Opinion by Judge McKEOWN; Concurrence by Judge NOONAN; Concurrence and Dissent by Judge WALLACE.
McKEOWN, Circuit Judge:
Marvin Bockting’s conviction for sexual abuse and life sentences stem from a trial in which the only witness to the conduct, his six-year old stepdaughter, Autumn Bockting, did not testify at trial, but whose interview with a detective was admitted as key evidence. Autumn’s statements at the interview contradicted her testimony at a preliminary hearing where she claimed not to remember what happened with her father. Admission of the interview evidence without cross-examination violated Bockt-ing’s constitutional right “to be confronted with the witnesses against him.” U.S. Const. amend. VI.
Although this case has been before the Nevada Supreme Court twice and before the United States Supreme Court on one occasion, resolution now rests on interpretation of an intervening Supreme Court case: Crawford v. Washington, 541 U.S. 36, 124 S.Ct. 1354, 158 L.Ed.2d 177 (2004). In Crawford, the Court definitively held that “[testimonial statements of witnesses absent from trial have been admitted only where the declarant is unavailable and only where the defendant has had a prior opportunity to cross-examine.” 124 S.Ct. at 1369. Because the little girl’s testimony, which was not subject to cross-examination,' was central to the conviction, its admission can hardly be classified as harmless error. Crawford dictates reversal.
The thorny issue is whether Crawford applies retroactively to this state habeas appeal. In an earlier case, we reserved this question for future consideration. See Leavitt v. Arave, 383 F.3d 809, 830 n. 22 (9th Cir.2004) (per curiam). If, as Judge Noonan argues, Crawford simply reiterates a longstanding rule and does not announce a new rule, then retroactivity falls out of our analysis. If, on the other hand, Crawford is characterized as a “new rule,” then we are faced with analyzing the retroactivity of Crawford in the framework of yet another recent Supreme Court case, Schriro v. Summerlin, — U.S. -, 124 S.Ct. 2519, 159 L.Ed.2d 442 (2004). New rules apply retroactively only where they place “certain kinds of primary, private individual conduct beyond the power of the criminal law-making authority to proscribe,” or where the new rule is “implicit in the concept of ordered liberty.” Teague v. Lane, 489 U.S. 288, 307, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989). The latter category is “reserved for watershed rules of criminal procedure.” Id. at 311.
The threshold question is whether Crawford constitutes a “new rule” under Teague. Judge Noonan’s approach— namely that Crawford does not announce a new rule but rather is a “correction of a misinterpretation,” Concurrence at 2015— has a certain appeal in light of the Court’s historical emphasis in Crawford. Indeed, one can read Crawford as intimating that the rule is longstanding. Unfortunately, Justice Scalia’s analysis is not entirely consistent with that viewpoint. Nonetheless, characterizing Crawford as something less than a new rule, as Judge Noonan does, is one legitimate way of interpreting Crawford. To do so leads to the same result here — the application of Crawford to Bockting’s pending habeas claims.
Despite the appeal of Judge Noonan’s reasoning, application of the Supreme Court’s guidance in Teague leads to the conclusion that Crawford announces a “new rule.” Because the Crawford rule is both a “watershed rule” and one “without [1013]*1013which the likelihood of an accurate conviction is seriously diminished,” Summerlin, — U.S. at -, 124 S.Ct. at 2523, the rule is retroactive.
I. FACTUAL BACKGROUND
Marvin Bockting lived with his wife, Laura, and his two daughters, Autumn and Honesty, in a motel in Las Vegas. Autumn had taken showers together with Laura and Bockting. She had also seen them having sex, and she was accustomed to the use of sexual language.
One Saturday night, when Laura was at home alone with the children, Autumn Bockting woke up crying. Her mother observed that “she looked like she had just woken up from a bad dream and she was quite upset.” At first she refused to tell Laura what was wrong. Laura asked why she wouldn’t tell. Autumn told her, “Because daddy said that you would make him leave and that he would beat my butt if I told you.” After reassurance from her mother, Autumn said “daddy put his pee-pee in her pee-pee, and that daddy put his pee-pee in her butt, and daddy made her suck on his pee-pee like it was a sucker.”
