[609]*609Justice Breyer
delivered the opinion of the Court.
California has brought a criminal prosecution after expiration of the time periods set forth in previously applicable statutes of limitations. California has done so under the authority of a new law that (1) permits resurrection of otherwise time-barred criminal prosecutions, and (2) was itself enacted after pre-existing limitations periods had expired. We conclude that the Constitution’s Ex Post Facto Clause, Art. I, §10, cl. 1, bars application of this new law to the present case.
I
In 1993, California enacted a new criminal statute of limitations governing sex-related child abuse crimes. The new statute permits prosecution for those crimes where “[t]he limitation period specified in [prior statutes of limitations] has expired” — provided that (1) a victim has reported an allegation of abuse to the police, (2) “there is independent evidence that clearly and convincingly corroborates the victim’s allegation,” and (3) the prosecution is begun within one year of the victim’s report. 1993 Cal. Stats, ch. 390, § 1 (codified as amended at Cal. Penal Code Ann. § 803(g) (West Supp. 2003)). A related provision, added to the statute in 1996, makes clear that a prosecution satisfying these three conditions “shall revive any cause of action barred by [prior statutes of limitations].” 1996 Cal. Stats, ch. 130, § 1 (codified at Cal. Penal Code Ann. § 803(g)(3)(A) (West Supp. 2003)). The statute thus authorizes prosecution for criminal acts committed many years beforehand — and where the original limitations period has expired — as long as prosecution begins within a year of a victim’s first complaint to the police.
In 1998, a California grand jury indicted Marion Stogner, the petitioner, charging him with sex-related child abuse committed decades earlier — between 1956 and 1973. With[610]*610out the new statute allowing revival of the State’s cause of action, California could not have prosecuted Stogner. The statute of limitations governing prosecutions at the time the crimes were allegedly committed had set forth a 3-year limitations period. And that period had run 22 years or more before the present prosecution was brought.
Stogner moved for the complaint’s dismissal. He argued that the Federal Constitution’s Ex Post Facto Clause, Art. I, § 10, cl. 1, forbids revival of a previously time-barred prosecution. The trial court agreed that such a revival is unconstitutional. But the California Court of Appeal reversed, citing a recent, contrary decision by the California Supreme Court, People v. Frazer, 21 Cal. 4th 737, 982 P. 2d 180 (1999), cert. denied, 529 U. S. 1108 (2000). Stogner then moved to dismiss his indictment, arguing that his prosecution is unconstitutional under both the Ex Post Facto Clause and the Due Process Clause, Amdt. 14, § 1. The trial court denied Stog-ner’s motion, and the Court of Appeal upheld that denial. Stogner v. Superior Court, 93 Cal. App. 4th 1229, 114 Cal. Rptr. 2d 37 (2001). We granted certiorari to consider Stogner’s constitutional claims. 537 U. S. 1043 (2002).
II
The Constitution’s two Ex Post Facto Clauses prohibit the Federal Government and the States from enacting laws with certain retroactive effects. See Art. I, §9, cl. 3 (Federal Government); Art. I, §10, cl. 1 (States). The law at issue here created a new criminal limitations period that extends the time in which prosecution is allowed. It authorized criminal prosecutions that the passage of time had previously barred. Moreover, it was enacted after prior limitations periods for Stogner’s alleged offenses had expired. Do these features of the law, taken together, produce the kind of retroactivity that the Constitution forbids? We conclude that they do.
[611]*611First, the new statute threatens the kinds of harm that, in this Court’s view, the Ex Post Facto Clause seeks to avoid. Long ago Justice Chase pointed out that the Clause protects liberty by preventing governments from enacting statutes with “manifestly unjust and oppressive” retroactive effects. Calder v. Bull, 3 Dall. 386, 391 (1798). Judge Learned Hand later wrote that extending a limitations period after the State has assured “a man that he has become safe from its pursuit... seems to most of us unfair and dishonest.” Falter v. United States, 23 F. 2d 420, 426 (CA2), cert. denied, 277 U. S. 590 (1928). In such a case, the government has refused “to play by its own rules,” Carmell v. Texas, 529 U. S. 513, 533 (2000). It has deprived the defendant of the “fair warning,” Weaver v. Graham, 450 U. S. 24, 28 (1981), that might have led him to preserve exculpatory evidence. F. Wharton, Criminal Pleading and Practice § 316, p. 210 (8th ed. 1880) (“The statute [of limitations] is ... an amnesty, declaring that after a certain time ... the offender shall be at liberty to return to his country ... and . .. may cease to preserve the proofs of his innocence”). And a Constitution that permits such an extension, by allowing legislatures to pick and choose when to act retroactively, risks both “arbitrary and potentially vindictive legislation,” and erosion of the separation of powers, Weaver, supra, at 29, and n. 10. See Fletcher v. Peck, 6 Cranch 87, 137-138 (1810) (viewing the Ex Post Facto Clause as a protection against “violent acts which might grow out of the feelings of the moment”).
