with whom The ChieF Justice, Justice O’Connor, and Justice Kennedy join, dissenting.
The Court today holds that the amended version of Article 38.07 of the Texas Code of Criminal Procedure reduces the amount of proof necessary to support a sexual assault conviction, and that its retroactive application therefore violates the Ex Post Facto Clause. In so holding, the Court misreads both the Texas statute and our precedents concerning the Ex Post Facto Clause. Article 38.07 is not, as the Court would have it, most accurately characterized as a “sufficiency of the evidence rule”; it is in its essence an evidentiary provision dictating the circumstances under which the jury may credit victim testimony in sexual offense prosecutions. The amended version of Article 38.07 does nothing more than accord to certain victims of sexual offenses full testimonial stature, giving them the same undiminished competency to testify that Texas extends to witnesses generally in the State’s judicial proceedings. Our precedents make clear that such a witness competency rule validly may be applied to offenses committed before its enactment. I therefore dissent.
*554* * *
Petitioner Scott Leslie Carmell began sexually abusing his stepdaughter, “K. M.,” in the spring of 1991, when K. M. was 13 years old. He continued to do so through March 1995. The specific question before the Court concerns Carmell’s sexual assault on K. M. in June 1992, when K. M. was 14.1 K. M. did not inform anyone about that assault or about any of Carmell’s other sexual advances toward her until sometime around March 1995, when she told a friend and then her mother, Eleanor Alexander. Alexander went to the police, and Carmell was arrested and charged in a 15-count indictment.
Under Article 38.07 of the Texas Code of Criminal Procedure as it stood at the time of the assault, a conviction for sexual assault was supportable on the uncorroborated testimony of the victim if the victim was younger than 14 years old at the time of the offense. If the victim was 14 years old or older, however, the victim’s testimony could support a conviction only if that testimony was corroborated by other evidence. One form of corroboration, specifically described in Article 38.07 itself, was known as “outcry”: The victim’s testimony could support a conviction if he or she had informed another person, other than the defendant, about the offense within six months of its occurrence. Tex. Code Crim. Proc. Ann., Art. 38.07 (Vernon 1983).
Article 38.07 was amended in 1993. Under the new version, which was in effect at the time of Carmell’s trial, the victim’s uncorroborated testimony can support a conviction as long as the victim was under 18 years of age at the time of the offense. Tex. Code Crim. Proc. Ann., Art. 38.07 (Vernon Supp. 2000). The corroboration requirement con*555tinues in force for victims aged 18 or older, with a modified definition of outcry not material here. Thus, under the version of Article 38.07 in effect at the time of Carmell’s trial but not the version in effect at the time of the offense, his conviction was supportable by the uncorroborated testimony of K. M. The new version of Article 38.07 was applied at Carmell’s trial, and he was convicted.2 Carmell argues that the application of the new version of Article 38.07 to his trial violated the Ex Post Facto Clause, U. S. Const., Art. I, §10, cl. 1.
I
A proper understanding of Article 38.07 of the Texas Code of Criminal Procedure is central to this case. Accordingly, I turn first to the effect and purpose of that statute.
The effect of Article 38.07 in sexual offense prosecutions is plain. If the victim is of a certain age, the jury, in assessing whether the prosecution has met its burden of demonstrating guilt beyond a reasonable doubt, must give no weight to her testimony unless that testimony is corroborated, either by other evidence going directly to guilt or by "outcry.”3 For victims (such as K. M.) who were between the ages of 14 and *55618 at the time of the offense, the 1993 amendment repealed this corroboration requirement. The amended version of Article 38.07 thus permits sexual assault victims between 14 and 18 to have their testimony considered by the jury in the same manner and with the same effect as that of witnesses generally in Texas prosecutions.
This sort of corroboration requirement — still embodied in Article 38.07 for victims aged 18 or older — is a common, if increasingly outmoded, rule of evidence. Its purpose is to rein in the admissibility of testimony the legislature has deemed insufficiently credible standing alone. Texas’ requirement of corroboration or outcry, like similar provisions in other jurisdictions, is premised on a legislative judgment that accusations made by sexual assault victims above a certain age are not independently trustworthy. See Villareal v. State, 511 S. W. 2d 500, 502 (Tex. Crim. App. 1974) (“The basis of this rule is that the failure to make an outcry or promptly report the rape diminishes the credibility of the prosecutrix.”); cf., e. g., Battle v. United States, 630 A. 2d 211, 217 (D. C. 1993) (evidence of outcry “rebuts an implied charge of recent fabrication, which springs from some jurors’ assumptions that sexual offense victims are generally lying and that the victim’s failure to report the crime promptly is inconsistent with the victim’s current statement that the assault occurred”).
