concurring in part and dissenting in part.
Although I agree with my colleagues that Greene’s claim is not. procedurally *107defaulted and join Part II of the majority opinion in full, I respectfully disagree with their determination of the controlling date for “clearly established Federal law” under 28 U.S.C. § 2254(d)(1). As my colleagues recognize, the authority on this question is conflicting and, save for a First Circuit Court opinion, unreasoned. But choosing the date of the relevant state-court decision, as our Court does today, leaves a twilight zone between the cutoff set by the majority here and the retroactivity analysis of the Supreme Court’s decisions in Griffith v. Kentucky, 479 U.S. 314, 107 S.Ct. 708, 93 L.Ed.2d 649 (1987), and Teague v. Lane, 489 U.S. 288, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989) (plurality).1 The consequence of the majority’s opinion today is that a criminal defendant who is denied the light under Griffith to apply a new constitutional rule to his or her case on direct appeal is left without later recourse to federal habeas review to correct that error.
I. Background
To set the stage here, Greene argued prior to his state trial that proceeding jointly against him and his co-defendants would prejudice his defense and his constitutional right to confront witnesses “when the Commonwealth offer[ed] into evidence [out-of-court] statements made by the co-defendants.” He cited for support the Supreme Court cases then known — Bruton v. United States, 391 U.S. 123, 88 S.Ct. 1620, 20 L.Ed.2d 476 (1968), and Richardson v. Marsh, 481 U.S. 200, 107 S.Ct. 1702, 95 L.Ed.2d 176 (1987). The motions judge, recognizing the prejudice of those statements fingering Greene, thought she could cancel out that prejudice by simply redacting Greene’s name when the statements were read to the jury and giving a limiting instruction to the jury.
Bruton (as clarified in Richardson) was not enough to win the point for Greene. He remained short of the winning line on his appeal when the Pennsylvania Superior Court ruled on December 16, 1997, that Bruton was not violated.
But there was hope, as Greene filed timely a petition for allowance of review to the Pennsylvania Supreme Court. That hope received a big boost when the Supreme Court of the United States decided Gray v. Maryland, 523 U.S. 185, 118 S.Ct. 1151, 140 L.Ed.2d 294 (1998), while his petition was pending. Gray held that redacting names from confessions using obvious blanks falls within the class of claims protected by Bruton. This was good news indeed for Greene. The Pennsylvania Supreme Court would no doubt take note of Gray and grant review.
Indeed it did, and it did so on the very issue decided by Gray. Yet for reasons we do not know, it abruptly dismissed its grant of appeal as “having been improvidently granted,” in effect leaving in place post-Gray the Superior Court’s pre-Grow/ decision on Greene’s Confrontation Clause rights.
Greene filed a petition seeking collateral relief under Pennsylvania’s Post-Conviction Relief Act (“PCRA”), 42 Pa. Cons. Stat. Ann. §§ 9541-46. He did not allege that his Confrontation Clause rights were violated (and thus did not cite Bruton or Gray), ostensibly because the PCRA proscribes relitigating matters already dealt with on direct appeal. 42 Pa. Cons.Stat. Ann. § 9543(a)(3). Nonetheless, my colleagues state that, had he done so, Greene *108would have “obtained a later, post-Gray state-court decision on the merits.” Maj. Op. at 103-04. They claim — -incorrectly, I believe, for the reasons stated more fully in note 13 below — that failure to raise Gray in his PCRA petition “shrank the universe of ‘clearly established Federal law’ available to him for his § 2254 petition.” Id. at 103. They say this while conceding that “Greene fairly presented the factual and legal substance of his Confrontation Clause claim to the Pennsylvania state courts.” Id. at 93. We do not consider federal claims procedurally defaulted, “even if the procedural rule is theoretically applicable to [the] facts,” unless the last state court rendering a judgment in the case “clearly and expressly states that its judgment rests on a state procedural bar.” Holloway v. Horn, 355 F.3d 707, 714 (3d Cir.2004) (citations omitted). There is no clear or express statement here.
Greene’s frustrating failure with the Pennsylvania court system was over, but all was not lost. He thought he could seek habeas review of his Confrontation Clause rights in a federal court. And he did.
This is where we come in after the District Court ruled against Greene: were his Confrontation Clause rights “clearly established [as] Federal law” when they were “adjudicated on the merits in State court proceedings”? 28 U.S.C. § 2254(d)(1). The answer is yes if we look to all United States Supreme Court decisions before his conviction became final,2 but no if we stop with the decision of the Pennsylvania Superior Court less than two and a half months before the Gray decision. Greene is in the unwelcome twilight zone where, in the United States Supreme Court’s own words, “uncertainty” currently exists. Smith v. Spisak, 558 U.S. --, 130 S.Ct. 676, 681, - L.Ed.2d - (2010).
II. Analysis
This is not a situation where Greene is seeking to take belated advantage of a rule to which he is not entitled. He is asking us to apply a case that should have been applied on direct review. Under the Supreme Court’s Griffith jurisprudence, he was entitled to the benefit of Gray. It is only because the Pennsylvania state courts failed to apply it to his case that we are evaluating it in the first instance on habeas review.