The next day, Laura confronted Bockt-ing. She asked him to leave, which he did. Two days later, a Tuesday, Laura called the rape crisis hotline and was told to take Autumn to the hospital, where they were met by Detective Zinovitch. Zinovitch tried to interview Autumn, but she was too upset. A rape examination was performed. The doctor found that Autumn’s rectal sphincter had been torn within the past week. She also found that Autumn’s hymenal ring was wide open, a rarity in a six-year old. The doctor determined that although she “couldn’t determine what kind of instrument or foreign body was used to cause the laxness of the hymen and the fissure of the rectum,” it had been caused by a blunt force. Two days later, Autumn was again interviewed by a detective. She repeated what she had told her mother, accurately describing the positions of the sex acts. She also demonstrated the acts with anatomically correct dolls.
At Bockting’s preliminary hearing, Autumn was able to answer questions about the difference between the truth and a lie, but became upset when she was asked about being touched by Bockting. Upon further questioning, she said she could not remember what occurred with her father and did not remember whether she had talked with the detective about the claimed assault. The judge declared Autumn an unavailable witness, and the preliminary hearing proceeded with the testimony of Laura and Zinovitch. At trial, the judge found that Autumn’s hearsay statements were admissible because she was effectively unavailable for trial. Without having the opportunity to cross-examine Autumn, Bockting was convicted and sentenced to life in prison.
Bockting appealed to the Nevada Supreme Court, which dismissed his claims. The United States Supreme Court later vacated the Nevada Supreme Court’s decision, remanding for consideration in light of Idaho v. Wright, 497 U.S. 805, 110 S.Ct. 3139, 111 L.Ed.2d 638 (1990). In 1993, the Nevada Supreme Court affirmed Bockt-ing’s conviction. He filed a second petition for postconviction relief, which the state district court denied in 1994.
Three years later, the Nevada Supreme Court again dismissed the appeal. Bockt-ing then sought relief in federal court. He timely filed a habeas petition in 1998, which he amended in 2000. The district court denied the petition, and Bockting filed the present appeal.
II. DISCUSSION
Because Bockting filed his petition after the effective date of the Antiterrorism and [1014]*1014Effective Death Penalty Act of 1996 (“AEDPA”), its provisions apply. See Woodford v. Garceau, 538 U.S. 202, 207, 123 S.Ct. 1398, 155 L.Ed.2d 363 (2003). Under AEDPA, habeas relief to a state prisoner is available only if the state court’s decision “resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States,” or was “based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding.” 28 U.S.C. § 2254(d)(1)-(2).
The Supreme Court also directs us that “in addition to performing any analysis required by AEDPA, a federal court considering a habeas petition must conduct a threshold Teague analysis when the issue is properly raised by the state.” See Horn v. Banks, 536 U.S. 266, 272, 122 S.Ct. 2147, 153 L.Ed.2d 301 (2002). Indeed, as the Court noted, “if our post-AEDPA cases suggest anything about AEDPA’s relationship to Teague, it is that the AEDPA and Teague inquiries are distinct.” Id.
In explaining Teague’s application, the Supreme Court recently explained that there are three steps to determining whether a rule of criminal procedure applies on collateral review:
First, the court must determine when the defendant’s conviction became final. Second, it must ascertain the legal landscape as it then existed, and ask whether the Constitution, as interpreted by precedent then existing, compels the rule. That is, the court must decide whether the rule is actually “new.” Finally, if the rule is new, the court must consider whether it falls within either of the two exceptions to nonretroactivity.
Beard v. Banks, — U.S. -, -, 124 S.Ct. 2504, 2510, 159 L.Ed.2d 494 (2004) (internal citations omitted). Because Bockting’s conviction became final in 1993, we must evaluate whether any subsequent rule of constitutional law is new against the benchmark of that year.
