 (1986) (express identification of statutory provision as sentencing factor). Even without any such help, however, §2119 at first glance has a look to it suggesting that the numbered subsections are only sentencing provisions. It begins with a principal paragraph listing a series of obvious elements (possession of a firearm, taking a motor vehicle, connection with interstate commerce, and so on). That paragraph comes close to standing on its own, followed by sentencing provisions, the first of which, subsection (1), certainly adds no further element. But the superficial impression loses clarity when one looks at the penalty subsections (2) and (3). These not only provide for steeply higher penalties, but condition them on further facts (injury, death) that seem quite as important as the elements in the principal paragraph (e.g., force and violence, intimidation). It is at best questionable whether the specification of facts sufficient to increase a penalty range by two-thirds, let alone from 15 years to life, was meant to carry none of the process safeguards that elements of an offense bring with them for a defendant's benefit. The "look" of the statute, then, is not a reliable guide to congressional intentions, and the Government accordingly advances two, more subtle structural arguments for its position that the fact specified in subsection (2) is merely a sentencing factor.

Like the Court of Appeals, the Government stresses that the statute's numbered subsections do not stand alone in defining offenses, most of whose elements on anyone's reckoning are set out in the statute's opening paragraph. This integrated structure is said to suggest that the statute establishes only a single offense. To the same point, the Government argues that the numbered subsections come after the word "shall," which often divides offense-defining provisions from those that specify sentences. Brief for United States 15 18. While these points are sound enough as far as they go, they are far short of dispositive even on their own terms, whereas they are weakened here by a number of countervailing structural considerations. First, as petitioner notes, Reply Brief for Petitioner 1 2, if the shorter subsection (2) of §2119 does not stand alone, neither does the section's more voluminous first paragraph. In isolation, it would merely describe some very obnoxious behavior, leaving any reader assuming that it must be a crime, but never being actually told that it is. Only the numbered subsidiary provisions complete the thought. Section 2119 is thus unlike most offense-defining provisions in the federal criminal code, which genuinely stand on their own grammatical feet thanks to phrases such as "shall be unlawful," see, e.g., 18 U.S.C. § 922(g), "shall be punished," see, e.g., §511A(a), or "shall be guilty of," see, e.g., 18 U.S.C. § 514 (1994 ed., Supp. II), which draw a provision to its close. Second, as for the significance of the word "shall," although it frequently separates offense-defining clauses from sentencing provisions, it hardly does so invariably. One of the robbery statutes that served as a model for §2119,3 see 18 U.S.C. § 2118(a)(3), (b)(3), for example, places elements of the offense on either side of "shall." And, of course, where the supposedly "elements" side is itself grammatically incomplete (as here), the placement of "shall" is oddly equivocal. Indeed, both the Government and the Courts of Appeals treat the statute perhaps most closely resembling this one, §1365(a) (consumer tampering), as defining basic and aggravated offenses, one of which is defined in terms of serious bodily injury. See, e.g., United States v. Meling, 47 F.3d 1546, 1551 (CA9 1995).

These clues derived from attention to structure and parsing of wording, like those the dissent holds up to distinguish the carjacking act both from the robbery statutes upon which it was modeled and state aggravated robbery statutes, see post, at 7 9, 10 11, turn out to move us only so far in our effort to infer congressional intent. The text alone does not justify any confident inference. But statutory drafting occurs against a backdrop not merely of structural conventions of varying significance, but of traditional treatment of certain categories of important facts, like the degree of injury to victims of crime, in relation to particular crimes. If a given statute is unclear about treating such a fact as element or penalty aggravator, it makes sense to look at what other statutes have done, on the fair assumption that Congress is unlikely to intend any radical departures from past practice without making a point of saying so.

We engaged in just such an enquiry this past Term in Almendarez-Torres, where we stressed the history of treating recidivism as a sentencing factor, and noted that, with perhaps one exception, Congress had never clearly made prior conviction an offense element where the offense conduct, in the absence of recidivism, was independently unlawful. 523 U.S., at 230. Here, on the contrary, the search for comparable examples more readily suggests that Congress had separate and aggravated offenses in mind when it employed the scheme of numbered subsections in §2119. Although Congress has explicitly treated serious bodily injury as a sentencing factor, see, e.g., 18 U.S.C. § 2262(b)(2) (interstate violation of a protection order); §248(b)(2) (free access to clinic entrances; bodily injury), it has unmistakably identified serious bodily injury as an offense element in any number of statutes, see, e.g., 10 U.S.C. § 928(b)(2) (assault by a member of the armed forces); 18 U.S.C. § 37(a)(1) (violence at international airports); §1091(a)(2) (genocide). The likelihood that Congress understood injury to be an offense element here follows all the more from the fact that carjacking is a type of robbery, and serious bodily injury has traditionally been treated, both by Congress and by the state legislatures, as defining an element of the offense of aggravated robbery. As the Government acknowledges, Brief for United States 20 21, and n. 8, Congress modeled the federal carjacking statute on several other federal robbery statutes.4 One of them, 18 U.S.C. § 2118 (robbery involving controlled substances), clearly makes causing serious bodily injury an element of the offense. It provides that "[w]hoever takes or attempts to take from the person or presence of another by force or violence or by intimidation any [of certain controlled substances] shall be fined or imprisoned not more than twenty years, or both, if (1) the replacement cost of the [controlled substance] was not less than $500, . . . or (3) another person was killed or suffered significant bodily injury as a result of such taking or attempt." §2118(a)(3); see also §2118(b)(3).5 A second model, §2113 (bank robbery), as the Government concedes, see Brief for United States 17, makes related facts of violence, that is, assault and jeopardizing life by using a dangerous weapon, elements defining an aggravated form of that type of robbery. See §§2113(d), (e); cf. Almendarez-Torres, supra, at 231 (citing bank robbery statute as example of statute establishing greater and lesser included offenses); McMillan, 477 U.S., at 88 (contrasting §2113(d) with provision defining a sentencing enhancement).

When pressed at oral argument, the Government proved unable to explain why Congress might have chosen one treatment of serious bodily harm or violence in defining two of the three offenses it used as its models for §2119 and a different treatment in writing the carjacking statute itself, see Tr. of Oral Arg. 41 44, and we are unable to imagine a convincing reason ourselves. We thus think it fair to say that, as in the earlier robbery statutes, so in the carjacking statute, Congress probably intended serious bodily injury to be an element defining an aggravated form of the crime.

State practice bolsters the conclusion. Many States use causation of serious bodily injury or harm as an element defining a distinct offense of aggravated robbery. See, e.g., Ala. Code §13A 8 41(a)(2) (1994) (robbery in the first degree defined in part by the causing of "serious physical injury"); Alaska Stat. Ann. §11.41.500(a)(3) (1996) (same); Ark. Code Ann. §5 12 103 (1997) (aggravated robbery; "[i]nflicts or attempts to inflict death or serious physical injury"); Conn. Gen. Stat. §53a 134(a)(1) (1994) (robbery in the first degree; "[c]auses serious physical injury"); Iowa Code §711.2 (1993) (robbery in the first degree; "purposely inflicts or attempts to inflict serious injury"); Kans. Stat. Ann. §21 3427 (1995) (aggravated robbery; "inflicts bodily harm"); Ky. Rev. Stat. Ann. §515.020(1)(a) (Michie 1990) (robbery in the first degree; "causes physical injury"); N. H. Rev. Stat. Ann. §636:1(III)(c) (1996) (class A felony of robbery; "[i]nflicted or attempted to inflict death or serious injury"); N. Y. Penal Law §160.15 (McKinney 1988) (robbery in the first degree; "[c]auses serious physical injury"); Ore. Rev. Stat. §164.415(1)(c) (1990) (robbery in the first degree; "[c]auses or attempts to cause serious physical injury"); Tex. Penal Code Ann. §29.03(a)(1) (1994) (aggravated robbery; "causes serious bodily injury"); Utah Code Ann. §76 6 302(1)(b) (1995) (aggravated robbery; "causes serious bodily injury"); Wash. Rev. Code §9A.56.200(1)(c) (1994) (robbery in the first degree; "[i]nflicts bodily injury"). While the state practice is not, admittedly, direct authority for reading the federal carjacking statute, it does show that in treating serious bodily injury as an element, Congress would have been treading a well-worn path.

