dissenting.
In 1980, respondent was charged with the murder of his wife. Despite respondent’s previous conviction for first-degree murder, and despite the serious character of the 1980 offense, California accepted his plea of nolo contendere to the offense of second-degree murder. The trial judge imposed a sentence of imprisonment for 15 years to life, under which respondent became eligible for parole in 1990.
The law in effect at the time of respondent’s offense entitled him to a hearing before the Board of Prison Terms (Board) in 1989 and in each year thereafter. In 1981, however, California amended its parole statute. The amended statute permitted the Board to delay parole hearings for multiple murderers for up to three years, provided the Board found that “it is not reasonable to expect that parole would be granted at a hearing during the following years.” Cal. Penal Code Ann. § 3041.5(b)(2) (West 1982). In 1989, the Board determined that respondent was not yet “suitable” for parole, and, after making the requisite findings, the Board deferred respondent’s next hearing for three years. The question before the Court is whether the California Legislature’s 1981 elimination of the statutory right to an annual parole hearing increased the punishment for respondent’s 1980 offense and thereby violated the Ex Post Facto Clause.
In answering that question, I begin with certain propositions of law that I do not understand the Court to dispute. *515Those propositions compel the conclusion that, as applied to the general prison population, replacing a statutory right to an annual parole hearing with a right to such a hearing every three years would violate the Ex Post Facto Clause of the Federal Constitution. Though nowhere disputing this conclusion, the majority holds that the 1981 amendment to the California parole statute is not ex post facto legislation because it applies only to a small subset of the prison population, namely multiple murderers, see ante, at 510, and because the Board must make a special finding before depriving a prisoner of an annual hearing, see ante, at 511. In my view, neither of these features is sufficient to save what is otherwise a plainly invalid statute.
I
The Constitution provides that “[n]o State shall . . . pass any ... ex post facto Law.” Art. I, §10. The Framers viewed the prohibition on ex post facto legislation as one of the.fundamental protections against arbitrary and oppressive government.1 Thus, for example, Madison noted that “ex post facto laws ... are contrary to the first principles of the social compact and to every principle of sound legislation.” The Federalist No. 44, p. 282 (C. Rossiter ed. 1961). Similarly, Hamilton counted the prohibition on ex post facto laws among the three protections that he described as “greater securities to liberty and republicanism than any [the Constitution] contains.” Id., No. 84, at 511.
Although the text of the Ex Post Facto Clause is not self-explanatory, its basic coverage has been well understood at least since 1798, when the Court in Calder v. Bull, 3 Dall. *516386, 390, identified four categories of ex post facto laws.2 The case before us today implicates the third Calder category, which consists of “[e]very law that changes the punishment, and inflicts a greater punishment, than the law annexed to the crime, when committed.” Ibid, (emphasis in original). This Court has consistently condemned laws falling in that category. Thus, in Beazell v. Ohio, 269 U. S. 167 (1925), Justice Stone noted that it “is settled, by decisions of this Court so well known that their citation may be dispensed with, that any statute ... which makes more burdensome the punishment for a crime, after its commission, . . . is prohibited as ex post facto.” Id., at 169-170. We reaffirmed Justice Stone’s observation only a few years ago: “The Beazell formulation is faithful to our best knowledge of the original understanding of the Ex Post Facto Clause: Legislatures may not retroactively . . . increase the punishment for criminal acts.” Collins v. Youngblood, 497 U. S. 37, 43 (1990).
In light of the importance that the Framers placed on the Ex Post Facto Clause, we have always enforced the prohibition against the retroactive enhancement scrupulously. Any statute that authorizes an increased term of imprisonment for a past offense is invalid. Thus, although the Court has carefully examined laws changing the conditions of confinement to determine whether they are favorable or unfavorable to the prisoner, see, e. g., Rooney v. North Dakota, 196 U. S. 319, 325 (1905); In re Medley, 134 U. S. 160, 171 (1890), no Member of the Court has ever voted to uphold a statute *517that retroactively increased the length of time that a prisoner must remain imprisoned for past offenses, see, e.g., Miller v. Florida, 482 U. S. 423 (1987) (unanimous opinion); Weaver v. Graham, 450 U. S. 24 (1981) (without dissent).