Legislatures in many States, including Texas, have enacted similar evidentiary provisions requiring corroboration for the testimony of other categories of witnesses, particularly accomplices. See, e. g., Tex. Code Crim. Proc. Ann., Art. 38.14 (Vernon Supp. 2000) (“A conviction cannot be had upon the testimony of an accomplice unless corroborated by other evidence tending to connect the defendant with the offense committed . . . .”). Such provisions — generally on the wane but still in force in several States — are, like Article 38.07, designed to ensure the credibility of the relevant witness. See, e. g., State v. Haugen, 448 N. W. 2d 191, 194 *557(N. D. 1989) (“The purpose of corroborating evidence is to show that accomplices are reliable witnesses and worthy of credit.”); Holladay v. State, 709 S. W. 2d 194, 196 (Tex. Crim. App. 1986) (“Because such a witness [i e., an accomplice] is usually deemed to be corrupt, his testimony is always looked upon with suspicion.”); Fleming v. State, 760 P. 2d 208, 209-210 (Okla. Crim. App. 1988) (“The purpose behind the requirement of corroboration is to protect an accused from being falsely implicated by another criminal in the hope of clemency, a desire for revenge, or for any other reason.”).
I make no judgment here as to the propriety of the Texas Legislature’s decision to view the testimony of certain sexual assault victims in the same light as that of accomplices. Ex post facto analysis does not depend on an assessment of a statute’s wisdom. For current purposes it suffices to note that Article 38.07’s corroboration requirement rests on the same rationale that underpins accomplice corroboration requirements: the notion that a particular witness, because of his or her role in the events at issue, might not give trustworthy testimony. See Reed v. State, 991 S. W. 2d 354, 361 (Tex. App. 1999) (“Generally speaking, the need to corroborate the testimony of a sexual assault victim stems from the notion that the victim, if over the age of consent, could be an accomplice rather than a victim.”); Hernandez v. State, 651 S. W. 2d 746, 751 (Tex. Crim. App. 1983) (concurring opinion adopted on rehearing) (Article 38.07’s corroboration requirement “was meant to deal only with testimony of a victim of a sexual offense who, for one reason or another, was held to be an ‘accomplice witness’ and, perforce, whose testimony must be corroborated.”).
The history of Article 38.07 bears out the view that its focus has always been on the competency and credibility of the victim as witness. The origins of the statute could be traced to the fact that in Texas, “for many years a seduced female was an incompetent witness as a matter of law.” Holladay, 709 S. W. 2d, at 200. See, e.g., Cole v. State, 40 *558Tex. 147 (1874); see also Hernandez, 651 S. W. 2d, at 751-752 (tracing the current Article 38.07 to the earlier seduction victim competency rule). In 1891, this common-law disability was lifted by statute and replaced by a corroboration requirement: “In prosecutions for seduction . . . the female alleged to have been seduced shall be permitted to testify; but no conviction shall be had upon the testimony of the said female, unless the same is corroborated by other evidence tending to connect the defendant with the offense charged.” Tex. Rev. Crim. Stat., Tit. 8, ch. 7, Art. 789 (1911). The application of this statute to offenses committed before its enactment was upheld by the Texas courts on the authority of Hopt v. Territory of Utah, 110 U. S. 574 (1884). See Mrous v. State, 31 Tex. Crim. App. 597, 21 S. W. 764 (1893). The corroboration requirement for seduction prosecutions, recodified in 1965 at Tex. Code Crim. Proc. Ann., Art. 38.07, remained in effect until 1973, when the entire 1925 Penal Code (including the offense of seduction) was repealed.
In 1975, Article 38.07 was enacted substantially in its present form. As revised, the article covered all sexual offenses in Chapter 21 of the Texas Penal Code; however, it contained no express exemption from the corroboration requirement for the testimony of the youngest victims. Tex. Code Crim. Proc. Ann., Art. 38.07 (Vernon 1979). The exemption for victims under the age of 14 was added in 1983, and extended in 1993 to cover those under the age of 18, as already described. As initially proposed, the 1993 change would have eliminated the eorroboration/outcry requirement altogether. House Research Organization, Texas House of Representatives, Daily Floor Report 13 (Mar. 15, 1998), Lodging of Petitioner. Supporters of the proposal maintained that “[victims in sexual assault cases are no more likely to fantasize or misconstrue the truth than the victims of most other crimes, which do not require corroboration of testimony or previous 'outcry/ Juries can decide if a witness is credi-ble_Most states no longer require this type of corrobora*559tion; neither should Texas.” Id., at 14. The historical development of Article 38.07 reveals a progressive alleviation of restrictions on the competency of victim testimony, not a legislative emphasis on the quantum of evidence needed to convict.
The version of Article 38.07 applied at Carmell’s trial was thus, in both effect and purpose, an evidentiary rule governing the weight that may be given to the testimony of sexual assault victims who had attained the age of 14. The Court’s efforts to paint it as something more than that are detached from the statute’s moorings and are consequently unpersuasive.