Despite these indications and the equivocal structural clues, the Government suggests that a 1996 amendment supports its reading of the carjacking statute as previously enacted. In the Carjacking Correction Act of 1996, 110 Stat. 3020, Congress provided that the term "serious bodily injury" in subsection (2) should include sexual abuse and aggravated sexual abuse as defined in §§2241 and 2242. The Government points to several statements in the 1996 amendment's legislative history in which subsection (2) is described as providing a "penalty enhancement," see, e.g., H. R. Rep. No. 104 787, pp. 2, 3 (1996), as showing that subsection (2) defines a sentencing factor. Even those of us disposed to treat legislative history as authority, however, find the quoted statements unimpressive. Assuming that "penalty enhancement" was meant to be synonymous with "sentencing factor," the legislative history also contains contrary indications in some of the statements made by the 1996 amendment's sponsors, suggesting an assumption that subsection (2) established an element or elements that had to be proven at trial. See 142 Cong. Rec. S9154 (July 30, 1996) (statement of Sen. Biden) ("[T]he defendant had been convicted of raping the woman") (emphasis added). This hardly seems the occasion to doubt that "subsequent legislative history is a 'hazardous basis for inferring the intent of an earlier' Congress." Pension Benefit Guaranty Corporation v. LTV Corp., 496 U.S. 633, 650 (1990) (quoting United States v. Price, 361 U.S. 304, 313 (1960)). Indeed, our leeriness of relying on hindsight expressed in legislative history is only confirmed by recognizing what oddity there would be in defining the fact of serious bodily injury by reference to a distinct offense with its own offense elements, like sexual abuse, while at the same time assuming that the fact so defined is merely a sentencing consideration.

Nor do we think the legislative history that attracted the Court of Appeals is any more helpful to the Government. See 60 F.3d, at 553. The Committee Reports and floor debate on the statute refer to its augmentation of the criminal law in the singular, not the plural, speaking only of a new federal "crime" or "offense" of carjacking in the singular. See, e.g., H. R. Rep. No. 102 851, pt. 1, p. 17 (1992); 138 Cong. Rec. 32500 (1992) (statement of Rep. Dingell). But what we make of the singular-plural distinction turns on the circumstances. Characterizing a cluster of provisions as enacting something to be described by the singular terms "offense" or "crime" would signify a good deal if the speakers or writers were addressing a point on which the distinction mattered. That is not, however, what they were doing in the passages cited, where those references couched in the singular did not occur in discussions of the issue of offense elements versus sentencing factors that we confront here. So, we think their significance is slight. On the subject of legislative history, we should add that we see nothing favorable to the Government in the fact that the statement in the House Report explaining that the drafters of the carjacking statute drew on the examples of other federal robbery statutes referred to an early version of the carjacking statute when it lacked any reference to the aggravated forms of the offense now defined by subsections (2) and (3). See H. R. Rep. No. 102 851, supra, at 17. As against the suggestion that Congress looked to the earlier robbery statutes only when it settled on the language contained in the carjacking statute's first paragraph, we think it would have been strange for Congress to find guidance in the other robbery statutes at the beginning of the legislative process and then just forget about them. As the Government itself suggests in a somewhat different context, there is no reason to think that Congress "might have abandoned [those] ready federal models" in developing the more fully elaborated version of the statute that it ultimately adopted. Brief for United States 21, n. 8.

While we think the fairest reading of §2119 treats the fact of serious bodily harm as an element, not a mere enhancement, we recognize the possibility of the other view. Any doubt that might be prompted by the arguments for that other reading should, however, be resolved against it under the rule, repeatedly affirmed, that "where 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); see also United States v. Jin Fuey Moy, 241 U.S. 394, 401 (1916). It is "out of respect for Congress, which we assume legislates in the light of constitutional limitations," Rust v. Sullivan, 500 U.S. 173, 191 (1991), that we adhere to this 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); see also United States v. X-Citement Video, Inc., 513 U.S. 64, 78 (1994).

As the Government would have us construe it, the statute would be open to constitutional doubt in light of a series of cases over the past quarter century, dealing with due process and the guarantee of trial by jury. The first of these, Mullaney v. Wilbur, 421 U.S. 684 (1975), reviewed a Maine murder statute providing that the element of malice (in the sense of want of provocation, Patterson v. New York, 432 U.S. 197, 215 (1977)) would be presumed upon proof of intent to kill resulting in death, subject to a defendant's right of rebuttal that he had acted on provocation in the heat of passion, which would reduce the offense to manslaughter. Mullaney, supra, at 686, and n. 3. The challenge was that the presumption subject to rebuttal relieved the State of its due process burden to prove every element of the crime beyond a reasonable doubt, as explained in In re Winship, 397 U.S. 358, 364 (1970). The State replied that the challenge was merely formalistic, that the State's law in effect established a generic crime of felonious homicide, Mullaney, supra, at 688, 696 697, on which view the fact subject to presumption and rebuttal would have gone simply to sentence, and Winship would not have been controlling. But the Court declined to accord the State this license to recharacterize the issue, in part because the State's reading left its statute at odds both with the centuries-old common law recognition of malice as the fact distinguishing murder from manslaughter and with the widely held modern view that heat of passion, once raised by the evidence, was a subject of the State's burden, 421 U.S., at 692 696, and in part because an unlimited choice over characterizing a stated fact as an element would leave the State substantially free to manipulate its way out of Winship, 421 U. S., at 698.

Two Terms later, in Patterson v. New York, supra, the Court ruled on a Winship challenge to a scheme defining murder as causing death with intent, subject to an affirmative defense of extreme emotional disturbance for which there was a reasonable explanation. 432 U.S., at 205 206. Unlike Maine's law, New York's raised no presumption of malice; malice was omitted from the elements of murder. Patterson contended that because the presence or absence of an extreme emotional disturbance affected the severity of sentence, Winship and Mullaney required the State to prove the absence of that fact beyond a reasonable doubt. We rejected this argument and "decline[d] to adopt as a constitutional imperative . . . that a State must disprove beyond a reasonable doubt every fact constituting any and all affirmative defenses related to the culpability of an accused." 432 U.S., at 210. We identified the use of a presumption to establish an essential ingredient of the offense as the curse of the Maine law, because the "shifting of the burden of persuasion with respect to a fact which the State deems so important that it must be either proved or presumed is impermissible under the Due Process Clause." Id., at 215. With one caveat, therefore, Patterson left the States free to choose the elements that define their crimes, without any impediment from Winship. The caveat was a stated recognition of some limit upon state authority to reallocate the traditional burden of proof, 432 U.S., at 210, which in that case was easily satisfied by the fact that "at common law the burden of proving" the mitigating circumstances of severe emotional disturbance "rested on the defendant." Id., at 202; see also id., at 211; Mullaney, supra, at 693 694. While a narrow reading of this limit might have been no more than a ban on using presumptions to reduce elements to the point of being nominal, a broader reading was equally open, that the State lacked the discretion to omit "traditional" elements from the definition of crimes and instead to require the accused to disprove such elements.

These cases about allocation of burden, with their implications about the charging obligation and the requisite quantum of proof, were succeeded by McMillan v. Pennsylvania, 477 U.S. 79 (1986), in which the Winship issue rose from a provision that a judge's finding (by a preponderance) of visible possession of a firearm would require a mandatory minimum sentence for certain felonies, but a minimum that fell within the sentencing ranges otherwise prescribed. Although the Court rejected the petitioner's claim insofar as it would have required a finding beyond a reasonable doubt of any fact upon which a mandatory minimum sentence depended (and rejected certain subsidiary arguments as well), it did observe that the result might have been different if proof of visible possession had exposed a defendant to a sentence beyond the maximum that the statute otherwise set without reference to that fact. 477 U.S., at 88.

McMillan is notable not only for acknowledging the question of due process requirements for factfinding that raises a sentencing range, but also for disposing of a claim that the Pennsylvania law violated the Sixth Amendment right to jury trial as well. The petitioner's basic argument was for a right to jury determination of all "ultimate facts concerning the offense committed," id., at 93, and although the Court disposed of this by reference back to its due process discussion, that discussion had broached the potential constitutional significance of factfinding that raised the sentencing ceiling.

McMillan, then, recognizes a question under both the Due Process Clause of the Fourteenth Amendment and the jury guarantee of the Sixth: when a jury determination has not been waived, may judicial factfinding by a preponderance support the application of a provision that increases the potential severity of the penalty for a variant of a given crime? The seriousness of the due process issue is evident from Mullaney's insistence that a State cannot manipulate its way out of Winship, and from Patterson's recognition of a limit on state authority to reallocate traditional burdens of proof; the substantiality of the jury claim is evident from the practical implications of assuming Sixth Amendment indifference to treating a fact that sets the sentencing range as a sentencing factor, not an element.6

The terms of the carjacking statute illustrate very well what is at stake. If serious bodily injury were merely a sentencing factor under §2119(2) (increasing the authorized penalty by two thirds, to 25 years), then death would presumably be nothing more than a sentencing factor under subsection (3) (increasing the penalty range to life). If a potential penalty might rise from 15 years to life on a nonjury determination, the jury's role would correspondingly shrink from the significance usually carried by determinations of guilt to the relative importance of low-level gatekeeping: in some cases, a jury finding of fact necessary for a maximum 15-year sentence would merely open the door to a judicial finding sufficient for life imprisonment. It is therefore no trivial question to ask whether recognizing an unlimited legislative power to authorize determinations setting ultimate sentencing limits without a jury would invite erosion of the jury's function to a point against which a line must necessarily be drawn.