My analysis differs from that of the majority. In a nutshell, subsection 2254(d)(1) does not choose any cutoff date. Thus, we are left with the retroactivity jurisprudence of Griffith and Teague. Because Gray was decided prior to the date Greene’s conviction became final, I believe Griffith requires its application to this case. I would therefore reverse the judgment of the District Court and remand for consideration of Gray.3
*109A. As noted, the question of whether § 2254(d)(1) sets a cutoff date is unresolved.
I agree with the majority that “clearly established Federal law” did not have any special meaning prior to AEDPA and the text of 28 U.S.C. § 2254(d)(1). See Maj. Op. at 98 n. 10. Nor does the text of 28 U.S.C. § 2254(d)(1) have an express time cutoff for “clearly established Federal law.” My colleagues read the statute implicitly to require that any Supreme Court decision handed down after the relevant state-court decision on the merits is to be ignored for purposes of habeas relief. I disagree that “[rjeading the language plainly,” or in a “straightforward” way, as my colleagues suggest, requires that the Supreme Court decision exist at the time of the state court’s substantive resolution. If § 2254(d)(1) were so plain or straightforward, why does the Supreme Court say it is uncertain? And why are my colleagues of the view that “there is no clear answer to the issue we face”? Maj. Op. at 96 n. 7. In the face of such uncertainty, I find it difficult to conclude that there is a “natural reading” of § 2254(d)(1) dictating a cutoff date.4
A primary reason for the Supreme Court’s uncertainty as to whether the text of § 2254(d)(1) provides a clear cutoff date is its own decision in Williams v. Taylor, 529 U.S. 362, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000). Inadvertently (no doubt), the Court had two different majorities identifying two different cutoffs.5 Justice Stevens, writing for six members of the Court in Part III of his opinion, stated that the applicable date for purposes of determining whether federal law is established is “the time [the habeas petitioner’s] state-court conviction became final.” 529 U.S. at 390, 120 S.Ct. 1495 (Stevens, J., for the Court). Justice O’Connor, writing for five members of the Court in Part II of her opinion, stated that “ ‘clearly established Federal lavd ... refers to the holdings, as opposed to the dicta, of [the Supreme] Court’s decisions as of the time of the relevant state-court decision.” Id. at 412, 120 S.Ct. 1495 (O’Connor, J., for the Court). Neither Justice Stevens nor Justice O’Connor appears to have chosen a cutoff based on the text of the statute, and they did not acknowledge the discrepancy in their respective opinions. Indeed, in Williams the choice of cutoff would not have mattered because the case focused on Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), a case decided prior to both the 1985 crime and the 1986 conviction in Williams, making the discussion of cutoff dicta because under both cutoffs Strickland was undoubtedly “clearly established Federal law.”
Our task is to reconcile the conflicting majorities in Williams regarding the cutoff for “clearly established Federal law” under AEDPA while maintaining consistency with the Court’s controlling decisions in Griffith and Teague. Recently, the Supreme Court recognized the “uncertainty” in temporal cutoff for “clearly established *110Federal law,” and declined to resolve it at that time. See Spisak, 130 S.Ct. at 681.6
Recognizing the Court’s statement in Spisak as the “most telling observation regarding the use of the date the conviction became final,” my colleagues dismiss it in a single sentence as “mere uncertainty [that] cannot counterbalance” the cases that select the date of the relevant state-court decision. Maj. Op. at 99. Moreover, they do so even though we agree that Supreme Court has never conducted a thorough analysis of the “clearly established” cutoff for AEDPA. Id. at 96-97. Though the Supreme Court has used the relevant state-court decision as the temporal cutoff in cases after Williams, I do not find this dispositive. Like our own checkered jurisprudence, it is not clear from the Supreme Court’s cases whether it recognized these divergent approaches inasmuch as it was not required in those cases to resolve whether the cutoff date was the relevant state-court decision date or the date the conviction became final.7
If, as the majority suggests, the clear answer is the date of the relevant state-court decision, the Spisak Court would not have noted the uncertainty, nor would it have assumed the date of finality. The Court is not in the business of offering advisory opinions, and if it were clear that its prior cases had selected the date of the relevant state-court decision, it would not have issued the opinion in Spisak. It would have held instead that, because Mills was decided after the final state-court decision on the merits, AEDPA did not permit consideration of the case. It would have stopped its analysis there instead of going on at great length to evaluate the Mills claim on the merits. Thus, post -Williams Supreme Court precedent offers little to clarify the temporal cutoff for “clearly established Federal law” under AEDPA.
B. The Supreme Court has not abandoned its retroactivity jurisprudence post-AEDPA.
AEDPA’s concern over whether a state court ruling in a criminal case was contrary to, or an unreasonable application of, *111“clearly established Federal law” stems from the desire to avoid disturbing final criminal judgments through collateral review. In particular, the use of the past tense (“established”) means that AEDPA is concerned with the law that should have been applied at the time of the state court proceedings. Where I diverge from my colleagues is how we determine what that body of law is.