Second, the kind of statute at issue falls literally within the categorical descriptions of ex post facto laws set forth by Justice Chase more than 200 years ago in Calder v. Bull, supra—a categorization that this Court has recognized as providing an authoritative account of the scope of the Ex Post Faceto Clause. Collins v. Youngblood, 497 U. S. 37, 46 (1990); Carmell, supra, at 539. Drawing substantially on Richard Wooddeson’s 18th-century commentary on the nature of ex post facto laws and past parliamentary abuses, [612]*612Chase divided ex post facto laws into categories that he described in two alternative ways. See 529 U. S., at 522-524, and n. 9. He wrote:
“I will state what laws I consider ex post facto laws, within the words and the intent of the prohibition. 1st. Every law that makes an action done before the passing of the law, and which was innocent when done, criminal; and punishes such action. 2d. Every law that aggravates a crime, or makes it greater than it was, when committed. 3d. Every law that changes the punishment, and inflicts a greater punishment, than the law annexed to the crime, when committed. 4th. Every law that alters the legal rules of evidence, and receives less, or different, testimony, than the law required at the time of the commission of the offence, in order to convict the offender. All these, and similar laws, are manifestly unjust and oppressive.” Calder, supra, at 390-391 (emphasis altered from original).
In his alternative description, Chase traced these four categories back to Parliament’s earlier abusive acts, as follows:
Category 1: “Sometimes they respected the crime, by declaring acts to be treason, which were not treason, when committed.”
Category 2: “[A]t other times they inflicted punishments, where the party was not, by law, liable to any punishment.”
Category 3: “[I]n other cases, they inflicted greater punishment, than the law annexed to the offence.”
Category 4: “[A]t other times, they violated the rules of evidence (to supply a deficiency of legal proof) by admitting one witness, when the existing law required two; by receiving evidence without oath; or the oath of the wife against the husband; or other testimony, which the courts of justice would not admit. ” 3 Dall., at 389 (emphasis altered from original).
[613]*613The second category — including any “law that aggravates a crime, or makes it greater than it was, when committed,” id., at 390 — describes California’s statute as long as those words are understood as Justice Chase understood them— i. e., as referring to a statute that “inflict[s] punishments, where the party was not, by law, liable to any punishment,” id., at 389. See also 2 R. Wooddeson, A Systematical View of the Laws of England 638 (1792) (hereinafter Wooddeson, Systematical View) (discussing the ex post facto status of a law that affects punishment by “making therein some innovation, or creating some forfeiture or disability, not incurred in the ordinary course of law” (emphasis added)). After (but not before) the original statute of limitations had expired, a party such as Stogner was not “liable to any punishment.” California’s new statute therefore “aggravated” Stogner’s alleged crime, or made it “greater than it was, when committed,” in the sense that, and to the extent that, it “inflicted punishment” for past criminal conduct that (when the new law was enacted) did not trigger any such liability. See also H. Black, American Constitutional Law § 266, p. 700 (4th ed. 1927) (hereinafter Black, American Constitutional Law) (“[A]n act condoned by the expiration of the statute of limitations is no longer a punishable offense”). It is consequently not surprising that New Jersey’s highest court long ago recognized that Chase’s alternative description of second category laws “exactly describes the operation” of the kind of statute at issue here. Moore v. State, 43 N. J. L. 203, 217 (1881) (emphasis added). See also H. Black, Constitutional Prohibitions Against Legislation Impairing the Obligation of Contracts, and Against Retroactive and Ex Post Facto Laws §235, p. 298 (1887) (hereinafter Black, Constitutional Prohibitions) (“Such a statute” “certainly makes that a punishable offense which was previously a condoned and obliterated offense”).
So to understand the second category (as applying where a new law inflicts a punishment upon a person not then sub[614]*614ject to that punishment, to any degree) explains why and how that category differs from both the first category (making criminal noncriminal behavior) and the third category (aggravating the punishment). And this understanding is consistent, in relevant part, with Chase’s second category examples — examples specifically provided to illustrate Chase’s alternative description of laws “‘inflicting] punishments, where the party was not, by law, liable to any punishment,’ ” Calder, 3 Dall., at 389.
Following Wooddeson, Chase cited as examples of such laws Acts of Parliament that banished certain individuals accused of treason. Id., at 389, and n. ‡; see also Carmell, 529 U. S., at 522-524, and n. 11. Both Chase and Wooddeson explicitly referred to these laws as involving “banishment.” Calder, supra, at 389, and n. ‡; 2 Wooddeson, Systematical View 638-639. This fact was significant because Parliament had enacted those laws not only after the crime’s commission, but under circumstances where banishment “was simply not a form of penalty that could be imposed by the courts.” Carmell, supra, at 523, n. 11; see also 11 W. Holdsworth, A History of English Law 569 (1938). Thus, these laws, like the California law at issue here, enabled punishment where it was not otherwise available “in the ordinary course of law,” 2 Wooddeson, Systematical View 638. As this Court previously recognized in Carmell, supra, at 523, and n. 11, it was this vice that was relevant to Chase’s purpose.
It is true, however, that Parliament’s Acts of banishment, unlike the law in this case, involved a punishment (1) that the legislature imposed directly, and (2) that courts had never previously had the power to impose. But these differences are not determinative. The first describes not a retroactivity problem but an attainder problem that Justice Chase’s language does not emphasize and with which the Constitution separately deals, Art. I, § 9, cl. 3; Art. I, § 10, cl. 1. The second difference seems beside the point. The example of [615]*615Parliament’s banishment laws points to concern that a legislature, knowing the accused and seeking to have the accused punished for a pre-existing crime, might enable punishment of the accused in ways that existing law forbids. That fundamental concern, related to basic concerns about retroactive penal laws and erosion of the separation of powers, applies with equal force to punishment like that enabled by California’s law as applied to Stogner — punishment that courts lacked the power to impose at the time the legislature acted. See Black, Constitutional Prohibitions §235, at 298 (“It would be superfluous to point out that such an act [reviving otherwise time-barred criminal liability] would fall within the evils intended to be guarded against by the prohibition in question”). Cf. 1 F. Wharton, Criminal Law §444a, pp. 347-348, n. b (rev. 7th ed. 1874) (hereinafter Criminal Law).