The question might well be less serious than the constitutional doubt rule requires if the history bearing on the Framers' understanding of the Sixth Amendment principle demonstrated an accepted tolerance for exclusively judicial factfinding to peg penalty limits. But such is not the history. To be sure, the scholarship of which we are aware does not show that a question exactly like this one was ever raised and resolved in the period before the Framing. On the other hand, several studies demonstrate that on a general level the tension between jury powers and powers exclusively judicial would likely have been very much to the fore in the Framers' conception of the jury right.

The fact that we point to no statutes of the earlier time exemplifying the distinction between elements and facts that elevate sentencing ranges is unsurprising, given the breadth of judicial discretion over fines and corporal punishment in less important, misdemeanor, cases, see, e.g., J. Baker, Introduction to English Legal History 584 (3d ed. 1990); 4 W. Blackstone, Commentaries on the Laws of England 372 (1769) (hereinafter Blackstone); Preyer, Penal Measures in the American Colonies: An Overview, 26 Am. J. Legal Hist. 326, 350 (1982), and the norm of fixed sentences in cases of felony, see Langbein, The English Criminal Trial Jury on the Eve of the French Revolution, in The Trial Jury in England, France, Germany 1700 1900, pp. 36 37 (A. Schioppa ed. 1987); 4 Blackstone 238 239; A. Scott, Criminal Law in Colonial Virginia 27 28, 103 106 (1930).

Even in this system, however, competition developed between judge and jury over the real significance of their respective roles. The potential or inevitable severity of sentences was indirectly checked by juries' assertions of a mitigating power when the circumstances of a prosecution pointed to political abuse of the criminal process or endowed a criminal conviction with particularly sanguinary consequences. This power to thwart Parliament and Crown took the form not only of flat-out acquittals in the face of guilt but of what today we would call verdicts of guilty to lesser included offenses, manifestations of what Blackstone described as "pious perjury" on the jurors' part. 4 Blackstone 238 239.7

Countervailing measures to diminish the juries' power were naturally forthcoming, with ensuing responses both in the mother country and in the colonies that validate, though they do not answer, the question that the Government's position here would raise. One such move on the government's side was a parliamentary practice of barring the right to jury trial when defining new, statutory offenses. See, e.g., Frankfurter & Corcoran, Petty Federal Offenses and the Constitutional Guaranty of Trial by Jury, 39 Harv. L. Rev. 917, 925 930 (1926); 4 Blackstone 277 279. This practice extended to violations of the Stamp Act and recurred in statutes regulating imperial trade, see C. Ubbelohde, Vice-Admiralty Courts and the American Revolution 16 21, 74 80 (1960); Wroth, The Massachusetts Vice Admiralty Court, in Law and Authority in Colonial America 32, 50 (G. Billias ed. 1965), and was one of the occasions for the protest in the Declaration of Independence against deprivation of the benefit of jury trial, see P. Maier, American Scripture 118 (1997). But even before the Declaration, a less revolutionary voice than the Continental Congress had protested against the legislative practice, in words widely read in America. The use of nonjury proceedings had "of late been so far extended," Blackstone warned in the 1760's, "as, if a check be not timely given, to threaten the disuse of our admirable and truly English trial by jury." 4 Blackstone 278. Identifying trial by jury as "the grand bulwark" of English liberties, Blackstone contended that other liberties would remain secure only "so long as this palladium remains sacred and inviolate, not only from all open attacks, (which none will be so hardy as to make) but also from all secret machinations, which may sap and undermine it; by introducing new and arbitrary methods of trial, by justices of the peace, commissioners of the revenue, and courts of conscience. And however convenient these may appear at first, (as doubtless all arbitrary powers, well executed, are the most convenient), yet let it be again remembered, that delays, and little inconveniences in the forms of justice, are the price that all free nations must pay for their liberty in more substantial matters." 4 Blackstone 342 344.

A second response to the juries' power to control outcomes occurred in attempts to confine jury determinations in libel cases to findings of fact, leaving it to the judges to apply the law and, thus, to limit the opportunities for juror nullification. Ultimately, of course, the attempt failed, the juries' victory being embodied in Fox's Libel Act in Britain, see generally T. Green, Verdict According to Conscience 318 355 (1985), and exemplified in John Peter Zenger's acquittal in the Colonies, see, e.g., J. Rakove, Original Meanings 300 302 (1996). It is significant here not merely that the denouement of the restrictive efforts left the juries in control, but that the focus of those efforts was principally the juries' control over the ultimate verdict, applying law to fact (or "finding" the law, see, e.g., id., at 301), and not the factfinding role itself.8 There was apparently some accepted understanding at the time that the finding of facts was simply too sacred a jury prerogative to be trifled with in prosecution for such a significant and traditional offense in the common law courts.9 That this history had to be in the minds of the Framers is beyond cavil. According to one authority, the leading account of Zenger's trial was, with one possible exception "the most widely known source of libertarian thought in England and America during the eighteenth century." L. Levy, Freedom of Speech and Press in Early American History 133 (1963). It is just as much beyond question that Americans of the period perfectly well understood the lesson that the jury right could be lost not only by gross denial, but by erosion. See supra, at 17-20. One contributor to the ratification debates, for example, commenting on the jury trial guarantee in Art. III, §2, echoed Blackstone in warning of the need "to guard with the most jealous circumspection against the introduction of new, and arbitrary methods of trial, which, under a variety of plausible pretenses, may in time, imperceptibly undermine this best preservative of LIBERTY." A [New Hampshire] Farmer, No. 3, June 6, 1788, quoted in The Complete Bill of Rights 477 (N. Cogan ed. 1997).

In sum, there is reason to suppose that in the present circumstances, however peculiar their details to our time and place, the relative diminution of the jury's significance would merit Sixth Amendment concern. It is not, of course, that anyone today would claim that every fact with a bearing on sentencing must be found by a jury; we have resolved that general issue and have no intention of questioning its resolution. The point is simply that diminishment of the jury's significance by removing control over facts determining a statutory sentencing range would resonate with the claims of earlier controversies, to raise a genuine Sixth Amendment issue not yet settled.

Our position that the Sixth Amendment and due process issues are by no means by the boards calls for a word about several cases that followed McMillan. Almendarez-Torres v. United States, 523 U.S. 224 (1998), decided last Term, stands for the proposition that not every fact expanding a penalty range must be stated in a felony indictment, the precise holding being that recidivism increasing the maximum penalty need not be so charged. But the case is not dispositive of the question here, not merely because we are concerned with the Sixth Amendment right to jury trial and not alone the rights to indictment and notice as claimed by Almendarez-Torres, but because the holding last Term rested in substantial part on the tradition of regarding recidivism as a sentencing factor, not as an element to be set out in the indictment. The Court's repeated emphasis on the distinctive significance of recidivism leaves no question that the Court regarded that fact as potentially distinguishable for constitutional purposes from other facts that might extend the range of possible sentencing. See id., at 230 ("At the outset, we note that the relevant statutory subject matter is recidivism"); ibid. ("With recidivism as the subject matter in mind, we turn to the statute's language"); id., at 243 ("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"); id., at 245 (distinguishing McMillan "in light of the particular sentencing factor at issue in this case recidivism"). One basis for that possible constitutional distinctiveness is not hard to see: unlike virtually any other consideration used to enlarge the possible penalty for an offense, and certainly unlike the factor before us in this case, a prior conviction must itself have been established through procedures satisfying the fair notice, reasonable doubt, and jury trial guarantees. Almendarez-Torres cannot, then, be read to resolve the due process and Sixth Amendment questions implicated by reading the carjacking statute as the Government urges.10

Nor is the question resolved by a series of three cases dealing with factfinding in capital sentencing. The first of these, Spaziano v. Florida, 468 U.S. 447 (1984), contains no discussion of the sort of factfinding before us in this case. It addressed the argument that capital sentencing must be a jury task and rejected that position on the ground that capital sentencing is like sentencing in other cases, being a choice of the appropriate disposition, as against an alternative or a range of alternatives. Id., at 459.