A. CRAWFORD ANNOUNCED A NEW RULE
The question before us is whether the Confrontation Clause principles stated in Crawford amount to a new rule. In Crawford, the Supreme Court considered whether Washington State’s use at trial of a witness’s tape-recorded statement to a police officer violated the Confrontation Clause. 124 S.Ct. at 1357. Writing for the Court, Justice Scalia engaged in a lengthy historical analysis of the Confrontation Clause, noting that “the principal evil at which the Confrontation Clause was directed was the civil-law mode of criminal procedure, and particularly its use of ex parte examinations as evidence against the accused.” Id. at 1363 (emphasis added). He went on to emphasize “that the Framers would not have allowed admission of testimonial statements of a witness who did not appear at trial unless he was unavailable to testify, and the defendant had had a prior opportunity for cross-examination.” Id. at 1365 (emphasis added).
Whether the rule in Crawford is new depends on whether it “was dictated by the then-existing precedent.” Beard, — U.S. at -, 124 S.Ct. at 2511. To be sure, the majority in Crawford is somewhat opaque as to the categorization of its ultimate holding. As the Court noted, “[ajlthough the results of our decisions have generally been faithful to the original meaning of the Confrontation Clause, the same cannot be said of our rationales.” 124 S.Ct. at 1369. The puzzler is whether the “results of [the Supreme Court] decisions” are coextensive with the rules of those decisions. If so, then the Crawford pronouncement could be legitimately [1015]*1015viewed as a continuation of Supreme Court jurisprudence. Careful scrutiny of the Crawford opinion suggests otherwise for at least two reasons: (1) Crawford deviates from the test announced in Ohio v. Roberts, 448 U.S. 56, 100 S.Ct. 2531, 65 L.Ed.2d 597 (1980); and (2) simply reaching the right “result” does not mean that the result flowed from a constant rule.
As the Court observed, “Roberts conditions the admissibility of all hearsay evidence on whether it falls under a ‘firmly rooted hearsay exception’ or bears particularized guarantees of trustworthiness.” Crawford, 124 S.Ct. at 1369 (quoting Roberts, 448 U.S. at 66, 100 S.Ct. 2531). Roberts rests on evidentiary principles of reliability and trustworthiness rather than on the constitutional principle of confrontation.
The Crawford majority quite clearly states that the Roberts “test departs from the historical principles” in that it is both “too broad” and “too narrow.” Crawford, 124 S.Ct. at 1369. Crawford did not announce the same rule as Roberts. Whereas Roberts countenanced the admission of witness testimony if trustworthy, under Crawford the testimony is admitted only if the witnesses are subject to cross-examination. The Court concluded that the cases applying Roberts have “remained faithful to the Framers’ understanding,” id. at 1369, that testimonial statements are admissible only when two criteria are met: unavailability and opportunity for cross-examination. In other words, “[wjhere testimonial statements are at issue, the only indicium of reliability sufficient to satisfy constitutional demands is the one the Constitution actually prescribes: confrontation.” Id. at 1374. Although historical principles are the bedrock for the Crawford decision, getting back to the- bedrock required the Court to disavow the Roberts test. The dissent in Crawford mirrors this view by concluding that “the Court of course overrules Ohio v. Roberts,” id. at 1378, and criticizing the “Court’s adoption of a new interpretation of the Confrontation Clause ....,” id. at 1374.
To say that the outcomes or “results” have been generally “faithful to the original' meaning” is not the same as saying that there has been a consistent rule throughout our history. Indeed, the result in an individual case may well be consistent with the Supreme Court’s new interpretation of Crawford, but that does not mean that the rule that dictated the prior result is necessarily consistent with Crawford. The majority in fact underscores the dichotomy between outcomes and rules in announcing that, “[i]f nothing else, the test we announce is an empirically accurate explanation of the results our cases have reached.” Id. at 1369 n. 9. Rules and outcomes are neither the same nor wholly overlapping.