Even though at first it seems conceptually difficult to say that a court unreasonably applied Supreme Court precedent that did not yet exist, retroactivity analysis becomes the tool for deciding. When a Supreme Court holding is retroactively applied to a prior proceeding, it is as if it existed at the time of that prior proceeding. The majority’s view ignores controlling Supreme Court precedent that allows, in certain circumstances, for the retroactive application of constitutional rules to criminal cases even though they are announced after a state court ruling on the merits.
Paramount to understanding the Supreme Court’s retroactivity jurisprudence is discerning its decisions in Griffith and Teague. They provide a distinction between “old rules” and “new rules,” terms that have a clear meaning only when used in relation to a given criminal conviction. Unhelpfully, the Supreme Court has used “new rule” to mean different things in the Griffith and Teague context.
A “new rule” for Griffith is one that is announced after a state court ruling on the merits. A “new rule” for Teague is one that is announced after a conviction becomes final. This means that in the application of Gray to Greene’s conviction, Gray is a “new rule” for Griffith purposes but is an “old rule” for Teague purposes. We are principally concerned with the Teague distinction between “old rules” and “new rules.” The following table may aid in understanding the discussion that follows.
[[Image here]]
*112C. The Supreme Court has a developed jurisprudence governing the application to cases on collateral review of its cases decided prefinality and those decided post-finality.
The retroactive application of newly announced constitutional rules in criminal cases has long troubled the Supreme Court. As noted, retroactivity takes that rule and transports it back in time to a proceeding that pre-dated the announcement of the rule, treating the rule as if it existed at the time of the prior proceeding. Because this fiction has the potential to upset settled proceedings, especially in the criminal context, over the years the Court came to adopt a bright line that splits the application of these rules into two domains of review.
Whether a new rule applies retroactively depends on whether a criminal conviction is on direct review or collateral review at the time of the Supreme Court decision announcing the new rule. If the conviction is on direct review when the new rule is announced, Griffith allows the retroactive application of the new rule to all criminal cases pending on direct review as a “basic norm[ ] of constitutional adjudication.” 479 U.S. at 322, 107 S.Ct. 708. If the conviction is on collateral review when the new rule is announced (ie., convictions that became final before the new rule is announced), Teague restricts the application of that new rule to narrow exceptions discussed below. This bright-line distinction was made due to the differing considerations between the two domains of review.
1. Griffith
The Griffith Court held that the “failure to apply a newly declared constitutional rule to criminal cases pending on direct review violates basic norms of constitutional adjudication.” 479 U.S. at 322, 107 S.Ct. 708. It was the very “integrity of judicial review” that required application of a new constitutional rule “to all similar cases pending on direct review.” Id. at 323, 107 S.Ct. 708. Two principles guided this decision. First, the Court recognized that
[a]s a practical matter, of course, we cannot hear each case pending on direct review and apply the new rule. But we fulfill our judicial responsibility by instructing the lower courts to apply the new rule retroactively to cases not yet final. Thus, it is the nature of judicial review that precludes us from “[sjimply fishing one case from the stream of appellate review, using it as a vehicle for pronouncing new constitutional standards, and then permitting a stream of similar cases subsequently to flow by unaffected by that new rule.”
Id. (citation omitted). Second, the Court recognized that
selective application of new rules violates the principle of treating similarly situated defendants the same. As we pointed out in United States v. Johnson, [457 U.S. 537, 102 S.Ct. 2579, 73 L.Ed.2d 202 (1982) ] the problem with not applying new rules to cases pending on direct review is “the actual inequity that results when the Court chooses which of many similarly situated defendants should be the chance beneficiary” of a new rule. Although the Court had tolerated this inequity for a time by not applying new rules retroactively to cases on direct review, we noted: “The time for toleration has come to an end.”
Id. (citations omitted) (emphasis in original). The Court therefore held “that a new rule for the conduct of criminal prosecutions is to be applied retroactively to all *113cases, state or federal, pending on direct review or not yet final, with no exception for cases in which the new rule constitutes a ‘clear break’ with the past.” Id. at 828, 107 S.Ct. 708. I note that “pending on direct review” is slightly different from “not yet final.” A case that has already exhausted the direct appeal as of right resulting in a state-court decision on the merits, but is not yet final, is still within the purview of Griffith. Finality is the key date.
The Griffith Court “instructed] the lower courts,” state and federal, “to apply the new rule retroactively to cases not yet final.” Id. at 323, 107 S.Ct. 708. It did not merely advise those courts to consider applying the rule subject to their discretion, but mandated application of the new rule. It was only through this mandate that “actual inequity” between “many similarly situated defendants” would be avoided.8
2. Teague
In Teague, the Supreme Court dealt with the other side of the retroactivity question. Collateral attacks such as habeas cotpus are not meant to be a substitute for direct review, and the Court has recognized an interest in leaving concluded litigation in a state of repose. 489 U.S. at 306, 109 S.Ct. 1060 (quoting Mackey v. United States, 401 U.S. 667, 682-83, 91 S.Ct. 1160, 28 L.Ed.2d 404 (1971) (Harlan, J., concurring in judgments in part and dissenting in part)). Quoting the second Justice Harlan, the Court noted that it was “ ‘sounder, in adjudicating habeas petitions, generally to apply the law prevailing at the time a conviction became final than it is to seek to dispose of [habeas ] cases on the basis of intervening changes in constitutional interpretation.’ ” Id. (quoting Mackey, 401 U.S. at 689, 91 S.Ct. 1171 (Harlan, J.) (alteration in original)). The Court identified only two exceptions to the general prohibition against the retroactive application of new post-finality rules to cases on collateral review: (1) new rules that place certain kinds of primary, private individual conduct beyond the power of the criminal law-making authority to proscribe, id. at 307, 109 S.Ct. 1060; and (2) new “watershed rules of criminal procedure ... [that] ‘alter our understanding of the bedrock procedural elements that must be found to vitiate the fairness of a particular conviction,’ ” id. at 311, 91 S.Ct. 1171 (emphasis in original) (citation omitted).