Spaziano was followed in a few years by Hildwin v. Florida, 490 U.S. 638 (1989) (per curiam), holding that the determination of death-qualifying aggravating facts could be entrusted to a judge, following a verdict of guilty of murder and a jury recommendation of death, without violating the Sixth Amendment's jury clause. Although citing Spaziano as authority, 490 U.S., at 639 640, Hildwin was the first case to deal expressly with factfinding necessary to authorize imposition of the more severe of alternative sentences, and thus arguably comparable to factfinding necessary to expand the sentencing range available on conviction of a lesser crime than murder. Even if we were satisfied that the analogy was sound, Hildwin could not drive the answer to the Sixth Amendment question raised by the Government's position here. In Hildwin, a jury made a sentencing recommendation of death, thus necessarily engaging in the factfinding required for imposition of a higher sentence, that is, the determination that at least one aggravating factor had been proved. Hildwin, therefore, can hardly be read as resolving the issue discussed here, as the reasoning in Walton v. Arizona, 497 U.S. 639 (1990), confirms.

Walton dealt with an argument only slightly less expansive than the one in Spaziano, that every finding underlying a sentencing determination must be made by a jury. Although the Court's rejection of that position cited Hildwin, it characterized the nature of capital sentencing by quoting from Poland v. Arizona, 476 U.S. 147, 156 (1986). See 497 U.S., at 648. There, the Court described statutory specifications of aggravating circumstances in capital sentencing as "standards to guide the choice between the alternative verdicts of death and life imprisonment." Ibid. (quoting Poland, supra, at 156 (internal quotation marks omitted)). The Court thus characterized the finding of aggravating facts falling within the traditional scope of capital sentencing as a choice between a greater and a lesser penalty, not as a process of raising the ceiling of the sentencing range available. We are frank to say that we emphasize this careful reading of Walton's rationale because the question implicated by the Government's position on the meaning of §2119(2) is too significant to be decided without being squarely faced.

In sum, the Government's view would raise serious constitutional questions on which precedent is not dispositive. Any doubt on the issue of statutory construction is hence to be resolved in favor of avoiding those questions.11 This is done by construing §2119 as establishing three separate offenses by the specification of distinct elements, each of which must be charged by indictment, proven beyond a reasonable doubt, and submitted to a jury for its verdict. The judgment of the Court of Appeals is accordingly reversed, and the case is remanded for proceedings consistent with this opinion.

It is so ordered.

1. Congress amended the statute in 1994 and 1996. In the Violent Crime Control and Law Enforcement Act of 1994, it deleted the phrase in the first paragraph concerning firearm possession and replaced it with the phrase, "with the intent to cause death or serious bodily harm." §60003(a)(14), 108 Stat. 1970. It also made death a possible punishment for offenses committed under subsection (3). Ibid. In the Carjacking Correction Act of 1996, Congress specified that the term "serious bodily injury" in subsection (2) includes certain sexual assaults. §2, 110 Stat. 3020.

2. The Ninth Circuit vacated another portion of the District Court's sentencing decision and remanded. United States v. Oliver, 60 F.3d 547, 555 556 (1995). On remand, the District Court reduced petitioner's carjacking sentence to 20 years and his total sentence to 25 years, and the Court of Appeals affirmed. App. 41 43; 116 F.3d 1487 (1997).

3. See n. 4, infra.

4. Legislative history identifies three such models. See H. R. Rep. No. 102 851, pt. 1, p. 17 (1992) ("The definition of the offense tracks the language used in other federal robbery statutes (18 U.S.C. § 2111 2113, 2118)"). One of them, 18 U.S.C. § 2111 (robbery in areas of federal maritime or territorial jurisdiction), lacks aggravated forms of the offense altogether, and thus is not on point here.

5. The dissent, in passing, questions our view that §2118(a) makes the causing of significant bodily injury an element of the offense defined by that section, see post, at 9, but it offers no reason to doubt our reading. Given that §2118(a) establishes only one maximum punishment, and that it makes eligibility for such punishment contingent on the establishment of at least one of three facts, one of which is the causing of death or significant bodily injury, we think our reading is the only sensible one.

6. The dissent repeatedly chides us for failing to state precisely enough the principle animating our view that the carjacking statute, as construed by the Government, may violate the Constitution. See post, at 1, 14, 17. The preceding paragraph in the text expresses that principle plainly enough, and we re-state it here: under the Due Process Clause of the Fifth Amendment and the notice and jury trial guarantees of the Sixth Amendment, any fact (other than prior conviction) that increases the maximum penalty for a crime must be charged in an indictment, submitted to a jury, and proven beyond a reasonable doubt. Because our prior cases suggest rather than establish this principle, our concern about the Government's reading of the statute rises only to the level of doubt, not certainty. Contrary to the dissent's suggestion, the constitutional proposition that drives our concern in no way "call[s] into question the principle that the definition of the elements of a criminal offense is entrusted to the legislature." Post, at 16 (internal quotation marks omitted). The constitutional guarantees that give rise to our concern in no way restrict the ability of legislatures to identify the conduct they wish characterize as criminal or to define the facts whose proof is essential to the establishment of criminal liability. The constitutional safeguards that figure in our analysis concern not the identity of the elements defining criminal liability but only the required procedures for finding the facts that determine the maximum permissible punishment; these are the safeguards going to the formality of notice, the identity of the factfinder, and the burden of proof.

7. For English practice, see, e.g., Langbein, Shaping the Eighteenth-Century Criminal Trial, 50 U. Chi. L. Rev. 1, 22, 52 54 (1983); Green, The English Criminal Trial Jury, in The Trial Jury in England, France, Germany 1700 1900, 41, 48 49 (A. Schioppa ed. 1987). For Colonial American practice, see, e.g., J. Goebell & T. Naughton, Law Enforcement in Colonial New York 673 674 (1944); State v. Bennet, 3 Brevard 515 (S. C. 1815).

8. The principle that the jury were the judges of fact and the judges the deciders of law was stated as an established principle as early as 1628 by Coke. See 1 E. Coke, Institutes of the Laws of England 155b (1628) ("ad questionem facti non respondent judices; ad questionem juris non respondent juratores"). See also Langbein, The English Criminal Trial Jury on the Eve of the French Revolution, in The Trial Jury in England, France, Germany, supra, at 34, n. 60. Even the traditional, jury-restrictive view of libel law recognized the jury's authority over matters of fact. See, e.g., King v. Francklin, 17 How. St. Tr. 626, 672 (K.B. 1731) ("These [publication and the words having the meaning ascribed to them] are the two matters of fact that come under your consideration; and of which you are proper judges. But then there is a third thing, to wit, Whether these defamatory expressions amount to a libel or not? This does not belong to the office of the jury, but to the office of the Court; because it is a matter of law, and not of fact; and of which the Court are the only proper judges"). Thus most participants in the struggle over jury autonomy in seditious libel cases viewed the debate as concerned with the extent of the jury's law-finding power, not its unquestioned role as the determiner of factual issues. See T. Green, Verdict According to Conscience 318 319 (1985). Some influential jurists suggested that it might also be seen as a struggle over the jury's right to find a particular fact, namely, the required criminal intent. See 10 W. Holdsworth, History of English Law 680 683 (1938).

9. See 4 Blackstone 354 (jurors could choose to stop at special verdicts if they wished).

10. The dissent insists that Almendarez-Torres "controls the question before us," post, at 13, 18, but in substantiating that assertion, it tellingly relies more heavily on the claims of the Almendarez-Torres dissenters than on the statements of the Almendarez-Torres majority. Neither source bears out the current dissent's conclusion. If, as the dissenters in this case suggest, Almendarez-Torres did not turn on the particular "sentencing factor at issue" there, 523 U.S., at 243, but instead stood for the broad proposition that any fact increasing the maximum permissible punishment may be determined by a judge by a preponderance, it is a mystery why the Almendarez-Torres majority engaged in so much discussion of recidivism, or why, at the crux of its constitutional discussion, it turned first to discuss the "tradition" of recidivism's treatment as a sentencing factor, ibid., or why it never announced the unqualified holding that today's dissenters claim to find in it. Admittedly, as the dissent here notes, the dissenters in Almendarez-Torres criticized the majority for what they considered the majority's unsupportable restraint in restricting their holding to recidivism. But that very criticism would have lacked its target if the Almendarez-Torres majority had not so doggedly refrained from endorsing the general principle the dissent in this case now attributes to them. The majority and the dissenters in Almendarez-Torres disagreed over the legitimacy of the Court's decision to restrict its holding to recidivism, but both sides agreed that the Court had done just that.