In finding that California’s law falls within the literal terms of Justice Chase’s second category, we do not deny that it may fall within another category as well. Justice Chase’s fourth category, for example, includes any “law that alters the legal rules of evidence, and receives less, or different, testimony, than the law required at the time of the commission of the offence, in order to convict the offender.” Calder, supra, at 390. This Court has described that category as including laws that diminish “the quantum of evidence required to convict.” Carmell, supra, at 532.
Significantly, a statute of limitations reflects a legislative judgment that, after a certain time, no quantum of evidence is sufficient to convict. See United States v. Marion, 404 U. S. 307, 322 (1971). And that judgment typically rests, in large part, upon evidentiary concerns — for example, concern that the passage of time has eroded memories or made witnesses or other evidence unavailable. United States v. Kubrick, 444 U. S. 111, 117 (1979); 4 W. LaFave, J. Israel, & N. King, Criminal Procedure § 18.5(a), p. 718 (1999); Wharton, Criminal Pleading and Practice §316, at 210. Indeed, this [616]*616Court once described statutes of limitations as creating “a presumption which renders proof unnecessary.” Wood v. Carpenter, 101 U. S. 135, 139 (1879).
Consequently, to resurrect a prosecution after the relevant statute of limitations has expired is to eliminate a currently existing conclusive presumption forbidding prosecution, and thereby to permit conviction on a quantum of evidence where that quantum, at the time the new law is enacted,,would have been legally insufficient. And, in that sense, the new law would “violate” previous evidence-related legal rules by authorizing the courts to “ ‘receiv[e] evidence . . . which the courts of justice would not [previously have] admitted]’ ” as sufficient proof of a crime, supra, at 612. Cf. Collins, 497 U. S., at 46 (“Subtle ex post facto violations are no more permissible than overt ones”); Cummings v. Missouri, 4 Wall. 277, 329 (1867) (The Ex Post Facto Clause “cannot be evaded by the form in which the power of the State is exerted”). Nonetheless, given Justice Chase’s description of the second category, we need not explore the fourth category, or other categories, further.
Third, likely for the reasons just stated, numerous legislators, courts, and commentators have long believed it well settled that the Ex Post Facto Clause forbids resurrection of a time-barred prosecution. Such sentiments appear already to have been widespread when the Reconstruction Congress of 1867 — the Congress that drafted the Fourteenth Amendment — rejected a bill that would have revived time-barred prosecutions for treason that various Congressmen wanted brought against Jefferson Davis and “his coconspirators,” Cong. Globe, 39th Cong., 2d Sess., 279 (1866-1867) (comments of Rep. Lawrence). Radical Republicans such as Roscoe Conkling and Thaddeus Stevens, no friends of the South, opposed the bill because, in their minds, it proposed an “ex post facto law,” id., at 68 (comments of Rep. Conkling), and threatened an injustice tantamount to “judicial murder,” id., at 69 (comments of Rep. Stevens). In this instance, Con[617]*617gress ultimately passed a law extending unexpired limitations periods, ch. 2B6, 15 Stat. 183 — a tailored approach to extending limitations periods that has also been taken in modern statutes, e.g., 18 U. S. C. §3293 (notes on effective date of 1990 amendment and effect of 1989 amendment); Cal. Penal Code Ann. §805.5 (West Supp. 2003).
Further, Congressmen such as Conkling were not the only ones who believed that laws reviving time-barred prosecutions are ex post facto. That view was echoed in roughly contemporaneous opinions by State Supreme Courts. E. g., State v. Sneed, 25 Tex. Supp. 66, 67 (1860); Moore, 43 N. J. L., at 216-217. Cf. State v. Keith, 63 N. C. 140, 145 (1869) (A State’s repeal of an amnesty was “substantially an ex post facto law”). Courts, with apparent unanimity until California’s decision in Frazer, have continued to state such views, and, when necessary, so to hold. E. g., People ex rel. Reibman v. Warden, 242 App. Div. 282, 285, 275 N. Y. S. 59, 62 (1934); United States v. Fraidin, 63 F. Supp. 271, 276 (Md. 1945); People v. Shedd, 702 P. 2d 267, 268 (Colo. 1985) (en banc) (per curiam); State v. Hodgson, 108 Wash. 2d 662, 667-669, 740 P. 2d 848, 851-852 (1987) (en banc), cert. denied sub nom. Fied v. Washington, 485 U. S. 938 (1988); Commonwealth v. Rochelean, 404 Mass. 129, 130-131, 533 N. E. 2d 1333, 1334 (1989); State v. Nunn, 244 Kan. 207, 218, 768 P. 2d 268, 277-278 (1989); State v. O’Neill, 118 Idaho 244, 247, 796 P. 2d 121, 124 (1990); State v. Hirsch, 245 Neb. 31, 39-40, 511 N. W. 2d 69, 76 (1994); State v. Schultzen, 522 N. W. 2d 833, 835 (Iowa 1994); State v. Comeau, 142 N. H. 84, 88, 697 A. 2d 497, 500 (1997) (citing State v. Hamel, 138 N. H. 392, 395-396, 643 A. 2d 953, 955-956 (1994)); Santiago v. Commonwealth, 428 Mass. 39, 42, 697 N. E. 2d 979, 981, cert. denied, 525 U. S. 1003 (1998). Cf. Thompson v. State, 54 Miss. 740, 743 (1877) (stating, without specifying further grounds, that a new law could not take away a vested statute-of-limitations defense); State v. Cookman, 127 Ore. App. 283, 289, 873 P. 2d 335, 338 (1994) (holding that a law resurrecting a time-barred crimi[618]*618nal case “violates the Due Process Clause”), aff’d on state-law grounds, 324 Ore. 19, 920 P. 2d 1086 (1996); Commonwealth v. Guimento, 341 Pa. Super. 95, 97-98, 491 A. 2d 166, 167-168 (1985) (enforcing a state ban on ex post facto laws apparently equivalent to the federal prohibition); People v. Chesebro, 185 Mich. App. 412, 416, 463 N. W. 2d 134, 135-136 (1990) (reciting “the general rule” that, “ ‘where a complete defense has arisen under [a statute of limitations], it cannot be taken away by a subsequent repeal thereof’ ”).
