JUSTICE O'CONNOR, with whom THE CHIEF JUSTICE joins, dissenting.
Today the Court applies a new rule of retroactivity to impose crushing and unnecessary liability on the States, precisely at a time when they can least afford it. Were the Court's decision the product of statutory or constitutional command, I would have no choice but to join it. But nothing in the Constitution or statute requires us to adopt the retroactivity rule the majority now applies. In fact, longstanding precedent requires the opposite result. Because I see no reason to abandon our traditional retroactivity analysis as articulated in Chevron Oil Co. v. Huson, 404 U. S. 97, 106107 (1971), and because I believe the Supreme Court of Virginia correctly applied Chevron Oil in this case, I would affirm the judgment below.
I
This Court's retroactivity jurisprudence has become somewhat chaotic in recent years. Three Terms ago, the case of American Trucking Assns., Inc. v. Smith, 496 U. S. 167 (1990), produced three opinions, none of which garnered a majority. One Term later, James B. Beam Distilling Co. v. Georgia, 501 U. S. 529 (1991), yielded five opinions; there, no single writing carried more than three votes. As a result, the Court today finds itself confronted with such disarray that, rather than relying on precedent, it must resort to vote counting: Examining the various opinions in Jim Beam, it discerns six votes for a single proposition that, in its view, controls this case. Ante, at 96-97.
If we had given appropriate weight to the principle of stare decisis in the first place, our retroactivity jurisprudence never would have become so hopelessly muddled. After all, it was not that long ago that the law of retroactivity for civil cases was considered well settled. In Chevron Oil Co., we explained that whether a decision will be nonretroactive depends on whether it announces a new rule, whether prospectivity would undermine the purposes of the rule, and whether retroactive application would produce injustice. 404 U. S., at 106-107. Even when this Court adjusted the retroactivity rule for criminal cases on direct review some six years ago, we reaffirmed the vitality of Chevron Oil, noting that retroactivity in civil cases "continues to be governed by the standard announced in Chevron Oil Co. v. Huson." Griffith v. Kentucky, 479 U. S. 314, 322, n. 8 (1987). In American Trucking Assns., supra, however, a number of Justices expressed a contrary view, and the jurisprudential equivalent of entropy immediately took over. Whatever the merits of any retroactivity test, it cannot be denied that resolution of the case before us would be simplified greatly had we not disregarded so needlessly our obligation to follow precedent in the first place.
I fear that the Court today, rather than rectifying that confusion, reinforces it still more. In the usual case, of course, retroactivity is not an issue; the courts simply apply their best understanding of current law in resolving each case that comes before them. James B. Beam, 501 U. S., at 534, 535-536 (SOUTER, J.). But where the law changes in some respect, the courts sometimes may elect not to apply the new law; instead, they apply the law that governed when the events giving rise to the suit took place, especially where the change in law is abrupt and the parties may have relied on the prior law. See id., at 534. This can be done in one of two ways. First, a court may choose to make the decision purely prospective, refusing to apply it not only to the parties before the court but also to any case where the relevant facts predate the decision. Id., at 536. Second, a court may apply the rule to some but not all cases where the operative events occurred before the court's decision, depending on the equities. See id., at 537. The first option is called "pure prospectivity" and the second "selective prospectivity."
As the majority notes, ante, at 96-97, six Justices in James B. Beam, supra, expressed their disagreement with selective prospectivity. Thus, even though there was no majority opinion in that case, one can derive from that case the proposition the Court announces today: Once "this Court applies a rule of federal law to the parties before it, that rule… must be given full retroactive effect in all cases still open on direct review." Ante, at 97. But no decision of this Court forecloses the possibility of pure prospectivity-refusal to apply a new rule in the very case in which it is announced and every case thereafter. As JUSTICE WHITE explained in his concurrence in James B. Beam, "[t]he propriety of prospective application of decision in this Court, in both constitutional and statutory cases, is settled by our prior decisions." 501 U. S., at 546 (opinion concurring in judgment).