11. In tones of alarm, the dissent suggests, see post, at 2, 17, that our decision will unsettle the efforts of many States to bring greater consistency to their sentencing practices through provisions for determinate sentences and statutorily or administratively established guidelines governing sentencing decisions. The dissent's concern is misplaced for several reasons. Most immediately, our decision today does not announce any new principle of constitutional law, but merely interprets a particular federal statute in light of a set of constitutional concerns that have emerged through a series of our decisions over the past quarter century. But even if we assume that the question we raise will someday be followed by the answer the dissenters seem to fear, that answer would in no way hinder the States (or the National Government) from choosing to pursue policies aimed at rationalizing sentencing practices. If the constitutional concern we have expressed should lead to a rule requiring jury determination of facts that raise a sentencing ceiling, that rule would in no way constrain legislative authority to identify the facts relevant to punishment or to establish fixed penalties. The constitutional guarantees that prompt our interpretation bear solely on the procedures by which the facts that raise the possible penalty are to be found, that is, what notice must be given, who must find the facts, and what burden must be satisfied to demonstrate them. Finally, while we disagree with the dissent's dire prediction about the effect of our decision on the States' ability to choose certain sentencing policies, it should go without saying that, if such policies conflict with safeguards enshrined in the Constitution for the protection of the accused, those policies have to yield to the constitutional guarantees. See, e.g., Burch v. Louisiana, 441 U.S. 130, 139 (1979) (Non-unanimous verdicts by six-person criminal juries "sufficiently threaten[] the constitutional principles [animating the jury trial guarantee] that any countervailing interest of the State should yield"); cf. Almeida-Sanchez v. United States, 413 U.S. 266, 273 (1973) ("The needs of law enforcement stand in constant tension with the Constitution's protections of the individual against certain exercises of official power. It is precisely the predictability of these pressures that counsels a resolute loyalty to constitutional safeguards").

ON WRIT OF CERTIORARI TO THE UNITED STATES COURT OF APPEALS FOR THE NINTH CIRCUIT

[March 24, 1999]

Justice Stevens, concurring.

Like Justice Scalia, see post, at 1, I am convinced that it is unconstitutional for a legislature to remove from the jury the assessment of facts that increase the prescribed range of penalties to which a criminal defendant is exposed. It is equally clear that such facts must be established by proof beyond a reasonable doubt. That is the essence of the Court's holdings in In re Winship, 397 U.S. 358 (1970), Mullaney v. Wilbur, 421 U.S. 684 (1975), and Patterson v. New York, 432 U.S. 197 (1977). To permit anything less "with respect to a fact which the State deems so important that it must either be proved or presumed is impermissible under the Due Process Clause." Id., at 215. This principle was firmly embedded in our jurisprudence through centuries of common law decisions. See, e.g., Winship, 397 U.S., at 361 364; Duncan v. Louisiana, 391 U.S. 145, 151 156 (1968). Indeed, in my view, a proper understanding of this principle encompasses facts that increase the minimum as well as the maximum permissible sentence, and also facts that must be established before a defendant may be put to death. If McMillan v. Pennsylvania, 477 U.S. 79 (1986), and Part II of the Court's opinion in Walton v. Arizona, 497 U.S. 639, 647 649 (1990), departed from that principle, as I think they did, see McMillan, 477 U.S., at 95 104 (Stevens, J., dissenting) and Walton, 497 U.S., at 709 714 (Stevens, J., dissenting), they should be reconsidered in due course. It is not, however, necessary to do so in order to join the Court's opinion today, which I do.

ON WRIT OF CERTIORARI TO THE UNITED STATES COURT OF APPEALS FOR THE NINTH CIRCUIT

[March 24, 1999]

Justice Scalia, concurring.

In dissenting in Almendarez-Torres v. United States, 118 S. Ct. 1219 (1998), I suggested the possibility, and in dissenting in Monge v. California, 118 S. Ct. 2246, 2255-2257 (1998), I set forth as my considered view, that it is unconstitutional to remove from the jury the assessment of facts that alter the congressionally prescribed range of penalties to which a criminal defendant is exposed. Because I think it necessary to resolve all ambiguities in criminal statutes in such fashion as to avoid violation of this constitutional principle, I join the opinion of the Court.

ON WRIT OF CERTIORARI TO THE UNITED STATES COURT OF APPEALS FOR THE NINTH CIRCUIT

[March 24, 1999]

Justice Kennedy, with whom The Chief Justice, Justice O'Connor, and Justice Breyer join, dissenting.

The question presented is whether the federal carjacking statute, prohibiting the taking of a motor vehicle from the person or presence of another by force and violence or by intimidation, contains in the first paragraph a complete definition of the offense, with all of the elements of the crime Congress intended to codify. 18 U.S.C. § 2119. In my view, shared by every Court of Appeals to have addressed the issue, it does. The Court adopts a contrary, strained reading according to which the single statutory section prohibits three distinct offenses.

Had it involved simply a question of statutory interpretation, the majority opinion would not have been cause for much concern. Questions of statutory interpretation can be close but nonetheless routine. That should have been so in today's case. The Court, however, is unwilling to rest its opinion on textual analysis. Rather, to bolster its statutory interpretation, the Court raises the specter of " 'grave and doubtful constitutional questions,' " ante, at 12, without an adequate explanation of the origins, contours, or consequences of its constitutional concerns. The Court's reliance on the so-called constitutional doubt rule is inconsistent with usual principles of stare decisis and contradicts the approach followed just last Term in Almendarez-Torres v. United States, 523 U.S. 224 (1998). Our precedents admit of no real doubt regarding the power of Congress to establish serious bodily injury and death as sentencing factors rather than offense elements, as we made clear in Almedarez-Torres. Departing from this recent authority, the Court's sweeping constitutional discussion casts doubt on sentencing practices and assumptions followed not only in the federal system but also in many States. Thus, among other unsettling consequences, today's decision intrudes upon legitimate and vital state interests, upsetting the proper federal balance. I dissent from this unfortunate and unnecessary result.

Before it departs on its troubling constitutional discussion, the Court analyzes the text of §2119. This portion of the Court's opinion, it should be acknowledged, is careful and comprehensive. In my submission, however, the analysis suggests the presence of more interpretative ambiguity than in fact exists and reaches the wrong result. Like the Court, I begin with the textual question.

* Criminal laws proscribe certain conduct and specify punishment for transgressions. A person commits a crime when his or her conduct violates the essential parts of the defined offense, which we refer to as its elements. As a general rule, each element of a charged crime must be set forth in an indictment, Hamling v. United States, 418 U.S. 87, 117 (1974), and established by the government by proof beyond a reasonable doubt, In re Winship, 397 U.S. 358, 364 (1970), as determined by a jury, assuming the jury right is invoked, Sullivan v. Louisiana, 508 U.S. 275, 277 278 (1993); Almendarez-Torres v. United States, 523 U.S., at 239. The same rigorous requirements do not apply with respect to "factors relevant only to the sentencing of an offender found guilty of the charged crime." Id., at 228; see also McMillan v. Pennsylvania, 477 U.S. 79, 93 (1986). "[T]he question of which factors are which is normally a matter for Congress." Almendarez-Torres v. United States, 523 U.S., at 228.

In determining whether clauses (1) (3) of §2119 set forth sentencing factors or define distinct criminal offenses, our task is to "look to the statute before us and ask what Congress intended." Ibid. The statute is as follows:

"Whoever, possessing a firearm as defined in section 921 of this title, takes a motor vehicle that has been transported, shipped, or received in interstate or foreign commerce from the person or presence of another by force and violence or by intimidation, or attempts to do so, shall

(1) be fined under this title or imprisoned not more than 15 years, or both,

(2) if serious bodily injury (as defined in section 1365 of this title) results, be fined under this title or imprisoned not more than 25 years, or both, and

(3) if death results, be fined under this title or imprisoned for any number of years up to life, or both." 18 U.S.C. § 2119 (1988 ed., Supp. V).

As the Court is quite fair to acknowledge, the first reading or initial look of the statute suggests that clauses (1) (3) are sentencing provisions. Ante, at 4. In my view, this conclusion survives further and meticulous examination.

Section 2119 begins by setting forth in its initial paragraph elements typical of a robbery-type offense. For all ordinary purposes, this is a complete crime. If, for instance, there were only a single punishment, as provided in clause (1), I think there could be no complaint with jury instructions drawn from the first paragraph of §2119, without reference to the punishment set forth in clause (1). The design of the statute yields the conclusion that the following numbered provisions do not convert each of the clauses into additional elements. These are punishment provisions directed to the sentencing judge alone. To be sure, the drafting could have been more clear, and my proffered interpretation would have been better implemented, if the word "shall" at the end of the first paragraph had been followed by a verb form (e.g., "be punished") and a period. Even as written, though, the statute sets forth a complete crime in the first paragraph. It is difficult to see why Congress would double back and insert additional elements for the jury's consideration in clauses (2) and (3). The more likely explanation is that Congress set forth the offense first and the punishment second, without intending to combine the two.