Even where courts have upheld extensions of unexpired statutes of limitations (extensions that our holding today does not affect, see supra, at 613), they have consistently distinguished situations where limitations periods have expired. Further, they have often done so by saying that extension of existing limitations periods is not ex post facto “provided,” “so long as,” “because,” or “if” the prior limitations periods have not expired — a manner of speaking that suggests a presumption that revival of time-barred criminal eases is not allowed. E. g., United States v. Madia, 955 F. 2d 538, 540 (CA8 1992) (“ ‘provided’ ”); United States v. Richardson, 512 F. 2d 105, 106 (CA3 1975) (“provided”); People v. Anderson, 53 Ill. 2d 437, 440, 292 N. E. 2d 364, 366 (1973) (“so long as”); United States v. Haug, 21 F. R. D. 22, 25 (ND Ohio 1957) (“so long as”), aff’d, 274 F. 2d 885 (CA6 1960), cert. denied, 365 U. S. 811 (1961); United States v. Kurzenknabe, 136 F. Supp. 17, 23 (NJ 1955) (“so long as”); State v. Duffy, 300 Mont. 381, 390, 6 P. 3d 453, 460 (2000) (“because”); State v. Davenport, 536 N. W. 2d 686, 688 (N. D. 1995) (“because”); Andrews v. State, 392 So. 2d 270, 271 (Fla. App. 1980) (“if”), review denied, 399 So. 2d 1145 (Fla. 1981). See, e. g., Shedd, supra, at 268 (citing Richardson, supra, and Andrews, supra, as directly supporting a conclusion that a law reviving time-barred offenses is ex post facto). Cf. Commonwealth v. Duffy, 96 Pa. 506, 514 (1881) (“[I]n any case where a right to acquittal has not been absolutely acquired by the completion of the period of limitation, that period [619]*619is subject to enlargement or repeal without being obnoxious to the constitutional prohibition against ex post facto laws”).
Given the apparent unanimity of pre-Frazer case law, legal scholars have long had reason to believe this matter settled. As early as 1887, Henry Black reported that, although “not at all numerous,” the “cases upon this point... unmistakably point to the conclusion that such an act would be ex post facto in the strict sense, and void.” Constitutional Prohibitions §235, at 297. Even earlier, in 1874, Francis Wharton supported this conclusion by emphasizing the historic role of statutes of limitations as “acts of grace or oblivion, and not of process,” “extinguishing] all future prosecution” and making an offense unable to “be again called into existence at the caprice of the prince.” 1 Criminal Law §444a, at 347-348, n. b. More modern commentators — reporting on the same and subsequent cases — have come to the same conclusion. E. g., 21 Am. Jur. 2d, Criminal Law § 294, pp. 349-350 (1998 and Supp. 2002); 16A C. J. S., Constitutional Law §420, p. 372 (1984 and Supp. 2002); 4 LaFave, Israel, & King, Criminal Procedure § 18.5(a), at 718, n. 6; 2 C. Antieau & W. Rich, Modern Constitutional Law §38.11, p. 445 (2d ed. 1997); Adlestein, Conflict of the Criminal Statute of Limitations with Lesser Offenses at Trial, 37 Wm. & Mary L. Rev. 199, 246 (1995); C. Corman, Limitation of Actions § 1.6, p. 35 (1993 Supp.); Black, Statutes of Limitations and the Ex Post Facto Clauses, 26 Ky. L. J. 42 (1937); Black, American Constitutional Law §266, at 700. Cf. H. Wood, Limitation of Actions § 13, p. 43 (3d ed. 1901) (The State “may be said” to be “estopped from prosecuting”). Likewise, with respect to the closely related case of a law repealing an amnesty — a case not distinguished by the dissent — William Wade concluded early on that “[s]uch an act would be as clearly in contravention of the inhibition of ex post facto laws as though it undertook to annex criminality to an act innocent when done.” Operation and Construction of Retroactive Laws [620]*620§286, p. 389 (1880). But cf. post, at 638-639 (Kennedy, J., dissenting).
This Court itself has not previously spoken decisively on this matter. On the one hand, it has clearly stated that the Fifth Amendment’s privilege against self-incrimination does not apply after the relevant limitations period has expired. Brown v. Walker, 161 U. S. 591, 597-598 (1896). And that rule may suggest that the expiration of a statute of limitations is irrevocable, for otherwise the passage of time would not have eliminated fear of prosecution.