Finally, the Court in Crawford pinpointed a.situation that was, in-fact, “arguably in tension with the rule requiring a prior opportunity for cross-examination when the. proffered statement is testimonial.” Id. at 1368 n. 8. Citing White v. Illinois, 502 U.S. 346, 112 S.Ct. 736, 116 L.Ed.2d 848 (1992), the Court described a case remarkably similar to ours, in which “statements of a child victim to an investigating police officer [were] admitted as spontaneous declarations.” Id. White rested on the issue of the unavailability requirement under the Confrontation Clause; had Crawford been the rule at the time, the lack of cross-examination would have been fatal to the admission of the evidence.
Although the long line of Confrontation Clause cases is “faithful to the original meaning” of the clause, id. at 1369, we cannot overlook Roberts and White. On balance, an analysis of the historical application of the Confrontation Clause cases [1016]*1016leads to the conclusion that Crawford announces a new rule that must be put through the Summerlin strainer. If, on the other hand, Crawford does not announce a new rule — the position taken by Judge Noonan — then the Summerlin analysis is not required. Either way, the result in this case is the same. Bockting’s petition should be granted.
B. SUMMERLIN CONTROLS THE RETROACTIVITY ANALYSIS
Because Crawford announces a new rule, we must ask whether it falls into one of the two Teague exceptions to the bar on retroactivity. The first Teague exception is for primary conduct that cannot be criminalized. The second is for bedrock rules of criminal procedure. Teague, 489 U.S. at 307, 109 S.Ct. 1060. It is the second exception that is at play in this case.
The Crawford rule does not narrow the scope of a criminal statute by interpreting its terms, nor is it a constitutional determination that places particular conduct or persons covered by the statute beyond the State’s power to punish. Summerlin, — U.S. at -, 124 S.Ct. at 2522. Rather, the rule in Crawford is procedural. See also Crawford, 124 S.Ct. at 1370 (stating the Confrontation Clause provides “a procedural rather than a substantive guarantee”). Therefore, Crawford merits retroactive application only if it implicates “the fundamental fairness and accuracy of the criminal proceeding,” Saffle v. Parks, 494 U.S. 484, 495, 110 S.Ct. 1257, 108 L.Ed.2d 415 (1990), and reworks our understanding of bedrock criminal procedure, Sawyer v. Smith, 497 U.S. 227, 242, 110 S.Ct. 2822, 111 L.Ed.2d 193 (1990).
To be sure, under Summerlin, the lower courts are tightly constrained in determining whether a new rule is a “watershed rule” that is not only fundamental, but one “without which the likelihood of an accurate conviction is seriously diminished.” Summerlin, — U.S. at -, 124 S.Ct. at 2523 (quoting Teague, 489 U.S. at 313, 109 S.Ct. 1060). Viewing Crawford and Sum-merlin together poses a conundrum. Justice Scalia wrote for the majority in both cases. In Summerlin, the Court admonished that “[t]his class of [retroactive] rules is extremely narrow, and it is unlikely that any ... ha[s] yet to emerge.” — U.S. at -, 124 S.Ct. at 2523 (quoting Tyler v. Cain, 533 U.S. 656, 667 n. 7, 121 S.Ct. 2478, 150 L.Ed.2d 632 (2001)) (internal quotations omitted). Nonetheless, the bar is not absolute and the Crawford rule meets the Court’s criteria. Admonitions such as in Summerlin offer discouragement but no guidance. Because our job is not to conjure up hidden meaning, we simply heed the warning and our analysis thus adheres faithfully to the holding and rationale of Summerlin. Viewing Crawford in light of Summerlin leads to the conclusion that the Crawford cross-examination requirement merits retroactive application.