Unlike the Court, I am unpersuaded by other factors that this common-sense reading is at odds with congressional intent. As to the substance of clauses (2) and (3), the harm from a crime including whether the crime, after its commission, results in the serious bodily injury or death of a victim has long been deemed relevant for sentencing purposes. Like recidivism, it is "as typical a sentencing factor as one might imagine," Almendarez-Torres v. United States, supra, at 230, a point the Court cannot dispute. To fix punishment based on the harm resulting from a crime has been the settled practice under traditional, discretionary sentencing regimes. See, e.g., U.S. Dept. of Justice, W. Rhodes & C. Conly, Analysis of Federal Sentencing X 13, XV 11 (Federal Justice Research Program Rep. No. FJRP 81/004, 1981) (under preguidelines practice, with respect to a variety of crimes, the amount of harm threatened or done to victims made a significant difference in the length of sentence). Even if we confine our attention to codified law, however, examples abound to prove the point. Other federal statutes, as the Court notes, treat serious bodily injury as a sentencing factor. Ante, at 7. As for state law, common practice discloses widespread reliance on victim-impact factors for sentencing purposes. See, e.g., Alaska Stat. Ann. §12.55.125(c)(2) (1998) ("physical injury"); Ariz. Rev. Stat. Ann. §13.702(C) (Supp. 1998 1999) ("serious physical injury"); Colo. Rev. Stat. §18 1 105(9)(f ) (1997) ("serious bodily injury"); Fla. Stat. Ann. §921.0016(3)(l) (Supp. 1999) ("permanent physical injury"); Haw. Rev. Stat. §706-662(5) (Supp. 1996) ("serious or substantial bodily injury" upon certain victims); Ill. Comp. Stat., ch. 730, §5/5 5 3.2(a) (1997) ("serious harm"); La. Code. Crim. Proc. Ann., Art. 894.1(B)(5) (West 1997) ("risk of death or great bodily harm to more than one person"); N. J. Stat. Ann. §2C:44 1(a)(2) (West 1995) ("gravity and seriousness of harm inflicted on the victim"); N. C. Gen. Stat. §15A 1340.16(d)(19) (1997) ("[t]he serious injury inflicted upon the victim is permanent and debilitating"); Ohio Rev. Code. Ann. §2929.12(B)(2) (1997) ("serious physical . . . harm"); Ore. Admin. Rules §213 008 0002(1)(b)(I) (1997) ("permanent injury"); Tenn. Code Ann. §40 35 114(12) (1997) ("death or serious bodily injury"); Utah Code of Judicial Admin., App. D, Form 2 (1998) ("substantial bodily injury"). Given this widespread understanding, there is nothing surprising or anomalous in the conclusion that Congress chose to treat serious bodily injury and resulting death as sentencing factors in §2119.

In addition, the plain reading of §2119 is reinforced by common patterns of statutory drafting. For example, in one established statutory model, Congress defines the elements of an offense in an initial paragraph ending with the phrase "shall be punished as provided in" a separate subsection. The subsection provides for graded sentencing ranges, predicated upon specific findings (such as serious bodily injury or death). See, e.g., 8 U.S.C. § 1324(a)(1). Section 2119 follows a similar logic. It is true that clauses (1) (3) are not separated into a separate subsection, thus giving rise to the textual problem we must resolve. Congress does not always separate sentencing factors into separate subsections, however. See, e.g., 18 U.S.C. § 1347 (1994 ed., Supp. III) (health-care fraud; enhanced penalties if the violation "results in serious bodily injury" or "results in death"). As with statutes like §1324, the structure of §2119 suggests a design which defines the offense first and the punishment afterward.

In addition, there is some significance in the use of the active voice in the main paragraph and the passive voice in clauses (2) and (3) of §2119. In the more common practice, criminal statutes use the active voice to define prohibited conduct. See, e.g., 18 U.S.C. § 1116 (1994 ed., Supp. III) ("[w]hoever kills or attempts to kill"); 18 U.S.C. § 2114 ("assaults," "robs or attempts to rob," "receives, possesses, conceals, or disposes"); Tex. Penal Code Ann. §29.03(a)(1), (2) (1994) (aggravated robbery; "causes serious bodily injury," or "uses or exhibits a deadly weapon"); cf. 18 U.S.C. § 248(b) (setting forth, as sentencing factors, "if bodily injury results," and "if death results"); United States Sentencing Commission, Guidelines Manual §2B3.1(b)(3) (Nov. 1998) (robbery guideline; "[i]f any victim sustained bodily injury").

These drafting conventions are not absolute rules. Congress uses active language in phrasing sentencing factors in some instances. See, e.g., 18 U.S.C. § 2262(b)(3) (1994 ed., Supp. III) ("if serious bodily injury to the victim results or if the offender uses a dangerous weapon during the offense"). Nevertheless, the more customary drafting conventions support, rather than contradict, the interpretation that §2119 sets forth but one offense.

The Court offers specific arguments regarding these background considerations, each deserving of consideration and response.

First, as its principal argument, the Court cites the three federal robbery statutes on which (according to the legislative history) §2119 was modeled. As the Court acknowledges, however, one of those statutes, 18 U.S.C. § 2111 does not refer to "serious bodily injury" or "death" "result[ing]" at all. Because of the omission, the Court deems this statute irrelevant for our purposes. Yet the Committee Report cited by the Court states that " '[t]he definition of the offense' " in §2119 " 'tracks the language used in other federal robbery statutes' " including §2111. Ante, at 8, n. 4 (quoting H. R. Rep. No. 102 851, pt. 1, p. 17 (1992)). The definition of the offense in §2119 includes "tak[ing]" or "attempt[ing]" to take a motor vehicle, "from the person or presence of another," "by force and violence or by intimidation." This is altogether consistent with the definition of the offense in §2111, which provides in part that "[w]hoever by force and violence, or by intimidation, takes or attempts to take from the person or presence of another" something of value "shall be imprisoned." Of course §§2111 and 2119 each include at least one element the other does not (e.g., "within the special maritime and territorial jurisdiction of the United States" in the former, "transported, shipped, or received in interstate or foreign commerce" in the latter). Those elements, however, are included in unambiguous fashion in the offense-defining part of the statutes. With respect to the debatable interpretive question whether serious bodily injury and death are part of the carjacking offense the circumstance that the definition of the offense in §2119 is based on §2111 and that §2111 does not include these elements, suggests §2119 does not include the elements either.

Passing over §2111, the Court suggests §§2113 and 2118 support its reading of §2119. I disagree. Section 2113, captioned "Bank robbery and incidental crimes," consists of eight subsections. The last three are definitional and irrelevant to the question at hand. The first subsection, subsection (a), proscribes the crime of bank robbery in language that tracks the definition of the offense in §2119, i.e., "tak[ing], or attempt[ing] to take," something of value "from the person or presence of another," "by force and violence, or by intimidation." Subsection (b) proceeds to define the offense of bank larceny and is cast in different terms as is natural in light of the different conduct proscribed. Subsections (d) and (e) of §2113, the two subsections relied upon by the Court, provide as follows:

"(d) Whoever, in committing, or in attempting to commit, any offense defined in subsections (a) and (b) of this section, assaults any person, or puts in jeopardy the life of any person by the use of a dangerous weapon or device, shall be fined under this title or imprisoned not more than twenty-five years, or both.

"(e) Whoever, in committing any offense defined in this section, or in avoiding or attempting to avoid apprehension for the commission of such offense, or in freeing himself or attempting to free himself from arrest or confinement for such offense, kills any person, or forces any person to accompany him without the consent of such person, shall be imprisoned not less than ten years, or if death results shall be punished by death or life imprisonment."

We have not held that subsections (d) and (e) set forth separate offenses. (The Court's citations to Almendarez-Torres and McMillan on this score are inapt. In neither case did we hold that §§2113(d) and (e) set forth distinct offenses.) Assuming they do, however, they fail to prove the Court's point, for two reasons. First, as a matter of structure, §2113 is divided into distinct subsections with a parallel form. Excluding the definitional provisions at the end, each of the five subsections begins with the word "[w]hoever," followed by specified conduct. Given that some of these subsections (e.g., subsections (a) and (b)) set forth distinct offenses, it is fair to presume their like structured neighbors do so as well. One finds no analogous subsections in §2119 with which clauses (1) (3) can be matched. On the contrary, clause (1) plainly fails to introduce anything that could be construed as an offense element, making it all the less likely that offense elements are introduced in clauses (2) and (3). Second, the phrases from §2113 cited by the Court "assaults any person" and "puts in jeopardy the life of any person by use of a dangerous weapon or device" are rather different from the "serious bodily injury results" and "death results" language of §2119. The former phrases occur before, not after, the punishment-introducing clause "shall be ." They are also phrased in the active voice, placing attention on the defendant's actions, rather than their consequences. The "or if death results" phrase at the end of subsection (e) is a closer analogue to clauses (2) and (3) of §2119, but there is no reason to assume that this phrase by itself as opposed to the preceding portion of subsection (e) defines an element of an offense.