In deciding Griffith and Teague, the Supreme Court has carefully set out the different concerns in the pre-finality (direct appeal) and post-finality (collateral attack) application of new rules. In the context of retroactivity for federal habeas review, the Teague Court focused on the distinction between intermediate judgments subject to appeal and final judgments subject only to collateral attack:
Application of constitutional rules not in existence at the time a conviction became final seriously undermines the principle of finality which is essential to the operation of our criminal justice system. Without finality, the criminal law is deprived of much of its deterrent effect. The fact that life and liberty are *114at stake in criminal prosecutions “shows only that ‘conventional notions of finality’ should not have as much place in criminal as in civil litigation, not that they should have none.”
Id. at 309, 109 S.Ct. 1060 (emphases in original) (citation omitted). With this view of finality, the Court held that “[u]nless they fall within an exception to the general rule, new constitutional rules of criminal procedure will not be applicable to those cases which have become final before the new rules are announced,” using finality, not the date of the relevant state-court decision, as the inflection point between Griffith and Teague. Id. at 310, 109 S.Ct. 1060.
D. Section 2254(d)(1) does not discard Griffith and Teague.
In a unanimous post-AEDPA and post-Williams decision, Whorton v. Bockting, 549 U.S. 406, 127 S.Ct. 1173, 167 L.Ed.2d 1 (2007), the Supreme Court held that Griffith and Teague
laid out the framework to be used in determining whether a rule announced in one of [the Court’s] opinions should be applied retroactively to judgments in criminal cases that are already final on direct review. Under the Teague framework, an old rule applies both on direct and collateral review, but a new rule is generally applicable only to cases that are still on direct review. See Griffith, 479 U.S. 314 [107 S.Ct. 708].9 A new *115rule applies retroactively in a collateral proceeding only if [the Teague requirements are met],
549 U.S. at 416, 127 S.Ct. 1173.10 Though Whorton dealt with an application of Teague, it explicitly -recognized that Griffith requires that “old rules” be applied both on direct and collateral review.11 To me this means that the Whorton Court unanimously endorsed Griffith and the idea that a Supreme Court decision handed down after the last state court ruling on the merits, but before finality, is an old rule that is applicable even under AEDPA and even if it is not a “watershed” ruling or does not place conduct beyond the power of the state to proscribe.
Given the Court’s retroactivity concerns, I believe the better reading of § 2254(d)(1) is that it does not set a definitive cutoff date for “clearly established Federal law.” It is the Supreme Court’s retroactivity jurisprudence of Griffith or Teague that determines applicability on collateral review, not AEDPA.
My colleagues’ reading of § 2254(d)(1) conflicts with Whorton. They refuse to include all “old rules” as “clearly established Federal law.” This reading contradicts the unanimous holding in Williams that all “old rules” for Teague purposes are “clearly established Federal law.”12 My colleagues recognize this contradiction, but they choose to ignore Griffith and Teague and adopt Justice O’Connor’s initial unreasoned declaration (that chose the date of the relevant state-court decision and cited no case) and not her later reasoned one (that referred to “old rules” under Teague and cited Supreme Court precedent). See 529 U.S. at 412, 120 S.Ct. 1495 (citing Stringer v. Black, 503 U.S. 222, 228, 112 S.Ct. 1130, 117 L.Ed.2d 367 (1992)). It is unclear to me why we would choose her statement of the law (a dictum, no less) in conflict with the Supreme Court’s decisions in Griffith and Teague instead of her statement of the law in harmony with those Supreme Court holdings and Whorton (and that actually invokes the controlling Supreme Court precedent of Teague).