That the Crawford requirement is fundamental to our legal regime is beyond dispute. Justice Scalia’s eloquent recitation of the history, purpose, and place of the Confrontation Clause and cross-examination answers this question. Crawford, 124 S.Ct. at 1359. Hundreds of years of tradition have embedded this notion as a fundamental role. Indeed, “[t]he Framers would be astounded to learn that ex parte testimony could be admitted against a criminal defendant because it was elicited by ‘neutral’ government officers.” Id. at 1373.
The question next posed is whether the rule implicates the accuracy of the criminal proceeding. Juxtaposed, Summerlin actually underscores why the Crawford rule implicates the “fundamental fairness and accuracy of the criminal proceeding,” Saffle, 494 U.S. at 495, and alters our under[1017]*1017standing of bedrock procedural principles, whereas the sentencing scheme in Summerlin did not. In Summerlin, the Court concluded that Ring v. Arizona, 536 U.S. 584, 122 S.Ct. 2428, 153 L.Ed.2d 556 (2002), which held that aggravating factors in a death penalty case must be proved to a jury rather than to a judge, was not a retroactive rule. Summerlin asked “whether judicial factfinding so seriously diminished] accuracy that there is an im-permissibly large risk of punishing conduct the law does not reach.” Summerlin, — U.S. at -, 124 S.Ct. at 2525 (citations and quotations omitted). The answer was no because “[t]he evidence is simply too equivocal to support that conclusion.” Id.
In Summerlin, the Court observed that “for every argument why juries are more accurate factfinders, there is another why they are less accurate.”1 In contrast, the evidence that cross-examination seriously decreases the possibility of inaccurate conviction is unequivocal.
The Supreme Court has repeatedly and without deviation held that the purpose of the Confrontation Clause is to promote accuracy. See, e.g., Crawford, 124 S.Ct. at 1370 (“This open examination of witnesses ... is much more conducive to the clearing up of truth.”) (quoting 3 Blackstone, Commentaries * 373); White v. Illinois, 502 U.S. 346, 355, 112 S.Ct. 736, 116 L.Ed.2d 848 (1992) (evaluating a Confrontation Clause claim against the benchmark of accuracy); Tennessee v. Street, 471 U.S. 409, 415, 105 S.Ct. 2078, 85 L.Ed.2d 425 (1985) (describing the “Confrontation Clause’s very mission” as “to advance the accuracy of the truth-determining process in criminal trials”) (internal quotation omitted); Ohio v. Roberts, 448 U.S. 56, 65, 100 S.Ct. 2531, 65 L.Ed.2d 597 (1980) (stating, the purpose of the Confrontation Clause is to “augment accuracy in the factfinding process”); Parker v. Randolph, 442 U.S. 62, 73, 99 S.Ct. 2132, 60 L.Ed.2d 713 (1979) (plurality) (“The right of confrontation conferred by the Sixth Amendment is a safeguard to ensure the fairness and accuracy of criminal trials.”); Chambers v. Mississippi, 410 U.S. 284, 295, 93 S.Ct. 1038, 35 L.Ed.2d 297 (1973) (“The right of cross-examination is more than a desirable rule of trial procedure. It is implicit in the constitutional right of confrontation, and helps assure the ‘accuracy [1018]*1018of the truth-determining process.’ ”); Dutton v. Evans, 400 U.S. 74, 89, 91 S.Ct. 210, 27 L.Ed.2d 213 (1970) (plurality) (“The decisions of this Court make it clear that the mission of the Confrontation Clause is to advance a practical concern for the accuracy of the truth-determining process in criminal trials....”); Pointer v. Texas, 380 U.S. 400, 404, 85 S.Ct. 1065, 13 L.Ed.2d 923 (1965) (“[P]robably no one, certainly no one experienced in the trial of lawsuits, would deny the value of cross-examination in exposing falsehood and bringing out the truth in the trial of a criminal case.”); Mattox v. United States, 156 U.S. 237, 242-43, 15 S.Ct. 337, 39 L.Ed. 409 (1895) (describing the “primary object” of the Confrontation Clause as “to prevent depositions or ex parte affidavits ... being used against the prisoner in lieu of a personal examination and cross-examination of the witness, in which the accused has an opportunity, not only of testing the recollection and sifting the conscience of the witness, but of compelling him to stand face to face with the jury in order that they may look at him, and judge by his demeanor upon the stand and the manner in which he gives his testimony whether he is worthy of belief’). See also M. Hale, History and Analysis of the Common Law of England 258 (1713) (adversarial testing “beats and bolts out the Truth much better”) (quoted in Crawford, 124 S.Ct. at 1370).