To begin with, it is beyond doubt that Article 38.07 does not establish an element of the offense. See Love v. State, 499 S. W. 2d 108, 108 (Tex. Crim. App. 1973) (“[Ojutery is not one of the elements of the offense charged.”). To convict a defendant of sexual assault in Texas today as before 1993, the prosecution need not introduce the victim’s testimony at all, much less any corroboration of that testimony. The Court is therefore less than correct in asserting that “[u]nder the law in effect at the time the acts were committed, the prosecution’s case was legally insufficient and petitioner was entitled to a judgment of acquittal, unless the State could produce both the victim’s testimony and corroborative evidence.” Ante, at 530. Under both the old and new versions of the statute, a conviction could be sustained on the testimony of a single third-party witness, on purely circumstantial evidence, or in any number of other ways — so long as the admissible evidence presented is sufficient to prove all of the elements of the offense beyond a reasonable doubt.4 And under either version of Article 38.07, of course, *560the accused could be convicted, like any other defendant, on the basis of a guilty plea or a voluntary confession. Article 38.07, in other words, does not define “sexual assault proven by corroborated victim testimony” as a distinct offense from “sexual assault.” Rather, the measure operates only to restrict the State’s method of proving its case.5
And it does so without affecting in any way the burden of persuasion that the prosecution must satisfy to support a conviction. Under both the old and new versions of the statute, the applicable standard is proof beyond a reasonable doubt. The amendment in 1993 that repealed the corroboration requirement for victims between the ages of 14 and 18 did nothing to change that standard.
The Court recognizes that Article 38.07 does not affect the applicable burden of persuasion, see ante, at 539, but several times it asserts that the amended version of the statute “changed the quantum of evidence necessary to sustain a conviction,” ante, at 530 (emphasis added). See also ante, at 531 (amended law “permitted petitioner to be convicted with less than the previously required quantum of evidence”); ante, at 532-533 (amended law “[r]educ[es] the quantum of evidence necessary to meet the burden of proof” (emphases added)). If by the word “quantum” the Court means to refer to the burden of persuasion, these statements are simply incorrect and contradict the Court’s own acknowledgment. And if, as appears more likely, “quantum” refers to some required quantity or amount of proof, the Court is also wrong. The partial repeal of Article 38.07’s corroboration requirement did not change the quantity of proof necessary to convict in every case, for the simple reason that Texas has never required the prosecution to introduce any particular
*561number of witnesses or items of proof to support a sexual assault conviction.6
The Court also declares several times that the amended version of Article 38.07 “subverts the presumption of innocence.” See ante, at 532; see also ante, at 533, nn. 22, 23, 546. The phrase comes from Cummings v. Missouri, 4 Wall. 277 (1867), in which the Court struck down a series of post-Civil War amendments to the Missouri Constitution that imposed penalties on persons unable or unwilling to swear an oath that they had not aided the Confederacy. The amendments, the Court said in Cummings, “subvert the presumptions of innocence” because “[t]hey assume that the parties are guilty [and] . . . call upon [them] to establish their innocence” by swearing the oath. Id., at 328. Nothing of the kind is involved here. Article 38.07 did not impose a presumption of guilt on Carmell and then saddle him with the task of overcoming it. The burden of persuasion remained at all times with the State. See Tex. Code Crim. Proc. Ann., Art. 38.03 (Vernon Supp. 2000). Carmell’s presumption of innocence is thus untouched by the current Article 38.07’s recognition of K. M.’s full testimonial stature.
The Court places perhaps its greatest weight on the “sufficiency of the evidence” label, see ante, at 547-552, but the label will not stick. As just noted, Article 38.07 has never dictated what it takes in all cases, quantitatively or qualitatively, for evidence to be sufficient to convict. To the contrary, under both the old and new versions of the statute the *562prosecution’s admissible evidence will be sufficient to support a conviction if a rational factfinder presented with that evidence could find the defendant guilty beyond a reasonable doubt. The 1993 repeal of the corroboration requirement for victims between the ages of 14 and 18 did not lower that “sufficiency of the evidence” hurdle; it simply expanded the range of methods the State could use to surmount it.
To be sure, one might descriptively say in an individual case that the uncorroborated testimony of the victim would be “sufficient” to convict under the new version of Article 38.07 and “insufficient” under the old. But that cannot be enough to invalidate a statute as ex post facto. If it were, then all evidentiary rules that work to the defendant’s detriment would be unconstitutional as applied to offenses committed before their enactment — an outcome our cases decisively reject. See infra, at 570-571 (discussing Thompson v. Missouri, 171 U. S. 380 (1898), and Hopt v. Territory of Utah, 110 U. S. 574 (1884), which upheld the retroactive application of evidentiary rules governing the authentication of documents and the competency of felons to testify, respectively). A defendant whose conviction turned, for example, on an item of hearsay evidence considered inadmissible at the time of the offense but made admissible by a later enacted statute might accurately describe the new statute as one that permits conviction on less evidence than was “sufficient” under prior law. But our precedents establish that such a defendant has no valid ex post facto claim. See infra, at 570-571. Neither does Carmell.
The Court attempts to distinguish Article 38.07 from garden-variety evidentiary rules by asserting that the latter “are ordinarily evenhanded, in the sense that they may benefit either the State or the defendant in any given case.” Ante, at 533, n. 23. The truth of this assertion is not at all clear. Evidence is never admissible in its own right; it must be admitted for some purpose. Rules of admissibility typically take that basic fact into account, often restricting the *563use of evidence in a way that systematically disadvantages one side. Consider, for example, a rule providing that evidence of a rape victim’s sexual relations with persons other than the accused is admissible to prove consent, or a rule providing that evidence of a sexual assault defendant’s prior sexual offenses is inadmissible to show a propensity to commit that type of crime. A statute repealing either of the above rules would “always run in the prosecution’s favor... [by] mak[ing] it easier to convict the accused.” Ante, at 546.7 Yet no one (until today) has suggested that such a statute would be ex post facto as applied to offenses committed before its enactment.