With respect to §2118, the Court asserts without citation to authority that the phrase "another person suffered significant bodily injury" in subsection (a)(3) is an element of the offense. Ante, at 8. Even assuming the Court is correct on the point, however, the differences in structure between that provision and §2119 show them not to be comparable. Clauses (1) (3) in §2119 set forth alternative sentences; but the three clauses in §2118(a) set forth alternative ways of qualifying for the only punishment provided. The more natural reading is that the drafters of §2119 took from §2118 the same thing they took from §§2111 and 2113: the language defining the basic elements of robbery. It is this language, and not other provisions, that is common to all four statutes.

In short, even indulging the Court's assumptions, the federal robbery statutes do not support the conclusion that §2119 contains three substantive offenses. Rather, all four statutes employ similar language to define the elements of a basic robbery-type offense. It is in this sense that §2119 is modeled on §§2111, 2113, and 2118.

The Court next relies on the consumer product-tampering statute, 18 U.S.C. § 1365(a), as support for its reading of §2119. It is indeed true, as the Court suggests, that the structure and phrasing of §1365(a) is similar to the carjacking statute. However, neither the Court nor, my research indicates, any Court of Appeals has held that §1365(a) creates multiple offenses. The only case cited for the proposition that "the Courts of Appeals treat the statute as defining basic and aggravated offenses," ante, at 6, establishes nothing of the kind. There, the Court of Appeals did no more than recite that the defendant had been charged and convicted on multiple counts of product tampering, under three subsections of §1365(a). United States v. Meling, 47 F.3d 1546, 1551 (CA9 1995). None of the issues presented turned on whether the subsections set forth additional elements.

The Court's final justification for its reading of §2119 rests on state practice. Of course, the Court cannot argue that States do not take factors like serious bodily injury into account at sentencing; as discussed above, they do. Instead, the Court says many States have created a distinct offense of aggravated robbery, requiring proof of serious bodily injury or harm. This is unremarkable. The laws reflect nothing more than common intuition that a forcible theft, all else being equal, is more blameworthy when it results in serious bodily injury or death. I have no doubt Congress was responding to this same intuition when it added clauses (2) and (3) to §2119. Recognizing the common policy concern, however, gives scant guidance on the question before us: whether Congress meant to give effect to the policy by making serious bodily injury and death elements of distinct offenses or by making them sentencing factors. I agree with the Court that these state statutes are not direct authority for the issue presented here. Ante, at 10.

The persuasive force of the Court's state-law citations is further undercut by the structural differences between those laws and §2119. Ten of the thirteen statutes cited by the Court follow the same pattern. One statutory section sets forth the elements of the basic robbery offense. Another section (captioned "Aggravated robbery" or "Robbery in the first degree") incorporates the basic robbery offense (either by explicit cross-reference or by obvious implication), adds the bodily or physical injury element (in the active voice), and then provides that the aggravated crime is subject to a higher penalty set forth elsewhere (e.g., "a class A felony"). Two of the remaining three statutes, N. Y. Penal Law §160.15 (McKinney 1988), and Ky. Rev. Stat. Ann. §515.020 (Michie 1990), deviate from this pattern in only minor respects while the third, N. H. Rev. Stat. Ann. §636:1 (1996), has a singular structure.

Had Congress wished to emulate this state practice in detail, one might have expected it to structure §2119 in a similar manner to the majority model. Cf. 18 U.S.C. § 2113(e),(d). It did not do so. This suggests to me either (i) that Congress chose a different structure than utilized by the States in order to show its intent to treat "serious bodily injury" as a sentencing factor, or (ii) that Congress simply did not concentrate on state practice in deciding whether "serious bodily injury" should be classed as an element or a sentencing factor. Neither possibility sustains the Court's interpretation of §2119.

Although the Court, in my view, errs in its reading of §2119 as a simple matter of statutory construction, of far greater concern is its constitutional discussion. In order to inject the rule of constitutional doubt into the case, the Court treats the relevant line of authorities from Winship to Almendarez-Torres as if it had been the Court's purpose to write them at odds with each other, not to produce a coherent body of case law interpreting the relevant constitutional provisions. This attempt to create instability is neither a proper use of the rule of constitutional doubt nor a persuasive reading of our precedents. We have settled more than the Court's opinion says.

In re Winship, 397 U.S. 358 (1970), made clear what has long been accepted in our criminal justice system. It is the principle that in a criminal case the government must establish guilt beyond a reasonable doubt. To implement this constitutional protection, it follows, there must be an understanding of the essential elements of the crime; and cases like this one will arise, requiring statutory analysis.

Nonetheless, the holding of the first case decided in the wake of Winship, Mullaney v. Wilbur, 421 U.S. 684 (1975), now seems straightforward. In homicide cases, Maine sought to presume malice from the fact of an intentional killing alone, subject to the defendant's right to prove he had acted in the heat of passion. This was so even though "the fact at issue the presence or absence of the heat of passion on sudden provocation has been, almost from the inception of the common law of homicide, the single most important factor in determining the degree of culpability attaching to an unlawful homicide." Id., at 696. As we later explained, Mullaney "held that a State must prove every ingredient of an offense beyond a reasonable doubt, and that it may not shift the burden of proof to the defendant by presuming that ingredient upon proof of the other elements of the offense." Patterson v. New York, 432 U.S. 197, 215 (1977).

In Patterson, the Court confronted a state rule placing on the defendant the burden of establishing extreme emotional disturbance as an affirmative defense to murder. As today's majority opinion recognizes, Patterson stands for the proposition that the State has considerable leeway in determining which factors shall be included as elements of its crimes. We determined that New York was permitted to place the burden of proving the affirmative defense on defendants because "nothing was presumed or implied against" them. Id., at 216.

In McMillan v. Pennsylvania, 477 U.S. 79 (1986), we upheld a State law requiring imposition of a mandatory minimum sentence upon the trial judge's determination that the defendant had visibly possessed a firearm during the commission of an enumerated offense. Today's majority errs, in my respectful view, by suggesting McMillan is somewhat inconsistent with Patterson. McMillan's holding follows easily from Patterson. McMillan confirmed the State's authority to treat aggravated behavior as a factor increasing the sentence, rather than as an element of the crime. The opinion made clear that we had already "rejected the claim that whenever a State links the 'severity of punishment' to 'the presence or absence of an identified fact' the State must prove that fact beyond a reasonable doubt." 477 U.S., at 84 (quoting Patterson v. New York, supra, at 214).

In today's decision, the Court chooses to rely on language from McMillan to create a doubt where there should be none. Ante, at 14-15. Yet any uncertainty on this score ought to have been put to rest by our decision last Term in Almendarez-Torres. To say otherwise, the majority must strive to limit Almendarez-Torres, just as it must struggle with Patterson and McMillan. Almendarez-Torres, however, controls the question before us.

As an initial matter, Almendarez-Torres makes clear that the constitutional doubt methodology employed by the Court today is incorrect. It teaches that the constitutional doubt canon of construction is applicable only if the statute at issue is "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." 523 U.S., at 238. For the reasons given in Part I, supra, the Court of Appeals' interpretation of §2119 is, in my view, superior to petitioner's reading. At a minimum, the question whether 8 U.S.C. § 1326(b) the statute at issue in Almendarez-Torres, set forth sentencing factors or elements of distinct offenses was a closer one than the statutory question presented here. Yet we found insufficient ambiguity to warrant application of the constitutional doubt principle there. 523 U.S., at 238. Unless we are to abandon any pretense of consistency in the application of the principle, it is incumbent on the Court to explain how it reconciles its analysis with Almendarez-Torres.

Not only is the proper construction of the statute clearer here, but there is less reason, in light of Almendarez-Torres itself, to question the constitutionality of the statute as construed by the Court of Appeals. The insubstantiality of the Court's constitutional concern is indicated by its quite summary reference to the principle of constitutional law the statute might offend. The Court puts the argument this way: "any fact (other than prior conviction) that increases the maximum penalty for a crime must be charged in an indictment, submitted to a jury, and proven beyond a reasonable doubt." Ante, at 16 n. 6. It suggests the carjacking statute violates this principle because absent a finding of serious bodily injury, a defendant may be sentenced to a maximum of 15 years' imprisonment and, absent a finding of death, he may be sentenced to a maximum of 25 years' imprisonment. A finding of serious bodily injury increases the maximum penalty for the crime of carjacking from 15 to 25 years' imprisonment and a finding of death increases the maximum to life imprisonment.