E. The majority’s cutoff creates a twilight zone
If the relevant cutoff date is the date of the last state-court decision on the merits, *116we would create a twilight zone for criminal defendants. Consider the possible times relative to a state court conviction when a decision by the 'Supreme Court is announced: (1) prior to the last state-court decision on the merits; (2) between the last state-court decision on the merits and finality; and (3) after the conviction is final. If it were decided in the first period (prior to the last state-court decision on the merits), a state court would have to apply it to be consistent with Griffith. If it were decided in the third period (after finality), habeas relief would be available as a “new rule” under Teague if the decision announced a “watershed” rule or placed certain conduct beyond the power of the state to proscribe. However, if it were decided in the second period (the twilight zone between the last state-court decision on the merits and before finality), the majority’s time cutoff would nonetheless consider it not to be “clearly established Federal law” and would bar habeas relief because the rule did not exist at the time of the last state-court decision on the merits. The majority reaches this conclusion even though, as discussed above, a rule announced pre-finality is an “old rule” for Teague purposes and Griffith requires its application on direct and collateral review.13
*117So inflexible is the “plain reading” the majority adopts that even “new rules” that pass the Teague test for retroactive application would not entitle a petitioner to habeas relief. “New rules” for Teague purposes are always decided after the date of the relevant state-court decision, as they come into being after finality. Yet the majority would not consider the “new rule” to be “clearly established Federal law” because the “new rule” did not yet exist, and no relief could be granted.14 Such a Catch-22 reading of § 2254(d)(1) effectively disregards Griffith and Teague even as the Supreme Court has maintained that both decisions remain viable.15
As discussed above, even though at first it seems conceptually difficult to say that a state court unreasonably applied Supreme Court precedent that did not yet exist, the Supreme Court’s retroactivity analysis *118treats the precedent as if it existed at the time of that prior state court-proceeding. Under Chiffith, Supreme Court decisions are retroactively applied to those convictions not yet final at the time of the decision. Furthermore, if the state court neglects to apply the rule retroactively to convictions not yet final, this can be still corrected after finality on collateral review. See Whorton, 549 U.S. at 416, 127 S.Ct. 1173 (“[O]ld rule[s] appl[y] both on direct and collateral review.” (emphasis added)). Under Teague, Supreme Court decisions are retroactively applicable even to convictions that were already final at the time of the decision if it announces a “watershed” rule or places certain conduct beyond the power of the state to proscribe. We know from Whorton that § 2254(d)(1) does not overrule Griffith and Teague, but by deeming irrelevant any case that postdates the relevant state-court decision, the majority implicitly disregards both Griffith and Teague.
While another Circuit Court has rejected the majority’s cutoff on fears of the potential for “state court ... subversion] ... by the simple expedient of summarily affirming a lower court’s decision,” Foxworth v. St. Amand, 570 F.3d 414, 432 (1st Cir.2009), its reasoning does not depend on a distrust of the judicial integrity of state courts. A well-meaning state court system could innocently neglect to apply Griffith after its final decision on the merits, but before the conviction becomes final. If a state court were to ignore the mandate to apply the new rule to all cases still pending on direct appeal or not yet final, it would similarly undermine the integrity of judicial review. That would leave collateral review by habeas corpus as the only remedy to correct the mistake. Surely a criminal defendant is entitled to recourse if the state courts simply forget to check for new, relevant Supreme Court precedent prior to finality.16 This helps to avoid the situation where similarly situated defendants receive disparate treatment based on the happenstance of state court attention (or inattention).17
Yet, under the majority’s selection of temporal cutoff, even that remedy would be foreclosed whenever the state courts declined to apply the rule without explanation. This would leave affected habeas petitioners as unfairly treated relative to othér similarly situated individuals who were lucky enough to have the state courts apply the new rule.
‡ ‡ ‡ ‡
It is not our place to second-guess the Supreme Court when it has held that: (1) Supreme Court decisions handed down pri- or to finality must be applied on both direct and collateral review under Griffith; *119(2) Teague and Griffith have continuing vitality after AEDPA; (3) all nine Justices in Williams agreed that an “old rule” under Teague qualifies as “clearly established Federal law”; and (4) its decisions since Williams have not definitively set a temporal cutoff. In the absence of an express statement to the contrary by the Supreme Court (and there is none), we are bound to apply the clearly expressed (and still controlling) jurisprudence of Griffith and Teague. The Court may wish, in the AEDPA context, to cut back on Griffith and Teague, but it, not us, possesses the power to overrule its precedent.
I would hold that the cutoff date for “clearly established Federal law” is not prescribed by 28 U.S.C. § 2254(d)(1). The retroactive application of constitutional rules to criminal cases is governed by Griffith and Teague, and I would look first to whether Gmy was decided before or after finality to determine which rule applies. As here Gray was decided prior to finality, the Pennsylvania Supreme Court should have considered it in the course of fulfilling its responsibilities under Griffith. When it did not do so, the District Court on habeas review needed to correct this failure to consider Gray. Accordingly, I would vacate its judgment and remand for application of Grray to Greene’s Confrontation Clause claim. For these reasons, I respectfully dissent from all but Part II of the majority opinion.
. "Although Teague was a plurality opinion that drew support from only four members of the Court, the Teague rule was affirmed and applied by a majority of the Court shortly thereafter.” Danforth v. Minnesota, 552 U.S. 264, 266 n. 1, 128 S.Ct. 1029, 169 L.Ed.2d 859 (2008) (citation omitted).
. Finality means that "a judgment of conviction has been rendered, the availability of appeal exhausted, and the time for a petition for certiorari elapsed or a petition for certiorari finally denied.” Griffith, 479 U.S. at 321 n. 6, 107 S.Ct. 708.