The Court resists the conclusion that Article 38.07 functions as a rule of witness competency by asserting that “[b]oth before and after the amendment, the victim’s testimony was competent evidence.” Ante, at 544. In all but the most technical sense that blanket statement is dubious. If the victim was 14 years old or older at the time of the offense (18 or older under the amended statute) and her testimony is unbolstered by corroboration or outcry, the jury may not credit that testimony in determining whether the State has met its burden of proof. Such a victim is of course not literally forbidden from testifying, but that cannot make the difference for Ex Post Facto Clause purposes between a sufficiency of the evidence rule and a witness competency rule. Evidence to which the jury is not permitted to assign weight is, in reality, incompetent evidence.
*564Perhaps the Court has been misdirected by the wording of Article 38.07, which speaks in both its old and new versions of evidence upon which a “conviction ... is supportable.” See ante, at 547. That sounds like a “sufficiency of the evidence rule,” until one realizes that any evidence admissible in a criminal case — i. e., any evidence that a jury is entitled to consider in determining whether the prosecution has met its burden of persuasion — is at least potentially evidence upon which a “conviction.., is supportable.” Conversely, as I have just said, evidence to which the jury may give no weight in making that determination is effectively inadmissible.8
In short, no matter how it is phrased, the corroboration requirement of Article 38.07 is functionally identical to a conditional rule of witness competency. If the former version of Article 38.07 had provided instead that “the testimony of the victim shall be inadmissible to prove the defendant’s guilt unless corroborated,” it would produce the *565same results as the actual statute in every case. Not “in certain instances,” ante, at 551, or “in some situations,” ante, at 550, but in every case.9 Recognizing this equivalency, the Texas Court of Criminal Appeals has noted that the Texas accomplice corroboration rule is “a mere rule of evidence” even though “statutorily worded as a sufficiency standard.” Malik v. State, 953 S. W. 2d 234, 240, n. 6 (1997).10
In sum, the function and purpose of the corroboration requirement embedded in the former version of Article 38.07 was to ensure the credibility of the victim’s testimony, not otherwise to impede the defendant’s conviction. Our precedents, I explain next, make clear that the retroactive repeal *566of such an evidentiary rule does not violate the Ex Post Facto Clause.
II
The Ex Post Facto Clause, this Court has said repeatedly, furthers two important purposes. First, it serves “to assure that legislative Acts give fair warning of their effect and permit individuals to rely on their meaning until explicitly changed.” Weaver v. Graham, 450 U. S. 24, 28-29 (1981).11 Second, it “restricts governmental power by restraining arbitrary and potentially vindictive legislation.” Id., at 29; see also Landgraf v. USI Film Products, 511 U. S. 244, 267 (1994); Miller v. Florida, 482 U. S. 423, 429-430 (1987). The latter purpose has much to do with the separation of powers; like its textual and conceptual neighbor the Bill of Attainder Clause, the Ex Post Facto Clause aims to ensure that legislatures do not meddle with the judiciary’s task of adjudicating guilt and innocence in individual cases. Weaver, 450 U. S., at 29, n. 10.
The Court does not even attempt to justify its extension of the Clause in terms of these two fundamental purposes. That is understandable, for today’s decision serves neither purpose. The first purpose (fair warning and reliance), vital as it is, cannot tenably be relied upon by Carmell. He had ample notice that the conduct in which he engaged was illegal. He certainly cannot claim to have relied in any way on the preamendment version, of Article 38.07: He tendered *567no reason to anticipate that K. M. would not report the assault within the outcry period, nor any cause to expect that corroborating evidence would not turn up sooner or later. Nor is the Clause’s second purpose relevant .here, for there is no indication that the Texas Legislature intended to single out this defendant or any class of defendants for vindictive or arbitrary treatment. Instead, the amendment of Article 38.07 simply brought the rules governing certain victim testimony in sexual offense prosecutions into conformity with Texas law governing witness testimony generally.
In holding the new Article 38.07 unconstitutional as applied to Carmell, the Court relies heavily on the fourth category of ex post facto statutes enumerated by Justice Chase in his opinion in Calder v. Bull, 3 Dall. 386, 390 (1798): “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.” Justice Chase’s formulation was. dictum, of course, because Calder involved a civil statute and the Court held that the statute was not ex post facto for that reason alone. Moreover, Justices Paterson and Iredell in their own seriatim opinions gave no hint that they considered rules of evidence to fall within the scope of the Clause. See id., at 395-397 (Paterson, J.); id., at 398-400 (Iredell, J.). Still, this Court has come to view Justice Chase’s categorical enumeration as an authoritative gloss on the Ex- Post Facto Clause’s reach. Just a decade ago in Collins v. Youngblood, 497 U. S. 37 (1990), for instance, this Court reiterated that “the prohibition which may not be evaded is the one defined by the Calder categories.” Id., at 46.