Rather than limiting its pronouncements to the question of selective prospectivity, the Court intimates that pure prospectivity may be prohibited as well. See ante, at 97 (referring to our lack of " 'constitutional authority… to disregard current law' "); ibid. (relying on "'basic norms of constitutional adjudication'" (quoting Griffith, supra, at 322)); see also ante, at 94 (touting the "fundamental rule of'retrospective operation'" of judicial decisions). The intimation is incorrect. As I have explained before and will touch upon only briefly here:"[W]hen the Court changes its mind, the law changes with it. If the Court decides, in the context of a civil case or controversy, to change the law, it must make [a] determination whether the new law or the old is to apply to conduct occurring before the law-changing decision. Chevron Oil describes our long-established procedure for making this inquiry." James B. Beam, supra, at 550 (O'CONNOR, J., dissenting) (internal quotation marks omitted).
Nor can the Court's suggestion be squared with our cases, which repeatedly have announced rules of purely prospective effect. See, e. g., Northern Pipeline Constr. Co. v. Marathon Pipe Line Co., 458 U. S. 50, 88 (1982); Chevron Oil, 404 u. S., at 106-107; Phoenix v. Kolodziejski, 399 U. S. 204, 214 (1970); Cipriano v. City of Houma, 395 U. S. 701, 706 (1969); see also American Trucking Assns., 496 U. S., at 188-200 (plurality opinion) (canvassing the Court's retroactivity jurisprudence); ante, at 110 (KENNEDY, J., concurring in part and concurring in judgment) (citing cases).
In any event, the question of pure prospectivity is not implicated here. The majority first holds that once a rule has been applied retroactively, the rule must be applied retroactively to all cases thereafter. Ante, at 97. Then it holds that Davis v. Michigan Dept. of Treasury, 489 U. S. 803 (1989), in fact retroactively applied the rule it announced. Ante, at 98-99. Under the majority's approach, that should end the matter: Because the Court applied the rule retroactively in Davis, it must do so here as well. Accordingly, there is no reason for the Court's careless dictum regarding pure prospectivity, much less dictum that is contrary to clear precedent.
Plainly enough, JUSTICE SCALIA would cast overboard our entire retroactivity doctrine with precisely the "unceremonious 'heave-ho'" he decries in his concurrence. See ante, at 109. Behind the undisguised hostility to an era whose jurisprudence he finds distasteful, JUSTICE SCALIA raises but two substantive arguments, both of which were raised in James B. Beam, 501 U. S., at 549 (SCALIA, J., concurring in judgment), and neither of which has been adopted by a majority of this Court. JUSTICE WHITE appropriately responded to those arguments then, see id., at 546 (opinion concurring in judgment), and there is no reason to repeat the responses now. As Justice Frankfurter explained more than 35 years ago:"We should not indulge in the fiction that the law now announced has always been the law…. It is much more conducive to law's self-respect to recognize candidly the considerations that give prospective content to a new pronouncement of law." Griffin v. Illinois, 351 U. S. 12, 26 (1956) (opinion concurring in judgment).
II
I dissented in James B. Beam because I believed that the absolute prohibition on selective prospectivity was not only contrary to precedent, but also so rigid that it produced unconscionable results. I would have adhered to the traditional equitable balancing test of Chevron Oil as the appropriate method of deciding the retroactivity question in individual cases. But even if one believes the prohibition on selective prospectivity desirable, it seems to me that the Court today takes that judgment to an illogical-and inequitable-extreme. It is one thing to say that, where we have considered prospectivity in a prior case and rejected it, we must reject it in every case thereafter. But it is quite another to hold that, because we did not consider the possibility of prospectivity in a prior case and instead applied a rule retroactively through inadvertence, we are foreclosed from considering the issue forever thereafter. Such a rule is both contrary to established precedent and at odds with any notion of fairness or sound decisional practice. Yet that is precisely the rule the Court appears to adopt today. Ante, at 96-97.
A
Under the Court's new approach, we have neither authority nor discretion to consider the merits of applying Davis v. Michigan Dept. of Treasury, supra, retroactively. Instead, we must inquire whether any of our previous decisions happened to have applied the Davis rule retroactively to the parties before the Court. Deciding whether we in fact have applied Davis retroactively turns out to be a rather difficult matter. Parsing the language of the Davis opinion, the Court encounters a single sentence it declares determinative:
"The State having conceded that a refund is appropriate in these circumstances, see Brief for Appellee 63, to the extent appellant has paid taxes pursuant to this invalid tax scheme, he is entitled to a refund." Id., at 817 (quoted in part, ante, at 98). According to the majority, that sentence constitutes "'consideration of remedial issues'" and therefore "'necessarily'" indicates that we applied the rule in Davis retroactively to the parties before us. Ante, at 98 (quoting James B. Beam, supra, at 539 (opinion of SOUTER, J.)). Ironically, respondent and its amici draw precisely the opposite conclusion from the same sentence. According to them, the fact that Michigan conceded that it would offer relief meant that we had no reason to decide the question of retroactivity in Davis. Michigan was willing to provide relief whether or not relief was required. The Court simply accepted that offer and preserved the retroactivity question for another day.