. Directing the District Court to consider Gray does not mean that Greene would get habeas relief. The Pennsylvania state courts may have applied Pennsylvania’s own existing Confrontation Clause jurisprudence that was not contrary to, or an unreasonable application. of, the constitutional rule announced in Gray even though Gray did not yet exist. AEDPA does not allow a writ of habeas corpus to be granted for simple errors of law; a writ is granted only when there is an unreasonable error of law. See Williams v. Taylor, 529 U.S. 362, 412, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000) (O’Connor, J., for the Court) ("[A]n unreasonable application of federal law is different from an incorrect or eironeous application of federal law.” (emphasis in original)).
. One might think this is the paradigm for invoking the rule of lenity. If even the Supreme Court is uncertain which time trigger is in play as to when federal law is "clearly established," it seems fair that, while others may in the future lose in the twilight zone, Greene should not. As that is not how the majority rules, Greene’s claim to Gray's benefits becomes simply sisyphean.
. The fractured decision had Justice Stevens announcing the judgment of the Court and delivering the opinion of the Court (6 votes) with respect to Parts I, III, and IV of his opinion, and a minority opinion (4 votes) with respect to Parts II and V. Justice O'Connor delivered the opinion of the Court (5 votes) with respect to Part II of her opinion (save for a footnote).
. Like this case, the timeline of Spisak has a "new rule” falling in the twilight zone between the last state-court decision on the merits and the date of finality. There, the last state-court decision on the merits was issued on April 13, 1988. State v. Spisak, 36 Ohio St.3d 80, 521 N.E.2d 800 (1988); see also Spisak v. Mitchell, 465 F.3d 684, 688 (6th Cir.2006). The conviction became final on March 6, 1989, when the Supreme Court of the United States denied certiorari. Spisak v. Ohio, 489 U.S. 1071, 109 S.Ct. 1354, 103 L.Ed.2d 822 (1989). Mills v. Maryland was decided on June 6, 1988, in the period between the last state-court decision on the merits and the date of finality. 486 U.S. 367, 108 S.Ct. 1860, 100 L.Ed.2d 384 (1988).
. The majority makes this observation with respect to our own precedent, yet does not reach the same conclusion with respect to the Supreme Court's rulings. In none of the cases cited by the majority as adopting the date of the relevant state-court decision was the Supreme Court required to determine the cutoff date, and in the only case where it had to confront the issue, Spisak, it sidestepped. We are not free to ignore that the Court itself has told us the law is "uncertain.” I caution that I do not take "mere uncertainty” alone to "counterbalance the numerous Supreme Court decisions,” but instead take the Supreme Court’s express reservation of the question in Spisak as telling us that the issue remains unresolved and that there is little to be gleaned from other cases that, as the majority notes, take one approach "without analysis.” Maj. Op. at 99. In that respect, the Supreme Court's § 2254(d)(1) precedent is like our precedent: it goes both ways and is unhelpful in resolving an inquiry that the decisions did not squarely address. Moreover, the selection of a cutoff date in Supreme Court cases is no more than dicta.
. For instance, it would undermine the integrity of judicial review if among ten similarly situated criminal defendants — all duly convicted in state court, all having only the petition for certiorari to the Supreme Court of the United States as their remaining recourse on direct appeal, and all raising the same question of constitutional law — only one were permitted to benefit from a new rule, leaving the other nine to "flow by unaffected by that new rule.” Griffith, 479 U.S. at 323, 107 S.Ct. 708 n. 6 (citation omitted). It is to avoid this inequality between a lucky defendant and an unlucky (but similarly situated) defendant that Griffith draws its bright line and requires application of the new rule.
. The Supreme Court here meant "old rule” (pre-finality) and "new rule” (post-finality) for Teague purposes. I disagree with my colleagues when they state that "[t]he sentence following the Griffith citation in the Whorton decision further confirms [their] understanding by explaining how new rules apply in collateral proceedings through citation to Teague, not Griffith.” Maj. Op. at 101. This case is not about "extending” Griffith to collateral review or "import[ing]” Griffith into Teague. This case is not even about "new rules” — it is about "old rules,” meaning all pre-finality Supreme Court precedent. The majority takes issue with applying "old rules” to collateral review, but if "old rules” do not apply on collateral review, which rules do? Griffith defines the body of law that should be applied on direct review, that is, all precedent that is decided prior to a conviction’s finality. This relates to collateral review because habeas is concerned with what body of law should have been applied on direct review. The relation of Griffith to collateral review is nothing novel or profound.
As my colleagues concede, Maj. Op. at 100, a rule that is handed down pre-finality is an old rule and need not pass the Teague test. Instead, it is governed by Griffith and is applicable "both on direct and collateral review.” Whorton, 549 U.S. at 416, 127 S.Ct. 1173. Furthermore, the Supreme Court has applied Griffith’s retroactivity principle in the habeas context, giving habeas petitioners the rights established by all Supreme Court decisions that pre-date a conviction’s finality:
Penry’s conviction became final on January 13, 1986, when this Court denied his petition for certiorari on direct review of his conviction and sentence. This Court’s decisions in Lockett v. Ohio [438 U.S. 586, 98 S.Ct. 2954, 57 L.Ed.2d 973 (1978)] and Eddings v. Oklahoma [455 U.S. 104, 102 S.Ct. 869, 71 L.Ed.2d 1 (1982)] were rendered before his conviction became final. Under the retroactivity principles adopted in Griffith v. Kentucky, Penry is entitled to the benefit of those decisions.