If those words are placed in the context of the full text of the Collins opinion, however, a strong case can be made that Collins pared the number of Calder categories down to three, eliminating altogether the fourth category on which the Court today so heavily relies. As long ago as 1925, in Beazell v. Ohio, 269 U. S. 167, the Court cataloged ex post *568facto laws without mentioning Chase’s fourth category at all. Id., at 169-170. And in Collins the Court cited with apparent approval BeazelVs omission of the fourth category, 497 U. S., at 43, n. 3, declaring that “[t]he Beazell formulation is faithful to our best knowledge of the original understanding of the Ex Post Facto Clause: Legislatures may not retroactively alter the definition of crimes or increase the punishment for criminal acts.” Id., at 43. Collins concluded by reciting in the plainest terms the prohibitions laid down by the Ex Post Facto Clause: A statute may not “punish as a crime an act previously committed, which was innocent when done; nor make more burdensome the punishment for a crime, after its commission; nor deprive one charged with crime of any defense available according to law at the time when the act was committed.” Id., at 52. This recitation conforms to Calder’s first three categories, but not the fourth; changes in evidentiary rules are nowhere mentioned.12
The majority asserts that the Court has repeatedly endorsed Justice Chase’s formulation, “including, in particular, the fourth category,” and it offers an impressive-looking string citation in support of the claim. Ante, at 525. Yet all of those eases simply quoted or paraphrased Chase’s enumeration, a mechanical task that naturally entailed a recitation of the fourth category. Not one of them depended on that category for the judgment the Court reached.13 Nei*569ther did Justice Washington’s opinion in Ogden v. Saunders, 12 Wheat. 213 (1827), which is quoted extensively by the Court, ante, at 532. In fact, the Court has never until today relied on the fourth Calder category to invalidate the application of a statute under the Ex Post Facto Clause.
It is true that the Court has on two occasions struck down as ex post facto the retroactive application of rules governing the functioning of the criminal trial process — but both decisions have since been overruled. In Kring v. Missouri, 107 U. S. 221 (1883), the Court held that Missouri was forbidden to apply retroactively a state constitutional amendment providing that a plea of guilty to second-degree murder would not automatically serve on retrial as an acquittal of the charge of first-degree murder. And in Thompson v. Utah, 170 U. S. 343 (1898), the Court held that a change in state law reducing the number of petit jurors in criminal trials from 12 to 8 was ex post facto because it deprived the defendant of “a substantial right involved in his liberty.” Id., at 352. The Court in Collins overruled both Kring and Thompson v. Utah, concluding that neither decision was “consistent with the understanding of the term ‘ex post facto law’ at the time the Constitution was adopted.” Collins, 497 U. S., at 47, 50, 51-52.
The Court today offers a different reading of Collins. It concludes that Collins overruled Kring and Thompson v. Utah because those eases improperly construed the Ex Post Facto Clause to cover all “substantial protections,” and that the fourth Calder category consequently remains intact. *570That is a plausible reading of Collins, and I might well be prepared to accept it, were the issue presented here. But it is not. For purposes of this case, it does not matter whether Collins eliminated the fourth Calder category or left it undisturbed. For even if the fourth category remains viable, our precedents make clear that it cannot be stretched to fit the statutory change at issue here. Those precedents— decisions that fully acknowledged the fourth Calder category — firmly establish that retroactively applied changes in rules concerning the admissibility of evidence and the competency of witnesses do not raise Ex Post Facto Clause concerns.
In Thompson v. Missouri, 171 U. S. 380 (1898), this Court upheld against ex post facto attack the retroactive application of a statute that permitted the introduction of previously inadmissible evidence to demonstrate the authenticity of disputed writings. The new statute, the Court reasoned, “did nothing more than remove an obstacle arising out of a rule of evidence that withdrew from the consideration of the jury testimony which, in the opinion of the legislature, tended to elucidate the ultimate, essential fact to be established, namely, the guilt of the accused.” Id., at 387.
The case most similar to the one before us is Hopt v. Territory of Utah, 110 U. S. 574 (1884). In that case, a statute in effect at the time of the offense but repealed by the time of trial provided that felons were incompetent to testify. The defendant, whose conviction for capital murder had been based in large part on the testimony of a felon, claimed that the application of the new law to his trial was ex post facto. The Court rejected the defendant’s claim, adopting reasoning applicable to the instant case:
“Statutes which simply enlarge the class of persons who may be competent to testify in criminal cases are not ex post facto in their application to prosecutions for crimes committed prior to their passage; for they do *571not attach criminality to any act previously done, and which was innocent when done; nor aggravate any crime theretofore committed; nor provide a greater punishment therefor than was prescribed at the time of its commission; nor do they alter the degree, or lessen the amount or measure, of the proof which was made necessary to conviction when the crime was committed.” Id., at 589.