Our ex post facto jurisprudence concerning increased punishment has established three important propositions. First, the Court has squarely held that an individual prisoner need not prove that the retroactive application of a law authorizing an increased punishment for a past offense has actually affected the sentence that that prisoner must serve. In Lindsey v. Washington, 301 U. S. 397 (1937), for example, petitioners were sentenced under a law that required' a sentence of 15 years; the law in effect at the time of the offense gave the judge discretion to impose a lesser sentence. The State contended that petitioners had failed to show that there was an ex post facto violation because petitioners might have received a 15-year sentence even under the old law. We unanimously rejected the State’s contention:
“[T]he ex post facto clause looks to the standard of punishment prescribed by a statute, rather than to the sentence actually imposed....
“Removal of the possibility of a sentence of less than fifteen years, at the end of which petitioners would be freed from further confinement and the tutelage of a parole revocable at will, operates to their detriment in the sense that the standard of punishment adopted by the new statute is more onerous than that of the old.” Id., at 401.
Only a few years ago, in Miller v. Florida, 482 U. S. 423 (1987), we unanimously reaffirmed the holding in Lindsey, noting that “Lindsey establishes ‘that one is not barred from challenging a change in the penal code on ex post facto grounds simply because the sentence he received under the new law was not more onerous than that which he might have received under the old.’” 482 U. S., at 432 (citation *518omitted). As we stated succinctly in Weaver v. Graham, 450 U. S., at 33, “[t]he inquiry looks to the challenged provision, and not to any special circumstances that may mitigate its effect on the particular individual.”
Second, we have noted that an impermissible increase in the punishment for a crime may result not only from statutes that govern initial sentencing but also from statutes that govern parole or early release. Thus, in Weaver v. Graham, we addressed a Florida statute that altered the availability of good-time credits. We rejected any notion that the removal of good-time credits did not constitute an increase in punishment, explaining that “a prisoner’s eligibility for reduced imprisonment is a significant factor entering into both the defendant’s decision to plea bargain and the judge’s calculation of the sentence to be imposed.” Id., at 32, citing Wolff v. McDonnell, 418 U. S. 539, 557 (1974); Warden v. Marrero, 417 U. S. 653, 658 (1974). See also Greenfield v. Scafati, 277 F. Supp. 644, 645 (Mass. 1967) (three-judge court) (“The availability of good conduct deductions is considered an essential element of the sentence”), summarily aff’d, 390 U. S. 713 (1968).
Finally, we have held that an increase in punishment occurs when the State deprives a person of the opportunity to take advantage of provisions for early release. Thus, in Weaver we emphasized that “petitioner is ... disadvantaged by the reduced opportunity to shorten his time in prison simply through good conduct.” 450 U. S., at 33-34. Our statement in Weaver was consistent with our holding in Lindsey that “[i]t is plainly to the substantial disadvantage of petitioners to be deprived of all opportunity to receive a sentence which would give them freedom from custody and control prior to the expiration of the 15-year term.” 301 U. S., at 401-402. See also Greenfield v. Scafati, 277 F. Supp. 644 (Mass. 1967) (three-judge court), summarily aff’d, 390 U. S. 713 (1968) (affirming judgment of a three-judge court that *519found an ex post facto violation in a statute that eliminated the opportunity to accumulate gain time for the first six months following parole revocation as applied to an inmate whose crime occurred before the statute’s enactment).
These settled propositions make perfectly clear that the retroactive application of a simple statute that changed the frequency of a statutorily mandated annual parole suitability hearing would constrict an inmate’s opportunity to earn early release and would thus constitute increased punishment in violation of the Ex Post Facto Clause. It is thus no surprise that nearly every Federal Court of Appeals and State Supreme Court to consider the issue has so held. See, e. g., 16 F. 3d 1001 (CA9 1994) (case below); Roller v. Cavanaugh, 984 F. 2d 120 (CA4), cert. dism’d, 510 U. S. 42 (1993); Akins v. Snow, 922 F. 2d 1558 (CA11), cert. denied, 501 U. S. 1260 (1991); Rodriguez v. United States Parole Commission, 594 F. 2d 170 (CA7 1979); State v. Reynolds, 642 A. 2d 1368 (N. H. 1994); Griffin v. State, 315 S. C. 285, 433 S. E. 2d 862 (1993), cert. denied, 510 U. S. 1093 (1994); Tiller v. Klincar, 138 Ill. 2d 1, 561 N. E. 2d 576 (1990), cert. denied, 498 U. S. 1031 (1991).3
The 1981 amendment at issue in this case, of course, is not such a simple statute. It is therefore necessary to consider whether the particular features of that amendment eliminate the ex post facto problems.