On the other hand, in Stewart v. Kahn, 11 Wall. 493, 503-504 (1871), this Court upheld a statute, enacted during the Civil War, that retroactively tolled all civil and criminal limitations for periods during which the war had made service of process impossible or courts inaccessible. Stewart, however, involved a civil, not a criminal, limitations statute. Id., at 500-501. Significantly, in reviewing this civil case, the Court upheld the statute as an exercise of Congress’ war powers, id., at 507, without explicit consideration of any potential collision with the Ex Post Facto Clause. Moreover, the Court already had held, independent of Congress’ Act, that statutes of limitations were tolled for “the time during which the courts in the States lately in rebellion were closed to the citizens of the loyal States ....” Id., at 503; see also Hanger v. Abbott, 6 Wall. 532, 539-542 (1868). Hence, the Court could have seen the relevant statute as ratifying a pre-existing expectation of tolling due to wartime exigencies, rather than as extending limitations periods that had truly expired. See id., at 541; see also Stewart, supra, at 507. In our view, Stewart therefore no more dictates the outcome here than does seemingly contrary precedent regarding the Fifth Amendment privilege.
Instead, we believe that the outcome of this case is determined by the nature of the harms that California’s law creates, by the fact that the law falls within Justice Chase’s second category as Chase understood that category, and by [621]*621a long line of authority holding that a law of this type violates the Ex Post Facto Clause.
III
In a prodigious display of legal and historical textual research, the dissent finely parses cases that offer us support, see post, at 633-637; shows appreciation for 19th-century dissident commentary, see post, at 638-639; discusses in depth its understanding of late 17th-century and early 18th-century parliamentary history, post, at 642-649; and does its best to drive a linguistic wedge between Justice Chase’s alternative descriptions of categories of ex post facto laws, post, at 640-641. All to what end? The dissent undertakes this Herculean effort to prove that it is not unfair, in any constitutionally relevant sense, to prosecute a man for crimes committed 25 to 42 years earlier when nearly a generation has passed since the law granted him an effective amnesty. Cf. post, at 649-653.
We disagree strongly with the dissent’s ultimate conclusion about the fairness of resurrecting a long-dead prosecution. See infra, at 630-632. Rather, like Judge Learned Hand, we believe that this retroactive application of a later-enacted law is unfair. And, like most other judges who have addressed this issue, see supra, at 617-618, we find the words “ex post facto” applicable to describe this kind of unfairness. Indeed, given the close fit between laws that work this kind of unfairness and the Constitution’s concern with ex post facto laws, we might well conclude that California’s law falls within the scope of the Constitution’s interdiction even were the dissent’s historical and precedent-related criticisms better founded than they are.
We need not examine that possibility here, however, because the dissent’s reading of the relevant history and precedent raises far too many problems to serve as a foundation for the reading of “ex post facto” that it proposes. In our view, that reading is too narrow; it is unsupported by prec[622]*622edent; and it would deny liberty where the Constitution gives protection.
A
In the dissent’s view, Chase’s historical examples show that “Calder’s second category concerns only laws” that both (1) “subjec[t] the offender to increased punishment” and (2) do so by “changing] the nature of an offense to make it greater than it was at the time of commission.” Post, at 642 (emphasis added). The dissent does not explain what it means by “changing] the nature of an offense,” but we must assume (from the fact that this language comes in a dissent) that it means something beyond attaching otherwise unavailable punishment and requires, in addition, some form of re-characterization of the crime. After all, the dissent seeks to show through its discussion of the relevant historical examples that a new law subjecting to punishment a person not then legally subject to punishment does not fall within the second category unless the new law somehow changes the kind of crime that was previously at issue.
The dissent’s discussion of the historical examples suffers from several problems. First, it raises problems of historical accuracy. In order to show the occurrence of a change in the kind or nature of the crime, the dissent argues that Parliament’s effort to banish the Earl of Clarendon amounted to an effort “to elevate criminal behavior of lower magnitude to the level of treason.” Post, at 643. The dissent supports this argument with a claim that “the allegations [against Clarendon] could not support a charge of treason.” Ibid. Historians, however, appear to have taken a different view. But cf. post, at 646. In their view, at least one charge against Clarendon did amount to treason.
Clarendon was charged with “betraying his majesty’s secret counsels to his enemies during the war.” Edward Earl of Clarendon’s Trial, 6 How. St. Tr. 292, 350 (1667) (hereinafter Clarendon’s Trial). In the words of one historian, this charge “undoubtedly contained treasonous matter.” [623]*623Roberts, The Impeachment of the Earl of Clarendon, 13 Camb. Hist. J. 1, 13 (1957) (hereinafter Roberts, Impeachment); accord, G. Miller, Edward Hyde, Earl of Clarendon 21-22 (1983); 10 Dictionary of National Biography 383 (L. Stephen & S. Lee eds. reprint 1922). See also Roberts, The Law of Impeachment in Stuart England: A Reply to Raoul Berger, 84 Yale L. J. 1419,1426 (1975); R. Berger, Impeachment: The Constitutional Problems 45, n. 193 (1974) (acknowledging and not contradicting the historian Henry Hal-lam’s conclusion that “ ‘one of the articles did actually contain an unquestionable treason'”). And it was on the basis of this specific charge — a charge of conduct that amounted to treason — that the House of Commons (which had previously refused to impeach Clarendon on other charges that did not amount to treason) “voted to impeach Clarendon for high treason.” Roberts, Impeachment 13; accord, Clarendon’s Trial 350-351.