One might very well debate the meaning of the single sentence on which everyone relies. But the debate is as meaningless as it is indeterminate. In Brecht v. Abrahamson, 507 U. S. 619 (1993), we reaffirmed our longstanding rule that, if a decision does not "squarely addres[s] [an] issue," this Court remains "free to address [it] on the merits" at a later date. Id., at 631. Accord, United States v. L. A. Tucker Truck Lines, Inc., 344 U. S. 33, 38 (1952) (issue not "raised in briefs or argument nor discussed in the opinion of the Court" cannot be taken as "a binding precedent on thee] point"); Webster v. Fall, 266 U. S. 507, 511 (1925) ("Questions which merely lurk in the record, neither brought to the attention of the court nor ruled upon, are not considered as having been so decided as to constitute precedents"). The rule can be traced back to some of the earliest of this Court's decisions. See statement of Marshall, C. J., as reported in the arguments of counsel in United States v. More, 3 Cranch 159, 172 (1805) ("No question was made, in that case, as to the jurisdiction. It passed sub silentio, and the court does not consider itself as bound by that case"). Regardless of how one reads the solitary sentence upon which the Court relies, surely it does not "squarely address" the question of retroactivity; it does not even mention retroactivity. At best, by addressing the question of remedies, the sentence implicitly "assumes" the rule in Davis to be retroactive. Our decision in Brecht, however, makes it quite clear that unexamined assumptions do not bind this Court. Brecht, supra, at 631 (That the Court "assumed the applicability of" a rule does not bind the Court to the assumption).
In fact, there is far less reason to consider ourselves bound by precedent today than there was in Brecht. In Brecht, the issue was not whether a legal question was resolved by a single case; it was whether our consistent practice of applying a particular rule, Chapman v. California, 386 U. S. 18, 24 (1967), to cases on collateral review precluded us from limiting the rule's application to cases on direct review. Because none of our prior cases directly had addressed the applicability of Chapman to cases on collateral review-each had only assumed it applied-the Court held that those cases did not bind us to any particular result. See Brecht, supra, at 630-631. I see no reason why a single retroactive application of the Davis rule, inferred from the sparse and ambiguous language of Davis itself, should carry more weight here than our consistent practice did in Brecht.
The Court offers no justification for disregarding the settled rule we so recently applied in Brecht. Nor do I believe it could, for the rule is not a procedural nicety. On the contrary, it is critical to the soundness of our decisional processes. It should go without saying that any decision of this Court has wide-ranging applications; nearly every opinion we issue has effects far beyond the particular case in which it issues. The rule we applied in Brecht, which limits the stare decisis effect of our decisions to questions actually considered and passed on, ensures that this Court does not decide important questions by accident or inadvertence. By adopting a contrary rule in the area of retroactivity, the Court now permanently binds itself to its every unexamined assumption or inattention. Any rule that creates a grave risk that we might resolve important issues of national concern sub silentio, without thought or consideration, cannot be a wise one.
This case demonstrates the danger of such a rule. The question of retroactivity was never briefed in Davis. It had not been passed upon by the court below. And it was not within the question presented. Indeed, at oral argument we signaled that we would not pass upon the retroactivity of the rule Davis would announce. After conceding that the Michigan Department of Taxation would give Davis himself a refund if he prevailed, counsel for the department argued that it would be unfair to require Michigan to provide refunds to the 24,000 taxpayers who were not before the Court. The following colloquy ensued:"[COURT]: So why do we have to answer that at all? "[MICHIGAN]: -if, if this Court issues an opinion stating that the current Michigan classification is unconstitutional or in violation of the statute, there are these 24,000 taxpayers out there."[COURT]: But that's not-it's not here, is it? Is that question here?"[MICHIGAN]: It is not specifically raised, no." Tr. of Oral Arg., O. T. 1988, No. 87-1020, pp. 37-38.