As the quoted passage shows, the Court in Hopt rejected the defendant’s Ex Post Facto Clause claim while retaining Calder’s fourth category. The same outcome should obtain today, for Hopt cannot meaningfully be distinguished from the instant case.
The Court asserts that “Article 38.07 plainly fits” the fourth Calder category, because “[Requiring only the victim’s testimony to convict, rather than the victim’s testimony plus other corroborating evidence is surely less testimony required to convict’ in any straightforward sense of those words.” Ante, at 530. Yet to declare Article 38.07 ex post facto on that basis is to overrule Hopt without saying so. For if the amended version of Article 38.07 requires “less testimony ... to convict,” then so do countless eviden-tiary rules, including the felon competency rule whose retroactive application we upheld in Hopt. In both this case and Hopt, a conviction based on evidence previously deemed inadmissible was sustained pursuant to a broadened rule regarding the competency of testimonial evidence. The mere fact that the new version of Article 38.07 makes some convictions easier to obtain cannot be enough to preclude its retroactive application. “Even though it may work to the disadvantage of a defendant, a procedural change is not ex post facto.” Dobbert v. Florida, 432 U. S. 282, 293 (1977).
In short, the Court’s expansive new reading of the Ex Post Facto Clause cannot be squared with this Court’s prior decisions. Rather than embrace such an unprecedented approach, I would advance a “commonsense understanding of *572Calder’s fourth category,” ante, at 530, one that comports with our precedents and with the underlying purposes of the Ex Post Facto Clause: Laws that reduce the burden of persuasion the prosecution must satisfy to win a conviction may not be applied to offenses committed before their enactment. To be sure, this reading would leave the fourth category with considerably less independent effect than it would have had in Justice Chase’s day, given our intervening decisions establishing the “beyond a reasonable doubt” standard as a constitutional minimum under the Due Process Clause. See, e. g., In re Winship, 397 U. S. 358 (1970); Jackson v. Virginia, 443 U. S. 307 (1979). But it is not a reading that necessarily renders the category meaningless even today. Imagine, for example, a statute requiring the prosecution to prove a particular sentencing enhancement factor — leadership role in the offense, say, or obstruction of justice — beyond a reasonable doubt. A new statute providing that the factor could be established by a mere preponderance of the evidence might rank as ex post facto if applied to offenses committed before its enactment. The same might be said of a statute retroactively increasing the defendant’s burden of persuasion as to an affirmative defense.
Burdens of persuasion are qualitative tests of sufficiency. Calder’s fourth category, however, encompasses quantitative sufficiency rules as well, for Justice Chase did speak of a law that “receives less... testimony, than the law required at the time of the commission of the offence.” 3 Dali., at 390 (emphasis added). Cf. Hopt, 110 U. S., at 590 (“Any statutory alteration of the legal rules of evidence which would authorize conviction upon less proof, in amount or degree, than was required when the offence was committed” might be ex post facto, (emphasis added)). Quantitative sufficiency rules are rare in modern Anglo-American law, but some do exist. Criminal statutes sometimes limit the prosecution to a particular form of proof, for example, the testimony of two witnesses to the same overt act. In modern Anglo-*573American law, such instances have been almost exclusively confined to two contexts: perjury, see Weiler v. United States, 323 U. S. 606 (1945), and treason, see U. S. Const., Art. III, §3, cl. 1 (“No Person shall be convicted of Treason unless on the Testimony of two Witnesses to the same overt Act, or on Confession in open Court.”). See generally Wig-more, Required Numbers of Witnesses; A Brief History of the Numerical System in England, 15 Harv. L. Rev. 83,100-108 (1901).
The treason statute in effect at the time of John Fenwick’s conspiracy, like the Treason Clause of our Constitution, embodied just such a quantitative sufficiency rule: As long as the accused traitor put the prosecution to its proof by pleading not guilty, the sworn testimony of two witnesses was necessary to support a conviction. The Court describes at great length the attainder of Fenwick, which served as a cautionary model for Justice Chase’s explication of the fourth category in Colder. See ante, at 526-530.14 This excursion into post-Restoration English history is diverting, but the Court’s statement that “the circumstances of petitioner’s case parallel those of Fenwick’s case 300 years earlier,” ante, at 530, simply will not wash. The preamendment version of Article 38.07 is nothing like the two-witness rule on which Fenwick vainly relied.15
First, the preamendment version of Article 38.07, unlike a two-witness rule, did not apply indifferently to all who testify. Rather, it branded a particular class of witnesses— *574sexual assault victims aged 14 or older — as less competent than others to speak in court. Second, as I have already, described, the Texas statute did not restrict the State to one prescribed form of proof. Both before and after the 1998 amendment, introduction of the victim’s corroborated testimony was neither required nor necessarily sufficient to sustain a conviction. Prosecutors’ compliance with both the old and new versions of Article 38.07 thus “says absolutely nothing about whether they have introduced a quantum of evidence sufficient to convict the offender.” Ante, at 547, 551-552.16 On the contrary, the only sufficiency rule applicable in Texas sexual offense prosecutions has always been a qualitative one: The State’s evidence must be sufficient to prove every element of the offense beyond a reasonable doubt.