The House of Lords initially thought that the Commons had failed to provide sufficient evidence because it failed to provide “special articles” laying out “particulars to prove it.” Roberts, Impeachment 14. The Lords and Commons deadlocked over whether a “general charge” was sufficient. Ibid. See. also Clarendon’s Trial 351-374. But Clarendon fled, thereby providing proof of guilt. 10 Dictionary of National Biography, supra, at 383; see also Clarendon’s Trial 389-390; 2 H. Hallam, Constitutional History of England: From the Accession of Henry VII to the Death of George II, p. 373 (8th ed. 1855). See also Berger, supra, at 44-45, and n. 189. The Lords and Commons then agreed to banish Clarendon. The Act of banishment — the only item in this complicated history explicitly cited by Chase — explained that Clarendon was being banished because he had “been impeached by the Commons ... of Treason and other misdemeanours” and had “fled whereby Justice cannot be done upon him according to his demerit.” 19 & 20 Car. II, c. 2 (1667-1668) (reprint 1963).
[624]*624In sum, Clarendon’s case involved Parliament’s punishment of an individual who was charged before Parliament with treason and satisfactorily proved to have committed treason, but whom Parliament punished by imposing “banishment” in circumstances where the party was not, in “the ordinary course of law,” liable to any “banishment” See supra, at 614. Indeed, because Clarendon had fled the country, it had become impossible to hold a proper trial to subject Clarendon to punishment through “ordinary” proceedings. See 19 & 20 Car. II, c. 2; Clarendon’s Trial 385-386. To repeat, the example of Clarendon’s banishment is an example of an individual’s being punished through legislation that subjected him to punishment otherwise unavailable, to any degree, through “the ordinary course of law” — just as Chase and his predecessor Wooddeson said. Calder, 3 Dall., at 389, and n. ‡; 2 Wooddeson, Systematical View 638. See also Carmell, 529 U. S., at 523, n. 11.
A second problem that the dissent’s account raises is one of historical completeness. That account does not explain how the second relevant example — the banishment of the Bishop of Atterbury — can count as an example of a recharac-terization of a pre-existing crime. The dissent concedes that Atterbury was charged with conduct constituting a “conspiracy to depose George I.” Post, at 647. It ought then to note (but it does not note) that, like the charge of “‘betraying his majesty’s secret counsels,’” supra, at 622, this charge was recognized as a charge of treason, see 2 J. Stephen, A History of the Criminal Law of England 266-267 (1883). As the dissent claims, the evidence upon which Parliament based its decision to banish may have been “meager,” and the punishment may even have been greater than some expected. Post, at 647. But the relevant point is that Parliament did not recharacterize the Bishop’s crime. Rather, through extraordinary proceedings that concluded with a punishment that only the legislature could impose, [625]*625Parliament aggravated a predefined crime by imposing a punishment that courts could not have imposed in “the ordinary course.”
Third, the dissent’s account raises a problem of vagueness. The dissent describes Justice Chase’s alternative description of the second category as “shed[ding] light on the meaning” of the category, post, at 641, and describes the historical references that accompany Chase’s alternative description as “illustrative examples,” post, at 649. But the question is would the dissent apply the term ex post facto to laws that fall within the alternative description — or would it not? If not, how does it reconcile its view with Carmell? See 529 U. S., at 522, n. 9; see also id., at 523 (Wooddeson’s categories “correlate precisely to Calder’s four categories”). If so, how does it explain the fact that the alternative description nowhere says anything about recharacterizing, or “changing the nature,” of a crime?
In our view, the key to the Atterbury and Clarendon examples lies not in any kind of recharacterization, or the like, but in the fact that Atterbury and Clarendon suffered the “same sentence” — “banishment.” 2 Wooddeson, Systematical View 638; see also Calder, supra, at 389, n. ‡ (using the word “banishment” to describe both examples). As we have argued, supra, at 614, Parliament aggravated the crimes at issue by imposing an otherwise unavailable punishment— namely, banishment — which was, according to Wooddeson, a “forfeiture or disability, not incurred in the ordinary course of law,” 2 Systematical View 638.
Fourth, the dissent’s initial account suffers from a technical problem of redundancy. Were the second category always to involve the recharacterization of an offense in a way that subjects it to greater punishment, see post, at 642, the second category would be redundant. Any law falling within it would also necessarily fall within the third category, which already encompasses “‘[e]very law that... inflicts a greater punishment,’ ” supra, at 612 (emphasis added).
[626]*626Fifth, the dissent’s historical account raises problems of pertinence. For one thing, to the extent that we are construing the scope of the Calder categories, we are trying not to investigate precisely what happened during the trials of Clarendon and Atterbury, but to determine how, several decades later, an 18th-century legal commentator and an 18th-century American judge who relied on that commentator — and, by extension, the Framers themselves — likely understood the scope of the words “ex post facto” Hence, the dissent’s account seems of little relevance once we recognize that:
(1) When Justice Chase set forth his alternative language for the second category (the language that the historical examples are meant to illuminate), he said nothing about recharacterizing crimes, Calder, 3 Dall., at 389;
(2) When Chase speaks of laws “declaring acts to be treason, which were not treason, when committed,” ibid., he uses this language for his alternative description of first category laws, and not second category laws, supra, at 612; and
(3) Wooddeson says nothing about recharacterizing crimes and instead uses the Clarendon and Atterbury examples to illustrate laws that “principally affect the punishment, making therein some innovation, or creating some forfeiture or disability, not incurred in the ordinary course of law,” 2 Systematical View 638 (some emphasis added).