But accuracy and reliability do not exist in a vacuum. Rather, “[t]he central concern of the Confrontation Clause is to ensure the reliability of the evidence against a criminal defendant by subjecting it to rigorous testing in the context of an adversary proceeding before the trier of fact.” “The word ‘confront,’ after all, also means a clashing of forces or ideas, thus carrying with it the notion of adversariness.” Maryland v. Craig, 497 U.S. 836, 845, 110 S.Ct. 3157, 111 L.Ed.2d 666 (1990). Even more to the point, the Court wrote, “[t]he combined effect of these elements of confrontation' — -physical presence, oath, cross-examination, and observation of demeanor by the trier of fact — serves the purposes of the Confrontation Clause by ensuring that evidence admitted against an accused is reliable and subject to the rigorous adversarial testing that is the norm of Anglo-American criminal proceedings.” Id. at 846, 110 S.Ct. 3157 (emphasis added).
Crawford itself answers the question of whether the absence of cross-examination “so seriously diminishe[s] accuracy that there is an impermissibly large risk of punishing conduct the law does not reach” Summerlin, — U.S. at -, 124 S.Ct. at 2525 (internal quotations omitted):
• “The framework is so unpredictable that it fails to provide meaningful protection from even core confrontation violations.” Crawford, 124 S.Ct. at 1371.
• “[T]he [general reliability] test is inherently, and therefore permanently, unpredictable.” Id. at 1374 n. 10. “It is difficult to imagine [the general reliability framework] providing any meaningful protection in [politically charged cases].” Id. at 1374.
• “[W]e view this as one of those rare cases in which the result below is so improbable that it reveals a fundamental failure on our part to interpret the Constitution in a way that secures its intended constraint on judicial discretion.” Id. at 1373.
Thus, at the heart of the Court’s concerns in Crawford was the reliability of admitted evidence. Where admitted evidence is unreliable, the accuracy of convictions is seriously undermined. That the rule in Crawford is one without which the accuracy of convictions would be seriously undermined is further born out by the Court’s own [1019]*1019description of its prior doctrine as a “rare case” of “fundamental failure.” Id. at 1373. The difference between pre- and post-Crawford Confrontation Clause jurisprudence is not the sort of change that can be dismissed as merely incremental. Instead, it is an “absolute pre-requisite to fundamental fairness.” Sawyer, 497 U.S. at 244, 110 S.Ct. 2822.
In Crawford, the Court itself faults the previous regime under Roberts. Crawford, 124 S.Ct. at 1371-73. .Indeed, to benchmark reliability against Roberts would undermine Crawford’s central thesis. Reliance on Roberts’ judicially-administered reliability test gives illusory comfort, as “[t]he [Confrontation] Clause ... reflects a judgment ... about how reliability can best be determined,” id. at 1370, and “[t]he legacy of Roberts in other courts vindicates the Framers’ wisdom in rejecting a general reliability exception,” id. at 1371.2
The retroactivity inquiry is not conducted in a vacuum. Rather, the analysis is modeled on the Summerlin requirements. But even if a “look back” to Roberts were appropriate, as Judge Wallace suggests, Crawford would be retroactive.