*520II
The first special feature that the majority identifies in the 1981 amendment, see ante, at 510, is that it applies only to the narrow class of prisoners who have “been convicted, in the same or different proceedings, of more than one offense which involves the taking of a life.” Cal. Penal Code Ann. § 3041.5(b)(2) (West 1982). In my view, the 1981 amendment’s narrow focus on that discrete class of prisoners implicates one of the principal concerns that underlies the constitutional prohibition against retrospective legislation — the danger that the legislature will usurp the judicial power and will legislate so as to administer justice unfairly against particular individuals. This concern has been at the forefront of our ex post facto jurisprudence. As Justice Harlan noted: “[T]he policy of the prohibition against ex post facto legislation would seem to rest on the apprehension that the legislature, in imposing penalties on past conduct,... may be acting with a purpose not to prevent dangerous conduct generally but to impose by legislation a penalty against specific persons or classes of persons.” James v. United States, 366 U. S. 213, 247, n. 3 (1961) (Harlan, J., concurring in part and dissenting in part). Our cases have thus consistently noted that the Ex Post Facto Clauses protect against the danger of such “vindictive legislation.” Miller v. Florida, 482 U. S., at 429; Weaver v. Graham, 450 U. S., at 29; see also Malloy v. South Carolina, 237 U. S. 180, 183 (1915). The narrower the class burdened by retroactive legislation, the greater the danger that the legislation has the characteristics of a bill of attainder.4 Cf. Plaut v. Spendthrift Farm, Inc., ante, at 241 *521(Breyer, J., concurring in judgment) (finding a separation-of-powers violation in part because of the statute’s “application to a limited number of individuals”); see generally ante, at 241-242 (discussing the dangers of statutes focused at particular groups of individuals).
I believe that the 1981 amendment implicates this core ex post facto concern. The narrow class of affected individuals belies the majority’s acceptance of the proposition that “the evident focus,” ante, at 507, of the 1981 amendment was to save costs. Surely, even today, multiple murderers make up but a small fraction of total parole hearings; eliminating those hearings would seem unlikely to create substantial savings. Indeed, though the majority gives credence to the budget-cutting rationale, petitioners are much more frank about their motivations, as they urge the Court to “reexamine” its ex post facto jurisprudence “[i]n view of the national trend towards the implementation of harsher penalties and conditions of confinement for offenders and inmates.” Brief for Petitioners 11 (footnote omitted).
I agree with petitioners’ implication that the 1981 amendment is better viewed as part of that national trend toward “get-tough-on-crime” legislation. The California statute challenged in this case is one of many currently popular statutes designed to cut back on the availability of parole. The California Legislature has adopted several similar provisions in recent years,5 and a number of other States have passed comparable legislation.6 Such measures are, of course, entirely legitimate when they operate prospectively, but their *522importance and prevalence surely justify careful review when those measures change the consequences of past conduct.
The danger of legislative overreaching against which the Ex Post Facto Clause protects is particularly acute when the target of the legislation is a narrow group as unpopular (to put it mildly) as multiple murderers. There is obviously little legislative hay to be made in cultivating the multiple murderer vote. For a statute such as the 1981 amendment, therefore, the concerns that animate the Ex Post Facto Clause demand enhanced, and not (as the majority seems to believe) reduced, judicial scrutiny.7
Ill
The second feature of the 1981 amendment on which the majority relies is the provision requiring that the Board make certain findings before it may defer the annual hearings. At the time of respondent’s crime, the Board was instructed either to set a parole date at an inmate’s initial parole hearing or, if it set no date, to provide the inmate with a written statement explaining the reasons for the denial and suggesting “activities in which he might participate that will benefit him while he is incarcerated.” The statute provided that the Board “shall hear each case annually thereafter.” Cal. Penal Code Ann. § 3041.5(b)(2) (West 1982).