Of course, we do not know whether Chase and Wooddeson, in using such language, had statutes of limitations specifically in mind. We know only that their descriptions of ex post facto laws and the relevant historical examples indicate an ex post facto category broad enough to include retroactive changes in, and applications of, those statutes. And we [627]*627know that those descriptions fit this case — the dissent’s historical exegesis notwithstanding.
More importantly, even were we to accept the dissent’s view that Chase’s second category examples involved some kind of recharacterization of criminal behavior (which they did not), why would recharacterization be the ex post facto touchstone? Why, in a case where (a) application of a previously inapplicable punishment and (b) recharacterization (or “changing the nature”) of criminal behavior do not come hand in hand, should the absence of the latter make a critical difference? After all, the presence of a recharacterization without new punishment works no harm. But the presence of the new punishment without recharacterization works all the harm. Indeed, it works retroactive harm — a circumstance relevant to the applicability of a constitutional provision aimed at preventing unfair retroactive laws. Perhaps that is why Justice Chase’s alternative description — which, like Wooddeson’s, speaks of laws “affect[ing] the punishment,” ibid.—does not mention recharacterization or the like.
B
The dissent believes that our discussion of the case law is “less persuasive than it may appear at a first glance.” Post, at 633. The dissent says that this case law is “deficient,” and that we rely on an “inapposite” case and other cases that “flatly contradict” the “principles” on which we rely. Post, at 634, 635.
Having reviewed the relevant cases and commentary, we continue to believe that our characterizations are accurate. We say that courts, “with apparent unanimity until California’s decision in Frazer, have continued to state” that “laws reviving time-barred prosecutions are ex post facto” and, “when necessary, so to hold.” Supra, at 617. That statement is accurate. The dissent refers to no case, outside of California, that has held, or even suggested, anything to the contrary.
[628]*628Of course, one might claim that the judges who wrote the cited opinions did not consider the matter as thoroughly as has the dissent or used precisely the same kind of reasoning. The dissent makes this kind of argument in its discussion of the old New Jersey case, Moore v. State, 43 N. J. L. 203 (1881)—a case that we believe supports our view. The dissent says that the Moore court “expressly stated that a statute reviving an expired limitations period ‘is not covered by any of [Justice Chase’s] classes.’” Post, at 635. And the dissent draws from this language the conclusion that Moore “flatly contradict^]” our views. Post, at 635.
The dissent, however, has taken the language that it quotes out of context. In context, the court’s statement reflects a conclusion that the language of Justice Chase’s first description of the categories (which Moore used the word “classes” to describe) does not fit cases in which a State revives time-barred prosecutions. The Moore court immediately adds, however, that Chase’s alternative description of second category laws does fit this case. Indeed, it “easily embraces” a statute that, like the statute here, retroactively extends an expired statute of limitations and “exactly describes [its] operation.” 43 N. J. L., at 216-217 (emphasis added). Had the New Jersey court had the benefit of Carmell, 529 U. S., at 522-524, and n. 9, or perhaps even of the dissent itself, post, at 641, would it not have recognized Chase’s alternative description as an authoritative account of elements of Chase’s “classes”? Would it then not have withdrawn its earlier statement, which the dissent quotes? Would it not have simply held that the statute did fall within the second category? Our reading of the case leads us to answer these questions affirmatively, but we leave the interested reader to examine the case and draw his or her own conclusions.
The dissent draws special attention to another case, State v. Sneed, 25 Tex. Supp. 66 (1860), arguing that it is “inappo-site” because it “avoided the issue” whether a law was [629]*629ex post facto “by holding that the statute was not meant to apply retroactively.” Post, at 634. Here is the court’s analysis, virtually in full:
“In this case the bar of the statute of limitations of one year was completed before the Code went into operation .... The state having neglected to prosecute within the time prescribed for its own action, lost the right to prosecute the suit. To give an Act of the Legislature, passed after such loss, the effect of reviving the right of action in the State, would give it an operation ex post facto, which we cannot suppose the Legislature intended.” 25 Tex. Supp., at 67.
The reader can make up his own mind.
Neither can we accept the dissent’s view that Judge Learned Hand’s like-minded comments in Falter were “unsupported,” post, at 637. In fact, Judge Hand’s comments had support in pre-existing case law, commentary, and published legislative debates, supra, at 616-620, and Hand’s opinion specifically cited Moore and two other early cases, Commonwealth v. Duffy, 96 Pa. 506 (1880), and People v. Buckner, 281 Ill. 340, 117 N. E. 1023 (1917). Falter, 23 F. 2d, at 425.