Now, however, the Court holds that the question was implicitly before us and that, even though the Davis opinion does not even discuss the question of retroactivity, it resolved the issue conclusively and irretrievably.
If Davis somehow did decide that its rule was to be retroactive, it was by chance and not by design. The absence of briefing, argument, or even mention of the question belies any suggestion that the issue was given thoughtful consideration. Even the author of the Davis opinion refuses to ac cept the notion that Davis resolved the question of retroactivity. Instead, JUSTICE KENNEDY applies the analysis of Chevron Oil to resolve the retroactivity question today. See ante, at 110-112 (opinion concurring in part and concurring in judgment).
The Court's decision today cannot be justified by comparison to our decision in Griffith v. Kentucky, 479 U. S. 314 (1987), which abandoned selective prospectivity in the criminal context. Ante, at 97. As I explained in American Trucking Assns., 496 U. S., at 197-200, there are significant differences between criminal and civil cases that weigh against such an extension. First, nonretroactivity in criminal cases historically has favored the government's reliance interests over the rights of criminal defendants. As a result, the generalized policy of favoring individual rights over governmental prerogative can justify the elimination of prospectivity in the criminal arena. The same rationale cannot apply in civil cases, as nonretroactivity in the civil context does not necessarily favor plaintiffs or defendants; "nor is there any policy reason for protecting one class of litigants over another." Id., at 198. More important, even a party to civil litigation who is "deprived of the full retroactive benefit of a new decision may receive some relief." Id., at 198199. Here, for example, petitioners received the benefit of prospective invalidation of Virginia's taxing scheme. From this moment forward, they will be treated on an equal basis with all other retirees, the very treatment our intergovernmental immunity cases require. The criminal defendant, in contrast, is usually interested only in one remedy-reversal of his conviction. That remedy can be obtained only if the rule is applied retroactively. See id., at 199.
Nor can the Court's rejection of selective retroactivity in the civil context be defended on equal treatment grounds. See Griffith, supra, at 323 (selective retroactivity accords a benefit to the defendant in whose case the decision is announced but not to any defendant thereafter). It may well be that there is little difference between the criminal defendant in whose case a decision is announced and the defendant who seeks certiorari on the same question two days later. But in this case there is a tremendous difference between the defendant in whose case the Davis rule was announced and the defendant who appears before us today: The latter litigated and preserved the retroactivity question while the former did not. The Michigan Department of Taxation did not even brief the question of retroactivity in Davis. Respondent, in contrast, actually prevailed on the question in the court below.
If the Court is concerned with equal treatment, that difference should be dispositive. Having failed to demand the unusual, prospectivity, respondent in Davis got the usualnamely, retroactivity. Respondent in this case has asked for the unusual. In fact, respondent here defends a judgment below that awarded it just that. I do not see how the principles of equality can support forcing the Commonwealth of Virginia to bear the harsh consequences of retroactivity simply because, years ago, the Michigan Department of Taxation failed to press the issue-and we neglected to consider it. Instead, the principles of fairness favor addressing the contentions the Virginia Department of Taxation presses before us by applying Chevron Oil today. It is therefore to Chevron Oil that I now turn.
B
Under Chevron Oil, whether a decision of this Court will be applied nonretroactively depends on three factors. First, as a threshold matter, "the decision to be applied nonretroactively must establish a new principle of law." 404 U. S., at 106. Second, nonretroactivity must not retard the new rule's operation in light of its history, purpose, and effect. Id., at 107. Third, nonretroactivity must be necessary to avoid the substantial injustice and hardship that a holding of retroactivity might impose. Ibid. In my view, all three factors favor holding our decision in Davis nonretroactive. As JUSTICE KENNEDY points out in his concurrence, ante, at 111, a decision cannot be made nonretroactive unless it announces "a new principle of law." Chevron Oil, 404 U. S., at 106. For purposes of civil retroactivity, Chevron Oil identifies two types of decisions that can be new. First, a decision is new if it overturns "clear past precedent on which litigants may have relied." Ibid.; ante, at 111 (KENNEDY, J., concurring in part and concurring in judgment). I agree with JUSTICE KENNEDY that Davis did not represent such a " 'revolutionary'" or "'avulsive change'" in the law. Ante, at 112 (quoting Hanover Shoe, Inc. v. United Shoe Machinery Corp., 392 U. S. 481, 499 (1968)).