The 1981 amendment allows the Board to defer the annual hearings for multiple murderers for up to three years if “the [Bjoard finds that it is not reasonable to expect that parole *523would be granted at a hearing during the following years and states the bases for the finding.” Ibid. The statute does not contain any provision authorizing any sort of review of a Board order dispensing with annual hearings. Nor does it provide any procedure for dealing with exceptional changed circumstances warranting the setting of a release date that might arise before the next scheduled hearing. In short, the amended statute vests unreviewable discretion in the Board to dispense with annual hearings for up to three years by making the required finding.
In my view, the requirement that the Board make this finding is insufficient to render the 1981 amendment constitutional. We have previously expressed doubts that an early release regime that substitutes administrative discretion for statutory requirements complies with the Ex Post Facto Clause. In Weaver v. Graham, we noted that the state statute at issue reduced the amount of gain time to which an inmate was “automatically entitled . . . simply for avoiding disciplinary infractions and performing his assigned tasks.” 450 U. S., at 35. The State argued, however, that the statute as a whole caused inmates no increase in punishment because the statute provided inmates with new ways to earn gain time. We rejected the State’s argument, noting that “the award of the extra gain time is purely discretionary, contingent on both the wishes of the correctional authorities and special behavior by the inmate.” Ibid.
The reasoning behind our skepticism in Weaver is applicable to this case. As is true under almost any factfinding regime, the Board will occasionally make mistakes and will defer parole hearings for inmates who would have been found suitable at those hearings.8 Because the parole hear*524ing is a prerequisite for early release, the inmates affected by the Board’s errors will have had their punishment increased. In my view, the Court’s speculation about possible methods of correcting the Board’s erroneous findings or of persuading the Board to reinstate a canceled hearing on the basis of new evidence is plainly insufficient to bridge the significant gap between the protection afforded by an unqualified right to annual hearings and the unreviewable discretion of an administrative agency to dispense with such hearings.9
IV
Two final elements of the majority’s opinion require comment. First, the majority suggests that a holding in respondent’s favor would require that we “invalidate” an “endless array of legislative adjustments,” thus plunging the judiciary into micromanagement of state parole procedures. Ante, at 508. The majority’s fear is completely unfounded. The provision of a parole hearing in California differs from all of the matters set forth by the majority in one critical way: It is an absolute prerequisite to release. For the three years in which respondent is denied his hearing, he is absolutely deprived of any parole opportunity. Though the changes to which the majority refers might well make it more difficult for prisoners to obtain release, none of them deprives prisoners of the opportunity for release. Our cases *525are absolutely clear that the retroactive deprivation of the opportunity for early release constitutes ex post facto legislation. The majority’s parade of hypothetical horribles is easily distinguishable from the case before us, and it thus provides no justification for diverging from our settled approach.
Second, the majority attempts to circumvent our ex post facto cases by characterizing the risk that the statute will actually increase any inmate’s punishment as “speculative.” In my view, the speculation runs in the other direction. Under the present California parole procedures, there is no possibility that an inmate will benefit from the 1981 amendment: Instead of an unqualified statutory right to an annual hearing, the amendment leaves the inmate with no protection against either the risk of a mistaken prediction or the risk that the Board may be influenced by its interest in curtailing its own workload. Moreover, the statute gives an inmate no right to advance favorable changed circumstances as a basis for a different result. Unlike the ex post facto law condemned in Weaver, and also unlike the statutes approved in Dobbert v. Florida, 432 U. S. 282 (1977), and Rooney v. North Dakota, 196 U. S. 319 (1905), the 1981 amendment contains no off-setting benefits for the inmate. By postponing and reducing the number of parole hearings, ostensibly for the sole purpose of cutting administrative costs, the amendment will at best leave an inmate in the same position he was in, and will almost inevitably delay the grant of parole in some cases.