Rules that are properly considered retroactive are those that “alter our understanding of the bedrock procedural elements essential to the fairness of a proceeding.” Sawyer v. Smith, 497 U.S. 227, 242, 110 S.Ct. 2822, 111 L.Ed.2d 193 (1990) (internal quotations omitted). As the Court recently stated, “[i]n providing guidance as to what might fall within this exception, we have repeatedly referred to the rule of Gideon v. Wainwright, 372 U.S. 335, 83 S.Ct. 792, 9 L.Ed.2d 799 (1963).” Beard, — U.S. at -, 124 S.Ct. at 2514 (citing Saffle, 494 U.S. at 495, 110 S.Ct. 1257, Gilmore v. Taylor, 508 U.S. 333, 364, 113 S.Ct. 2112, 124 L.Ed.2d 306 (1993) (Blackmun, J., dissenting)). Recognizing that bedrock procedural rules are very few in number, it is no leap to conclude that the right of cross-examination as an adjunct to the constitutional right of confrontation joins the very limited company of Gideon.
We join one other circuit that has concluded that Crawford announces a new rule, although its retroactivity analysis differs from ours. See Brown v. Uphoff, 381 F.3d 1219 (10th Cir.2004). The Second Circuit did not directly address the new rule issue but concluded that even if Crawford did announce a new rule, it would not be retroactive. Mungo v. Duncan, 393 F.3d 327, 336 (2d Cir.2004).
Addressing the question of a new rule, the Tenth Circuit reasoned as follows:
As we explained above, prior to the decision in Crawford, Roberts provided the appropriate framework for determining whether the admission of hearsay statements violated the Confrontation Clause. The Supreme Court itself noted that the logic of Roberts was inconsistent with the Court’s conclusion in Crawford that the Confrontation Clause requires an opportunity to cross-examine before testimonial hearsay may be admitted against the defendant. Crawford, — U.S. at -, 124 S.Ct. at 1369. Thus, Roberts and its progeny did not dictate the result in Crawford and we conclude that it announces a new rule of constitutional law. See Crawford, at -, 124 S.Ct. at 1374 (Rehnquist, C.J., dissenting) (referring to the majority’s holding [1020]*1020as a “new interpretation of the Confrontation Clause”).
Brown, 381 F.3d at 1226. Our analysis is in accord.
But the Tenth Circuit diverges from us in holding that Crawford is not retroactive because it does not set forth a watershed rule of criminal procedure. Id. Stating that unless 2008 a new rule is “on the magnitude of the rule announced in Gideon v. Wainwright,” it will not fit within the Teague watershed rule exception, the court explained:
Unlike Gideon, Crawford does not ‘alter[ ] our understanding of what constitutes basic due process,’ but merely sets out new standards for the admission of certain kinds of hearsay. Confrontation Clause violations are subject to harmless error analysis and thus may be excused depending on the state of the evidence at trial. It would, therefore, be difficult to conclude that the rule in Crawford alters rights fundamental to due process.
Id. at 1226-27 (internal citations omitted).
The Supreme Court’s Confrontation Clause jurisprudence in Crawford cannot be dismissed as a mere tweak on the admissibility of hearsay. See Brown, 381 F.3d at 1226. The Supreme Court surely did not conceive of it as such. Rather, the Court describes the right of confrontation as a “bedrock procedural guarantee,” notes that it “dates back to Roman times” and was part of the common law known to the founding generation. Crawford, 124 S.Ct. at 1359. The Court also contrasts “exclusion under the hearsay rules” with “the civil-law abuses the Confrontation Clause targeted.” Crawford, 124 S.Ct. at 1364. In a rare mea culpa, the Court faults itself for not enunciating the Crawford rule earlier, stating that “it reveals a fundamental failure on our part to interpret the Constitution in a way that secures its intended constraint on judicial discretion.” Id. at 1373. There is nothing “mere” about the Crawford rule.