That should not be surprising. It makes little sense in our modern legal system to conceive of standards of proof in quantitative terms. In a civil case, the winner is the party that produces better evidence, not the party that produces more evidence. Similarly, in a criminal trial the prosecution need not introduce any fixed amount of evidence, so long as the evidence it does introduce could persuade a rational fact-finder beyond a reasonable doubt. “Our system of justice rests on the general assumption that the truth is not to be determined merely by the number of witnesses on each side of a controversy. In gauging the truth of conflicting evidence, a jury has no simple formulation of weights and measures on which to rely. The touchstone is always credibility; the ultimate measure of testimonial worth is quality and not *575quantity.” Weiler, 323 U. S., at 608. If the Court wishes to rely on the fourth Colder category to render Texas’ altered evidentiary rule prospective only, it should do so forthrightly by overruling Hopt and Thompson v. Missouri, rather than by attempting to portray Article 38.07 as a quantitative sufficiency rule indistinguishable from the two-witness requirement that figured in John Fenwick’s case.
* * *
In sum, it is well settled (or was until today) that retroactive changes to rules concerning the admissibility of evidence and the competency of witnesses to testify cannot be ex post facto. Because Article 38.07 is in both function and purpose a rule of admissibility, Thompson v. Missouri, Hopt, Beazell, and Collins dictate that its retroactive application does not violate the Ex Post Facto Clause. That conclusion comports perfectly with the dual purposes that underlie the Clause: ensuring fair notice so that individuals can rely on the laws in force at the time they engage in conduct, and sustaining the separation of powers while preventing the passage of vindictive legislation. The Court today thus not only brings about an “undefined enlargement of the Ex Post Facto Clause,” Collins, 497 U. S., at 46, that conflicts with established precedent, it also fails to advance the Clause’s fundamental purposes. For these reasons, I dissent.
The Court correctly notes that Carmell’s ex post facto challenge applies equally to three other counts on which he was convicted. Ante, at 518-519. This Court’s grant of review, however, was limited to the first question presented in Carmell’s petition for certiorari, which encompassed only the count charging the June 1992 assault. Pet. for Cert. 4.
The Texas Court of Appeals did not rule on whether the State in fact did corroborate K. M.’s testimony at trial. I note the testimony of K. M.’s mother that when she visited Carmell in jail and told him he needed to confess if he was sorry for what he had done, he wrote “‘adultery with [K. M.]’” on a piece of paper. 963 S. W. 2d 833, 835 (Tex. App. 1998). That testimony might count as corroboration. Because this question is outside the grant of certiorari, I (like the Court, see ante, at 519, n. 4) do not further address it.
At first glance one might object that the statute permits the jury to give such testimony some weight, just not enough to support a conviction. See, e. g., ante, at 546, n. 33 (contending that under the old Article 38.07, “the victim’s testimony alone is not inadmissible, it is just insufficient”). A moment’s reflection should reveal, however, that this distinction is illusory. If a particular item of evidence cannot by itself support a conviction, then the jury will not be permitted to consider it unless and until corroborating evidence is introduced.
Not only is corroborated victim testimony not necessary for a conviction under the former version of Article 38.07, it is not always sufficient. Under both the old and new versions of the statute, the prosecution’s evidence will not support a conviction unless it is adequate to prove all the elements of the offense beyond a reasonable doubt.
By the same reasoning, the repeal of the corroboration requirement for victims between the ages of 14 and 18 plainly did not deprive sexual assault defendants of any defense they previously enjoyed.
Moreover, even in a case founded on the victim’s testimony, the pre-1993 version of Article 38.07 would permit the prosecution to corroborate that testimony without introducing any additional evidence going to the defendant’s guilt, because corroboration could be provided by outcry, which is hearsay and inadmissible to prove the truth of the matter asserted. See Heckathorne v. State, 697 S. W. 2d 8, 12 (Tex. App. 1985) (“[A]n outcry should not be admitted for its truth, but merely as evidence that the victim informed someone of the offense.”).
Cf. Fed. Rules Evid. 412(a)(1) (restricting admissibility of “[e]vidence offered to prove that any alleged victim [of sexual misconduct] engaged in other sexual behavior”); 412(b)(1)(B) (providing that “evidence of specific instances of sexual behavior by the alleged victim with respect to the person accused” is admissible to prove consent); 413(a) (providing that “evidence of the defendant’s commission of another offense or offenses of sexual assault is admissible” in sexual assault eases notwithstanding Rule 404(b)’s general prohibition on the introduction of prior bad acts evidence “to show action in conformity therewith”).
It is thus no wonder that before 1986 the general rule of witness competency was codified at Article 38.06 of the Texas Code of Criminal Procedure, and the statute now at issue immediately followed it. Article 38.07 was an exception to the general rule laid out in Article 38.06. It is logical to put an exception right after the rule. Yet the Court draws the opposite inference from that juxtaposition. See ante, at 545, n. 32.