Penry v. Lynaugh, 492 U.S. 302, 314-15, 109 S.Ct. 2934, 106 L.Ed.2d 256 (1989) (citations omitted), abrogated on other grounds by Atkins v. Virginia, 536 U.S. 304, 122 S.Ct. 2242, 153 L.Ed.2d 335 (2002).
We have likewise done so in our own habeas cases, applying pre-finality decisions on habeas review under the authority of Griffith. See, e.g. Lewis v. Horn, 581 F.3d 92, 102 n. 5 (3d Cir.2009) ("Although Abu-Jamal, like Lewis, was convicted prior to the Supreme Court’s decision in Batson, his direct appeal was still pending when the Supreme Court decided Batson, and therefore Batson applied retroactively to his case.” (citing Griffith, 479 U.S. at 328, 107 S.Ct. 708)); Simmons v. Beyer, 44 F.3d 1160, 1164-65 (3d Cir.1995) ("Had Simmons received a timely review, his conviction would have been final before 1986. In a sense, he is a 'chance beneficiary’ of the Batson rule ... [, but] we see no reason to *115bend the rule in Griffith to deny Simmons the constitutional protection afforded in Batson.” (citing Griffith, 479 U.S. at 323, 107 S.Ct. 708 n. 6)).
. In Whorton, the Court analyzed whether Crawford v. Washington, 541 U.S. 36, 124 S.Ct. 1354, 158 L.Ed.2d 177 (2004), a case that was decided after the habeas petitioner’s conviction was final, was applicable to his case as a “new rule.” Implicit in the analysis is the recognition that cases decided after the last state-court decision on the merits may nonetheless be applicable to cases under AEDPA so long as they meet the appropriate retroactivity test. Otherwise there would be no need to consider whether Crawford applied because it did not exist at the time of the last state-court decision on the merits.
. As noted above, an "old” rule is any rule that existed prior to finality or was dictated by the governing precedent at the time of finality-
. Notwithstanding the apparent conflict in time cutoff in Williams, Justice O’Connor and Justice Stevens did agree that “whatever would qualify as an old rule under [the Court’s] Teague jurisprudence will constitute 'clearly established Federal law, as determined by the Supreme Court of the United States’ under § 2254(d)(1).” 529 U.S. at 412, 120 S.Ct. 1495 (O’Connor, J., for the Court (joined by Rehnquist, Kennedy, Thomas, & Scalia, JJ.)); see also id. at 379-81, 120 S.Ct. 1495 (Stevens, J., 4-vote portion (joined by Souter, Ginsburg, & Breyer, JJ.)) (distinguishing “new rules” from “clearly established” rules). In other words, the court was unanimous in holding that all "old rules” under Teague constitute “clearly established Federal law” under AEDPA.
. As noted above, the majority states that Greene could have raised a Bruton Confrontation Clause claim on PCRA review, thereby receiving a post-Gray stale-court decision on the merits. Maj. Op. at 102-03 & n. 12. I believe this is not correct. The PCRA entitles state prisoners to relief only if they plead and prove an allegation of error that "has not been previously litigated or waived.” 42 Pa. Cons.Stat. Ann. § 9543(a)(3). Greene had previously litigated his Confrontation Clause claim on direct appeal and thus could not bring a PCRA action.
"[A]n issue has been previously litigated if ... the highest appellate court in which the petitioner could have had review as a matter of right has ruled on the merits of the issue.” Id. § 9544(a)(2). "Whether an issue was previously litigated turns on whether '[the issue] constitutes a discrete legal ground or merely an alternative theory in support of the same underlying issue that was raised on direct appeal.’ ” Commonwealth v. Small, 602 Pa. 425, 980 A.2d 549, 569 (2009) (alteration in original); see also Commonwealth v. Collins, 585 Pa. 45, 888 A.2d 564, 570 (2005). Furthermore, citing to different authority for the same issue "does not change the fact [that] th[e] issue was previously litigated. To hold otherwise would mean a PCRA petitioner could raise an issue on direct appeal, and raise it again in a PCRA petition merely by citing a different case to support the original theory.” Small, 980 A.2d at 569.
My colleagues and I agree that Greene has not procedurally defaulted his Bruton claim, having raised it on direct appeal. We agree that he litigated his Bruton claim in the Pennsylvania Superior Court. We agree that the Superior Court ruled on the merits of this claim. We agree that he raised it in his petition for allowance of appeal with the Pennsylvania Supreme Court. In other words, there can be no dispute that the Bruton claim was "previously litigated” for purposes of the PCRA. While my colleagues suggest that Gray was a discrete legal ground because it answered the question left open in Marsh regarding pronoun or symbol substitutions, the fact remains that the Pennsylvania Superior Court directly addressed the issue of pronoun substitutions in Greene’s direct appeal. App. 990 (Pa.Super.Ct.Op.) (citing Commonwealth v. Miles, 545 Pa. 500, 681 A.2d 1295, 1300 (1996)). Though Gray was decided later, the "discrete legal ground” was before the Superior Court and it decided that issue. In any event, to the extent that my colleagues believe that the Pennsylvania courts might have relaxed their PCRA standards to allow Greene to make the claim, that is pure speculation.