Nonetheless, Chevron also explains that a decision may be "new" if it resolves "an issue of first impression whose resolution was not clearly foreshadowed." Chevron Oil, supra, at 106 (emphasis added). Thus, even a decision that is "controlled by the… principles" articulated in precedent may announce a new rule, so long as the rule was "sufficiently debatable" in advance. Arizona Governing Comm. for Tax Deferred Annuity and Deferred Compensation Plans v. Norris, 463 U. S. 1073, 1109 (1983) (O'CONNOR, J., concurring). Reading the Davis opinion alone, one might get the impression that it did not announce a new rule even of that variety. The opinion's emphatic language suggests that the outcome was not even debatable. See ante, at 111 (KENNEDY, J., concurring in part and concurring in judgment). In my view, however, assertive language is not itself determinative. As THE CHIEF JUSTICE explained for the Court in a different context:"[T]he fact that a court says that its decision… is 'controlled' by a prior decision, is not conclusive for purposes of deciding whether the current decision is a 'new rule'…. Courts frequently view their decisions as being 'controlled' or 'governed' by prior opinions even when aware of reasonable contrary conclusions reached by other courts." Butler v. McKellar, 494 U. S. 407, 415 (1990).
In Butler, we determined that the rule announced in Arizona v. Roberson, 486 U. S. 675 (1988), was "new" for purposes of Teague v. Lane, 489 U. S. 288 (1989), despite Roberson's repeated assertions that its rule was "directly controlled" by precedent. Indeed, we did not even feel bound by the opinion's statement that it was not announcing a new rule at all but rather declining to create an exception to an existing rule. While Teague and its progeny may not provide the appropriate standard of novelty for Chevron Oil purposes, their teaching-that whether an opinion is new depends not on its language or tone but on the legal landscape from which it arose-obtains nonetheless.
In any event, JUSTICE STEVENS certainly thought that Davis announced a new rule. In fact, he thought that the rule was not only unprecedented, but wrong: "The Court's holding is not supported by the rationale for the intergovernmental immunity doctrine and is not compelled by our previous decisions. I cannot join the unjustified, court-imposed restriction on a State's power to administer its own affairs." 489 U. S., at 818-819 (dissenting opinion). And just last Term two Members of this Court expressed their disagreement with the decision in Davis, labeling its application of the doctrine of intergovernmental immunity "perverse." Barker v. Kansas, 503 U. S. 594, 606 (1992) (STEVENS, J., joined by THOMAS, J., concurring). Although I would not call our decision in Davis perverse, I agree that its rule was sufficiently debatable in advance as to fall short of being "clearly foreshadowed." The great weight of authority is in accord.*
*Swanson v. Powers, 937 F.2d 965, 968, 970, 971 (CA4 1991) ("The most pertinent judicial decisions" were contrary to a holding of immunity and "the rationale behind the precedent might have suggested a different re In fact, before Davis was announced, conventional wisdom seemed to be directly to the contrary. One would think that, if Davis was "clearly foreshadowed," some taxpayer might have made the intergovernmental immunity argument before. No one had. Twenty-three States had taxation schemes just like the one at issue in Davis; and some of those schemes were established as much as half a century before Davis was decided. See Harper v. Virginia Dept. of Taxation, 241 Va. 232, 237, 401 S. E. 2d 868, 871 (1991). Yet not a single taxpayer ever challenged one of those schemes on intergovernmental immunity grounds until Davis challenged Michigan's in 1984. If Justice Holmes is correct that "[t]he prophecies of what the courts will do in fact, and nothing more pretentious" are "law," O. Holmes, The Path of the Law, in Collected Legal Papers 167, 173 (1920), then surely Davis announced new law; the universal "prophecy" before Davis seemed to be that such taxation schemes were valid.