The Court’s related observation that Texas’ general witness competency statute “already contains its own provision respecting child witnesses,” ante, at 544-545, is true but irrelevant. Article 38.07’s corroboration requirement has nothing to do with the diminished credibility of child witnesses. Indeed, the statute has always permitted juries to credit fully the testimony of sexual offense victims below a certain age (first 14, then 18) without any corroboration, the reason apparently being that the legislature considers victims under a certain age to be too young to consent to sex and then lie about it. See, e. g., Scoggan v. State, 799 S. W. 2d 679, 681 (Tex. Crim. App. 1990); Hernandez v. State, 651 S. W. 2d 746, 752-753 (Tex. Crim. App. 1983) (concurring opinion adopted on rehearing). The corroboration requirement attaches only to victims above a certain age, and thus would not be appropriate for inclusion in a “provision respecting child witnesses.”
The Court contends that the effect of Article 38.07 is distinct from that of a witness competency rule because noncompliance with the former dictates acquittal ex proprio vigore while noncompliance with the latter dictates acquittal “in combination with the normally operative sufficiency rule.” Ante, at 552, n. 35. This is a distinction without a difference, because the “normally operative sufficiency rule” in question — when the prosecution submits no admissible evidence, its case will be deemed insufficient — is a bedrock requirement of due process, applicable in every criminal trial.
The Court observes that the characterization of a state law under the Ex Post Facto Clause is a federal question. Ante, at 544, n. 31. This undoubtedly correct observation stands in some tension, however, with the Court’s rebanee on the assertion that “Texas courts treat Article 38.07 as a sufficiency of the evidence rule.” Ante, at 518, n. 2. In any event, the latter assertion is inaccurate, as Malik's discussion of the accomplice coiToboration rule suggests. It is true that a trial court’s failure to comply with Article 38.07 results on appeal in the entry of an order of acquittal. But it is not true that the remedy on appeal for the introduction of inadmissible evidence is always a remand for a new trial. When the only evidence introduced by the prosecution is evidence that may not be considered by a jury in determining the defendant’s guilt, the proper result is always acquittal. By the same reasoning, as this Court decided just this Term, when a court of appeals has found that evidence was improperly admitted in a civil trial and that the remaining evidence is insufficient, it may enter judgment as a matter of law rather than ordering a new trial. Weisgram v. Marley Co., 528 U. S. 440 (2000).
Today's opinion apart, see ante, at 531, n. 21, this Court has consistently stressed “ ‘lack of fair notice’ ” as one of the “central concerns of the Ex Post Facto Clause.” Lynce v. Mathis, 519 U. S. 433, 441 (1997) (quoting Weaver v. Graham, 450 U. S. 24, 30 (1981)). See also Landgraf v. USI Film Products, 511 U. S. 244, 266-267 (1994); Miller v. Florida, 482 U. S. 423, 430 (1987); Weaver, 450 U. S., at 28-29; Marks v. United States, 430 U. S. 188, 191-192 (1977). The implausibility of ex ante reliance on rules of admissibility like the one at issue here helps explain why the Ex Post Facto Clause has never been held to apply to changes in such rules.
In California Dept. of Corrections v. Morales, 514 U. S. 499, 504-505 (1995), the Court similarly enumerated the categories of ex post facto laws without mentioning the fourth category.
The Court in Cummings v. Missouri, 4 Wall. 277 (1867), invoked the fourth category, see id., at 328, but that invocation was hardly necessary to the Court’s holding. In Cummings, as already noted, the Court invalidated on Bill of Attainder Clause and Ex Post Facto Clause grounds state constitutional amendments that imposed punishment on persons unable to swear an oath that they had not taken up arms against the Union in the Civil War. The Court recognized that the challenged amendments, though framed in terms of a method of proof, were “aimed at past acts, *569and not future acts,” id., at 327, for only those who had aided the Confederacy would be unable to take the expurgatory oath. The Court held that the amendments violated Calder’s first category by retroactively creating new offenses, 4 Wall., at 327-328, and violated the third category by retroactively imposing new punishments, id., at 328. As for Calder’s fourth category, the Court said only that the amendments “subvert[ed] the presumptions of innocenee” by “assumfing] that the parties [we]re guilty.” 4 Wall, at 328. As already discussed, supra, at 561, that analysis is of no help to Carmell here.
Tellingly, the Court offers no evidence that anyone at the time of the Framers considered witness eorroboration requirements of the type involved here to fall within the scope of the ex post facto prohibition. 15
When the Texas Legislature wants to enact a two-witness rule, it knows how to do so. See Tex. Code Crim. Proc. Ann., Art. 38.15 (Vernon Supp. 2000) (“No person can be convicted of treason except upon the testimony of at least two witnesses to the same overt act, or upon his own confession in open eourt.”); Art. 38.18(a) (“No person may be convicted of perjury or aggravated perjury if proof that his statement is false rests solely upon the testimony of one witness other than the defendant.”).
Noncompliance with the former version of Article 38.07 does say something: The statute mandates acquittal if the prosecution comes forward with no evidence beyond the victim’s testimony, which is deemed unreliable standing alone. But as the Court itself recognizes, “a witness competency rule that. . . has the practical effect of telling us what evidence would result in acquittal does not really speak to Calder’s fourth category.” Ante, at 551.