The Court concludes, nevertheless, that it is “speculative” to say that the statute will increase inmates’ punishment. To draw such a conclusion, the Court “speculates” about the accuracy of the Board’s predictions, it “speculates” about the parole suitability of a class of prisoners, it “speculates” about the length of time that elapses between an eventual parole hearing and the ultimate release date, and it “speculates” as to the availability of procedures to deal with unexpected *526changes in circumstances. To engage in such pure speculation while condemning respondent’s assertion of increased punishment as “speculative” seems to me not only unpersuasive, but actually perverse.
I respectfully dissent.
That the Framers included two separate clauses in the Constitution prohibiting ex post facto legislation, see Art. I, §9, cl. 3 (“No Bill of Attainder or ex post facto Law shall be passed”); Art. I, § 10 (“No State shall . . . pass any Bill of Attainder, ex post facto Law, or Law impairing the Obligation of Contracts”), highlights the Framers’ appraisal of the importance of that 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.” Calder v. Bull, 3 Dall. 386, 390 (1798) (emphasis in original).
The two contrary decisions cited by the parties, see Bailey v. Gardening, 940 F. 2d 1150 (CA8 1991); In re Jackson, 39 Cal. 3d 464, 703 P. 2d 100 (1985), do not undermine my thesis. In Bailey v. Gardebring, a 2-to-l decision, the Court of Appeals found no ex post facto violation when Minnesota failed to provide a prisoner with an annual parole hearing. However, one member of the majority premised his conclusion on the view that the Minnesota parole regulations were not “laws”; the other member of the majority concurred only in the result, but authored no opinion. In In re Jackson, the California Supreme Court upheld the very amendment at issue in this case and thus did not speak to the more general situation I have described in the text.
“A bill of attainder is a legislative act which inflicts punishment without a judicial trial... . [Legislative acts, no matter what their form, that apply either to named individuals or to easily ascertainable members of a group in such a way as to inflict punishment on them without a judicial trial are bills of attainder prohibited by the Constitution.” United States v. Lovett, 328 U. S. 303, 315-316 (1946) (internal quotation marks omitted). The prohibitions on ex post facto laws and on bills of attainder *521are obviously closely related. See, e. g., Fletcher v. Peck, 6 Cranch 87, 138-139 (1810).
The California Legislature appears to have altered the frequency of parole hearings for some prisoners on at least three occasions since the 1981 amendment. See 1986 Cal. Stats., ch. 248, § 166; 1990 Cal. Stats., ch. 1053, § 1; 1994 Cal. Stats., ch. 560, § 1.
See, e. g., 1992 N. H. Laws, ch. 254:13; Mich. Comp. Laws Ann. § 791.234 (West 1992); Ill. Rev. Stat., ch. 38, ¶ 1003-3-5(f) (1987); S. C. Code Ann. § 24-21-645 (Supp. 1987).
Though the Court suggests that “multiple murderers” have a particularly low likelihood of parole, ante, at 510-611, n. 7, the statute in effect at the time of respondent’s offense determined that even multiple murderers were sufficiently likely candidates for early release to be entitled to an annual parole hearing. The grant of that statutory right reflected the California Legislature’s judgment that such a hearing provided an important avenue to reduced punishment. The Ex Post Facto Clause, properly construed, should prevent the legislature from revising that judgment retroactively.
There may be reasons to be particularly skeptical of the reliability of the Board’s findings with respect to deferrals under the 1981 amendment. The Board’s determination that the inmate is not currently suitable for parole and the determination that the inmate will not be suitable for parole in the next three years are expected to be separate determina*524tions. In the state-court litigation over the constitutionality of this statute, the State argued that compliance with the requirement of separate determinations was “ ‘virtually impossible’ ” because “ ‘[b]oth the decision to deny parole and to delay a subsequent hearing for two years must be the same.’ ” In re Jackson, 39 Cal. 3d, at 478, 703 P. 2d, at 109. Indeed, in respondent’s case, the findings on parole suitability and on the possibility of future parole are remarkably similar. The Board’s findings on which the majority relies so heavily thus seem of particularly questionable utility.
I find it somewhat ironic that the majority posits the existence of nonstatutory, extraordinary remedies as a cure for legislation ostensibly motivated entirely by an interest in administrative efficiency.