We add that, whatever the exact counts of categories of cases that we cite, cf. post, at 633, it is not surprising that most of these cases involve dicta, while only a handful involve clear holdings. Where the law has long been accepted as clearly settled, few cases are likely to arise, and cases that do arise most likely involve bordering areas of law, such as new limitations statutes enacted prior to expiration of preexisting limitations periods. Consistent with this expectation, one commentator noted in 1993 that the question whether to give retroactive effect to the extension of unexpired limitations periods had “become timely due to state legislature amendments during the early 1980s that lengthen the limitation period for the crimes of rape and sexual intercourse with a child.” Corman, Limitation of Actions §1.6, [630]*630at 36. The law at issue today represents a kind of extreme variant that, given the legal consensus of unconstitutionality, has not likely been often enacted in our Nation’s history. Cf. 1 J. Bishop, Criminal Law § 219a, p. 127 (rev. 4th ed. 1868) (declining to answer whether a law reviving time-barred prosecutions was ex post facto in part because “it is not likely to come before the courts”).
Neither should it be surprising if the reasoning in a string of cases stretching back over nearly 150 years is not perfectly consistent with modern conceptions of how legal analysis should proceed. After all, Beazell v. Ohio, 269 U. S. 167 (1925), an opinion relied on by the dissent, post, at 640, is itself vulnerable to criticism that its “method of analysis is foreclosed by this Court’s precedents,” post, at 637. See Collins, 497 U. S., at 45-46. In assessing the case law, we find the essential fact to be the unanimity of judicial views that the kind of statute before us is ex post facto. See supra, at 617-619.
The situation is similar with respect to commentators. Here, the essential fact is that, over a span of well over a century, commentators have come to the same conclusion, and have done so with virtual unanimity. See supra, at 619-620. We say “virtual,” for the dissent identifies one commentator who did not, namely, Joel Bishop — the same commentator relied on 122 years ago by the dissent in Moore, supra, at 240. The Moore majority rejected Bishop’s Conclusion. So did other contemporary courts and commentators. Supra, at 617-620. We do the same.
C
The dissent says it is a “fallacy” to apply the label “ ‘unfair and dishonest’ ” to this statute,, a law that revives long-dead prosecutions. Post, at 650. The dissent supports this conclusion with three arguments. First, it suggests that “retroactive extension of unexpired statutes of limitations” is no less unfair. Ibid. Second, the dissent refers to the small [631]*631likelihood that “criminals keep calendars” to mark the expiration of limitations periods, and it mocks the possibility that revival “destroys a reliance interest.” Ibid. Third, the dissent emphasizes the harm that child molestation causes, a harm that “will plague the victim for a lifetime,” and stresses the need to convict those who abuse children. Post, at 651.
In making the first argument, the dissent reverses field, abandoning its historical literalism to appeal to practical consequences. But history, case law, and constitutional purposes all are relevant. At a minimum, the first two of these adequately explain the difference between expired and unexpired statutes of limitations, and Chase’s alternative description of second category laws itself supports such a distinction. See supra, at 613-614, 618-619.
In making its second argument, which denies the existence of significant reliance interests, the dissent ignores the potentially lengthy period of time (in this case, 22 years) during which the accused lacked notice that he might be prosecuted and during which he was unaware, for example, of any need to preserve evidence of innocence. See supra, at 609-610. Memories fade, and witnesses can die or disappear. See supra, at 615-616. Such problems can plague child abuse cases, where recollection after so many years may be uncertain, and “recovered” memories faulty, but may nonetheless lead to prosecutions that destroy families. See, e.g., Holdsworth, Is It Repressed Memory with Delayed Recall or Is It False Memory Syndrome? The Controversy and Its Potential Legal Implications, 22 Law & Psychol. Rev. 103, 103-104 (1998). Regardless, a constitutional principle must apply not only in child abuse cases, but in every criminal case. And, insofar as we can tell, the dissent’s principle would permit the State to revive a prosecution for any kind of crime without any temporal limitation. Thus, in the criminal context, the dissent goes beyond our prior statements of what is constitutionally permissible even in the [632]*632analogous civil context. Chase Securities Corp. v. Donaldson, 325 U. S. 304, 312, n. 8, 315-316 (1945) (acknowledging that extension of even an expired civil limitations period can unconstitutionally infringe upon a “vested right”); William Danzer & Co. v. Gulf & Ship Island R. Co., 268 U. S. 633, 637 (1925) (holding the same). But see post, at 638, 653. It is difficult to believe that the Constitution grants greater protection from unfair retroactivity to property than to human liberty.
As to the dissent’s third argument, we agree that the State’s interest in prosecuting child abuse cases is an important one. But there is also a predominating constitutional interest in forbidding the State to revive a long-forbidden prosecution. And to hold that such a law is ex post facto does not prevent the State from extending time limits for the prosecution of future offenses, or for prosecutions not yet time barred.
In sum, California’s law subjects an individual such as Stogner to prosecution long after the State has, in effect, granted an amnesty, telling him that he is “at liberty to return to his country . . . and that from henceforth he may cease to preserve the proofs of his innocence,” Wharton, Criminal Pleading and Practice §316, at 210. See also Moore, 43 N. J. L., at 223-224. It retroactively withdraws a complete defense to prosecution after it has already attached, and it does so in a manner that allows the State to withdraw this defense at will and with respect to individuals already identified. See supra, at 611. “Unfair” seems to us a fair characterization.
IV
The statute before us is unfairly retroactive as applied to Stogner. A long line of judicial authority supports characterization of this law as ex post facto. For the reasons stated, we believe the law falls within Justice Chase’s second category of ex post facto laws. We conclude that a law enacted after expiration of a previously applicable limita[633]*633tions period violates the Ex Post Facto Clause when it is applied to revive a previously time-barred prosecution. The California court’s judgment to the contrary is
Reversed.