A similar situation arose in another Pennsylvania case cited by the Commonwealth. Commonwealth v. Washington, 592 Pa. 698, 927 A.2d 586, 608-09 (2007). There, Washington unsuccessfully raised a Bruton claim very similar to Greene's at trial and on direct appeal to the Pennsylvania Supreme Court. Subsequent to Washington's conviction becoming final, Gray was decided. On PCRA review, he invoked Gray, reasserted his Bruton claim, and argued that it was not previously litigated. The Pennsylvania Supreme *117Court rejected this argument and held the Bruton claim “previously litigated because [it] reviewed the claim and ruled on the merits of the issue” on direct appeal. Id. at 609. Likewise, Greene could not raise the claim in his PCRA petition as the majority suggests he should have done. He had pursued his color-able Bruton claim based on federal law "as far as possible in the state courts.” See Maj. Op. at 104.
Of course, the issue is not whether Washington is factually on point with Greene’s case, or whether it is theoretically possible that Greene could have raised a claim based on Gray on PCRA review, but whether the meaning of "clearly established Federal law” under § 2254(d)(1) means different things in different cases depending on the availability of state collateral review. What my colleagues essentially hold is that when a defendant has “previously litigated" a novel constitutional claim on direct appeal, but the Supreme Court recognizes the constitutional rule in a different case before his conviction becomes final, he has no federal habeas recourse if the state courts do not apply the constitutional rule to his case on direct review and the state collateral process prohibits him from re-raising the claim.
. Although another portion of AEDPA, 28 U.S.C. § 2244(d)(1)(C), resets the one-year period of limitation when the Supreme Court newly recognizes a constitutional right and makes it retroactively applicable to cases on collateral review, it is § 2254(d)(1) that governs the substantive power of the federal courts to grant a federal remedy to defendants in state custody. Section 2254(d)(1) states that the writ of habeas corpus “shall not be granted” unless it was contrary to, or involved an unreasonable application of, “clearly established Federal law.” The majority’s reading of § 2254(d)(1) means a new rule that satisfies Teague and is made retroactively applicable to cases on collateral review would not be time-barred by § 2244, but a federal court would still be powerless to grant relief under § 2254. Thus, if a watershed decision such as Gideon v. Wainwright, 372 U.S. 335, 83 S.Ct. 792, 9 L.Ed.2d 799 (1963), were to come down tomorrow as a “new rule,” under the majority's holding one would be entitled to file for habeas corpus, but would not be entitled to habeas relief. The majority’s reading of § 2254(d)(1) obviates the need for § 2244(d)(1)(C).
. In fact, the Supreme Court reversed our Court when we called into question the continuing relevance of Teague. Horn v. Banks, 536 U.S. 266, 272, 122 S.Ct. 2147, 153 L.Ed.2d 301 (2002), rev’g Banks v. Horn, 271 F.3d 527 (3d Cir.2001). The Supreme Court instead held the AEDPA and Teague inquiries to be distinct and required federal courts to address Teague when properly raised. Id. There appears to be additional confusion in the majority opinion when it suggests that I “would sub silentio codify Teague." Maj. Op. at 100. But I agree that AEDPA and Teague are distinct inquiries — AEDPA is concerned with the unreasonableness of the state court decision, while Teague is concerned with the yardstick we use to measure unreasonableness. Satisfying Teague may make a Supreme Court decision retroactive and “clearly established Federal law,” but it does not necessarily mean that a state court decision was unreasonable. Likewise, if the relevant state-court decision is rendered post-finality (in the case of a state-court decision rendered during collateral review), Teague may still operate as a bar to federal habeas relief if the relevant Supreme Court decision is also post-finality. Even under my reading, AEDPA and Teague are distinct inquiries.
. To the extent that a defendant is not constitutionally entitled to counsel on discretionary appeals, it seems unreasonable to expect that defendant, who is likely incarcerated, to be fully abreast of Supreme Court rulings and the first person to bring intervening Supreme Court decisions to the attention of the state courts within the window for discretionary review. The state courts have an independent obligation under Griffith to ensure the application of intervening Supreme Court decisions to all cases that are not yet final.
. It is important to note that retroactively applying (under Griffith/Teague) Supreme Court decisions announced after the last state-court decision as "clearly established” does not impugn the state court that fails to predict a later Supreme Court ruling. After all, a state court is not faulted when Griffith directs the application of a "new rule” to cases that are not yet final, but already adjudicated on the merits. In that instance, the state court need only apply the rule to pending cases. Nor does it demonstrate a mistrust of state courts when Teague directs the application of a “new rule” to cases that are already final. Again, to fulfill their judicial obligations, state courts need only apply the retroactive rule to the affected cases.