If the Court is to be taken at its word, Congress could comply with this principle by making only minor changes of phraseology that would leave the statutory scheme, for practical purposes, unchanged. Congress could leave the initial paragraph of §2119 intact, and provide that one who commits the conduct described there shall "be imprisoned for any number of years up to life." It could then add that "if the sentencing judge determines that no death resulted, one convicted under this section shall be imprisoned not more than 25 years" and "if the sentencing judge determines that no serious bodily injury resulted, one convicted under this section shall be imprisoned not more than 15 years." The practical result would be the same as the current version of §2119 (as construed by the Court of Appeals): the jury makes the requisite findings under the initial paragraph, and the court itself sentences the defendant within one of the prescribed ranges based on the judge's own determination whether serious bodily injury or death resulted.

The Court does not tell us whether this version of the statute would pass constitutional muster. If so, the Court's principle amounts to nothing more than chastising Congress for failing to use to the approved phrasing in expressing its intent as to how carjackers should be punished. No constitutional values are served by so formalistic an approach, while its constitutional costs in statutes struck down or, as today, misconstrued, are real.

If, on the other hand, a rephrased §2119 would still violate the Court's underlying constitutional principle, the Court ought to explain how it would determine which sentencing schemes cross the constitutional line. For example, a statute that sets a maximum penalty and then provides detailed sentencing criteria to be applied by a sentencing judge (along the lines of the federal Sentencing Guidelines) would be only a more detailed version of the rephrased §2119 suggested above. We are left to guess whether statutes of that sort might be in jeopardy. (Further, by its terms, Justice Scalia's view "that it is unconstitutional to remove from the jury the assessment of facts that alter the congressionally prescribed range of penalties to which a criminal defendant is exposed," ante, at 1 (concurring opinion) would call into question the validity of judge-administered mandatory minimum sentencing provisions, contrary to our holding in McMillan. Once the facts triggering application of the mandatory minimum are found by the judge, the sentencing range to which the defendant is exposed is altered.) In light of these uncertainties, today's decision raises more questions than the Court acknowledges.

In any event, the Court's constitutional doubts are not well founded. In Almendarez-Torres, we squarely rejected the petitioner's argument that "any significant increase in a statutory maximum sentence would trigger a constitutional 'elements' requirement"; as we said, the Constitution "does not impose that requirement." 523 U.S., at 247. See also Monge v. California, 524 U.S. 721, 729 (1998) ("[T]he Court has rejected an absolute rule that an enhancement constitutes an element of the offense any time that it increases the maximum sentence to which a defendant is exposed"). Indeed, the dissenters in Almendarez-Torres had no doubt on this score. 523 U.S., at 260 (opinion of Scalia, J.) (arguing 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").

The Court suggests two bases on which Almendarez-Torres is distinguishable, neither of which is persuasive. First, the Court suggests that this case is "concerned with the Sixth Amendment right to jury trial and not alone the rights to indictment and notice as claimed by Almendarez-Torres." Ante, at 21. This is not a valid basis upon which to distinguish Almendarez-Torres. The petitioner in Almendarez-Torres claimed that "the Constitution requires Congress to treat recidivism as an element of the offense" and that, as a corollary, "[t]he Government must prove that 'element' to a jury." 523 U.S. at 239.

The Court has not suggested in its previous opinions, moreover, that there is a difference, in the context relevant here, between, on the one hand, a right to a jury determination, and, on the other, a right to notice by indictment and to a determination based upon proof by the prosecution beyond a reasonable doubt. The Court offers no reason why the concept of an element of a crime should mean one thing for one inquiry and something else for another. There would be little to guide us in formulating a standard to differentiate between elements of a crime for purposes of indictment, jury trial, and proof beyond a reasonable doubt. Inviting such confusion is a curious way to safeguard the important procedural rights of criminal defendants.

Second, the Court is eager to find controlling significance in the fact that the statute at issue in Almendarez-Torres made recidivism a sentencing factor, while the sentencing factor at issue here is serious bodily injury. This is not a difference of constitutional dimension, and Almendarez-Torres does not say otherwise. It is true that our statutory analysis was informed in substantial measure by the fact that recividism is a common sentencing factor. Id., at 230. In our constitutional analysis we invoked the long history of using recidivism as a basis for increasing an offender's sentence to illustrate the novel and anomalous character of the petitioner's proposed constitutional rule i.e., that under McMillan v. Pennsylvania any factor that increases the maximum penalty for a crime must be deemed an element of the offense. We proceeded to reject that rule. Almendarez-Torres v. United States, 523 U.S., at 247. The dissenters there (like the Court today) misunderstood the import of this discussion, but they were correct in their observation that "[i]t 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 case where some other sentencing factor is at issue." Id., at 258 (opinion of Scalia, J.).

The constitutional portion of Almendarez-Torres also rejected the argument that constitutional concerns were raised by a "different 'tradition' that of courts having treated recidivism as an element of the related crime." Id., at 246. We found this argument unconvincing because "any such tradition is not uniform." Ibid. Of course, the same is true with respect to the sentencing factors at issue here. See infra, at 4 5. In sum, "there is no rational basis for making recidivism an exception." 523 U.S., at 258 (Scalia, J., dissenting) (emphasis omitted).

If the Court deems its new direction to be a justified departure from stare decisis, it does not make the case. There is no support for the view that Almendarez-Torres was based on a historical misunderstanding or misinterpretation. By the Court's own submission, its historical discussion demonstrates no more than that "the tension between jury safeguard and exclusively judicial powers" would probably and generally have informed the Framers' conception of the jury right. Ante, at 17. That must be correct, but it does not call into question the principle that " '[t]he definition of the elements of a criminal offense is entrusted to the legislature.' " Staples v. United States, 511 U.S. 600, 604 (1994) (quoting Liparota v. United States, 471 U.S. 419, 424 (1985)).

The Court's historical analysis might have some bearing on the instant case if §2119 disclosed the intent to serve the real objective of punishing (without constitutional safeguards) those who caused serious bodily harm, rather than to prevent the underlying conduct of carjacking. See Almendarez-Torres v. United States, supra, at 243, 246. No such inference or implication can be drawn from the text and statutory history of the offense here under consideration. In fact, the Court makes no attempt to argue that anything particular to the carjacking statute suggests the jury's role has been unconstitutionally diminished. The gravamen of the offense is carjacking coupled with a threat of bodily harm. The jury resolves these issues, i.e., whether a vehicle is taken "by force and violence or by intimidation." Indeed, whether serious bodily injury results can be outside of the defendant's control. As already explained, it is not in the least a novel view that after the offense is established, the extent of the harm caused is taken into account in the sentencing phase. In this respect, today's case is far easier than McMillan, where the sentencing factor was inherent in the criminal conduct itself.

The rationale of the Court's constitutional doubt holding makes it difficult to predict the full consequences of today's holding, but it is likely that it will cause disruption and uncertainty in the sentencing systems of the States. Sentencing is one of the most difficult tasks in the enforcement of the criminal law. In seeking to bring more order and consistency to the process, some States have sought to move from a system of indeterminate sentencing or a grant of vast discretion to the trial judge to a regime in which there are more uniform penalties, prescribed by the legislature. See A. Campbell, Law of Sentencing §§1:3, 4:6 4:8 (2d ed. 1991). These States should not be confronted with an unexpected rule mandating that what were once factors bearing upon the sentence now must be treated as offense elements for determination by the jury. This is especially so when, as here, what is at issue is not the conduct of the defendant, but the consequences of a completed criminal act.

A further disconcerting result of today's decision is the needless doubt the Court's analysis casts upon our cases involving capital sentencing. For example, while in Walton v. Arizona, 497 U.S. 639, 648 (1990), we viewed the aggravating factors at issue as sentencing enhancements and not as elements of the offense, the same is true of serious bodily injury under the reading of §2119 the Court rejects as constitutionally suspect. The question is why, given that characterization, the statutory scheme in Walton was constitutionally permissible. Under the relevant Arizona statute, Walton could not have been sentenced to death unless the trial judge found at least one of the enumerated aggravating factors. See Ariz. Rev. Stat. Ann. §13 703 (1989). Absent such a finding, the maximum potential punishment provided by law was a term of imprisonment. If it is constitutionally impermissible to allow a judge's finding to increase the maximum punishment for carjacking by 10 years, it is not clear why a judge's finding may increase the maximum punishment for murder from imprisonment to death. In fact, Walton would appear to have been a better candidate for the Court's new approach than is the instant case. In Walton, the question was the aggravated character of the defendant's conduct, not, as here, a result that followed after the criminal conduct had been completed.

In distinguishing this line of precedent, the Court suggests Walton did not "squarely fac[e]" the key constitutional question "implicated by the Government's position on the meaning of §2119(2)." Ante, at 24. The implication is clear. Reexamination of this area of our capital jurisprudence can be expected.

* * *

The Court misreads §2119 and seeks to create constitutional doubt where there is none. In my view, Almendarez-Torres controls this case. I would hold §2119 as interpreted by the Court of Appeals constitutional, and I dissent from the opinion and judgment of the Court.