An examination of the decision in Davis and its predecessors reveals that Davis was anything but clearly foreshadowed. Of course, it was well established long before Davis that the nondiscrimination principle of 4 U. S. C. § 111 and the doctrine of intergovernmental immunity prohibit a State from imposing a discriminatory tax on the United States or
suIt in [Davis itself]"; "how the intergovernmental tax immunity doctrine and 4 U. S. C. § 111 applied to [plans like the one at issue in Davis] was anything but clearly established prior to Davis"); Harper v. Virginia Dept. of Taxation, 241 Va. 232, 238, 401 S. E. 2d 868, 872 (1991) ("[T]he Davis decision established a new rule of law by deciding an issue of first impression whose resolution was not clearly foreshadowed"); Swanson v. State, 329 N. C. 576, 583, 407 S. E. 2d 791, 794 (1991) ("[T]he decision of Davis was not clearly foreshadowed"); Bass v. State, 302 S. C. 250, 256, 395 S. E. 2d 171, 174 (1990) (Davis "established a new principle of law"); Bohn v. Waddell, 164 Ariz. 74, 92, 790 P. 2d 772, 790 (1990) (Davis "established a new principle of law"); Note, Rejection of the "Similarly Situated Taxpayer" Rationale: Davis v. Michigan Department of Treasury, 43 Tax Lawyer 431, 441 (1990) ("The majority in Davis rejected a long-standing doctrine"). those who do business with it. The income tax at issue in Davis, however, did not appear discriminatory on its face. Like the Virginia income tax at issue here, it did not single out federal employees or retirees for disfavored treatment. Instead, federal retirees were treated identically to all other retirees, with a single and numerically insignificant exception-retirees whose retirement benefits were paid by the State. Whether such an exception rendered the tax "discriminatory" within the meaning of the intergovernmental immunity doctrine, it seems to me, was an open question. On the one hand, the tax scheme did distinguish between federal retirees and state retirees: The former were required to pay state taxes on their retirement income, while the latter were not. But it was far from clear that such was the proper comparison. In fact, there were strong arguments that it was not.
As JUSTICE STEVENS explained more thoroughly in his Davis dissent, 489 U. S., at 819-and as we have recognized since McCulloch v. Maryland, 4 Wheat. 316 (1819)-intergovernmental immunity is necessary to prevent the States from interfering with federal interests through taxation. Because the National Government has no recourse to the state ballot box, it has only a limited ability to protect itself against excessive state taxes. But the risk of excessive taxation of federal interests is eliminated, and "[a] 'political check' is provided, when a state tax falls" not only on the Federal Government, but also "on a significant group of state citizens who can be counted upon to use their votes to keep the State from raising the tax excessively, and thus placing an unfair burden on the Federal Government." Washington v. United States, 460 U. S. 536, 545 (1983) (emphasis added). Accord, United States v. County of Fresno, 429 U. S. 452, 462-464 (1977); South Carolina v. Baker, 485 U. S. 505, 526, n. 15 (1988).
There can be no doubt that the taxation scheme at issue in Davis and the one employed by the Commonwealth of Virginia provided that necessary "political check." They exempted only a small group of citizens, state retirees, while subjecting the remainder of their citizens-federal retirees, retirees who receive income from private sources, and nonretirees alike-to a uniform income tax. As a result, any attempt to increase income taxes excessively so as to interfere with federal interests would have caused the similarly taxed populace to "use their votes" to protect their interests, thereby protecting the interests of the Federal Government as well. There being no risk of abusive taxation of the National Government, there was a good argument that there should have been no intergovernmental immunity problem either. See Davis, 489 U. S., at 821-824 (STEVENS, J., dissenting).
In addition, distinguishing between taxation of state retirees and all others, including private and federal retirees, was justifiable from an economic standpoint. The State, after all, does not merely collect taxes from its retirees; it pays their benefits as well. As a result, it makes no difference to the State or the retirees whether the State increases state retirement benefits in an amount sufficient to cover taxes it imposes, or whether the State offers reduced benefits and makes them tax free. The net income level of the retirees and the impact on the state fisc is the same. Thus, the Michigan Department of Taxation had a good argument that its differential treatment of state and federal retirees was "directly related to, and justified by, [a] significant differenc[e] between the two classes," id., at 816 (internal quotation marks omitted): Taxing federal retirees enhances the State's fisc, whereas taxing state retirees does not.
I recite these arguments not to show that the decision in Davis was wrong-I joined the opinion then and remain of the view that it was correct-but instead to point out that the arguments on the other side were substantial. Of course, the Court was able to "ancho[r] its decision in precedent," ante, at 112 (KENNEDY, J., concurring in part and con curring in judgment). But surely that cannot be dispositive. Few decisions are so novel that there is no precedent to which they may be moored. What is determinative is that the decision was "sufficiently debatable" ex ante that, under Chevron Oil, nonretroactivity cannot be precluded. Arizona Governing Committee v. Norris,