The Tenth Circuit mistakenly concluded that rules of constitutional law subject to harmless error review can never be considered bedrock rules of procedure. The two inquiries hinge on different questions. Whether a rule is a bedrock rule of procedure depends on whether it 'increases the likelihood of accurate conviction. Summerlin, — U.S. at -, 124 S.Ct. at 2523. Whether a rule is subject to harmless error analysis depends on whether the impact of the error can be measured. See Arizona v. Fulminante, 499 U.S. 279, 307-08, 111 S.Ct. 1246, 113 L.Ed.2d 302 (1991). Therefore, a rule of constitutional law could be essential to promote accurate convictions, but still subject to harmless error review if the impact of misapplication of the rule were easily measurable. In short, because accuracy and measurability are different concepts, whether a rule1 of constitutional law is subject to harmless error review does not answer the question whether it is a bedrock rule of procedure.
After assuming that Crawford announced a new rule, the Second Circuit rejected retroactivity, reasoning that the Crawford rule would not improve overall accuracy because “it is likely to improve accuracy in some circumstances and diminish it in others.” Mungo, 393 F.3d at 335. The flaw in this analysis is that the Second Circuit has substituted its judgment of whether the Crawford rule is one without which the accuracy of conviction is seriously diminished, for the Supreme Court’s considered judgment. The Court has found repeatedly that the purpose of the Confrontation Clause is to promote accuracy, see, e.g., Tennessee v. Street, 471 U.S. 409, 415, 105 S.Ct. 2078, 85 L.Ed.2d 425 (1985), and thus Crawford rejected the Roberts framework as reflective of “a fun[1021]*1021damental failure on our part to interpret the Constitution in a way that secures its intended constraint on judicial discretion,” Crawford, 124 S.Ct. at 1373. Viewing these holdings together leads to the conclusion that the Craivford rule is one without which the likelihood of accurate conviction is seriously diminished.
C. BOCKTING MERITS RELIEF UNDER AEDPA
Having determined that Crawford is retroactive, the remaining task is to determine whether, under AEDPA, the Nevada Supreme Court’s analysis was either “contrary to, or involved an unreasonable application of, clearly established Federal law.” 28 U.S.C. § 2254(d)(1)-(2). The Supreme Court has yet to address directly whether AEDPA was intended or should be read to adopt the Teague exceptions. Williams v. Taylor, 529 U.S. 362, 380, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000) (Stevens, J., for four justices) (“AEDPA codifies Teague to the extent that Teague requires federal habeas courts to deny relief that is contingent upon a rule of law not clearly established at the time the state conviction became final.”). Application of Teague is the means by which new rules are made retroactive. As noted earlier, the Court has clarified that the Teague and AEDPA inquiries are separate. Horn, 536 U.S. at 272, 122 S.Ct. 2147. But in directing us to undertake both inquiries in an AEDPA case, the Court has impliedly endorsed the application of Teague in the AEDPA context. Further, it appears that Congress intended to preserve the Teague exceptions because AEDPA explicitly provides for their application in proceedings involving state habeas petitions. See 28 U.S.C. § 2254(e) (“If the applicant has failed to develop the factual basis of a claim in State court proceedings, the court shall not hold an evidentiary hearing on the claim unless the applicant shows that ... the claim relies on a new rule of constitutional law, made retroactive to cases on collateral review by the Supreme •Court.”) But even if Congress’ intent is unclear, the constitutional doubt canon of construction mandates that we read the statute to incorporate the Teague exceptions to avoid the serious constitutional problem raised by depriving individuals of bedrock principles of Due Process. Ferguson v. Palmateer, 321 F.3d 820, 823 (9th Cir.2003).
The Nevada Supreme Court relied on Roberts to conclude that Autumn was “unavailable” and that her statements bore the requisite “particularized guarantees of trustworthiness.” Bockting v. State, 109 Nev. 103, 847 P.2d 1364, 1367-70 (1993). As we now know, but the Nevada Supreme Court could not have divined, Crawford dictates the right to cross-examine the witness. The progeny of Roberts, such as the multi-factor reliability test developed in Idaho v. Wright, 497 U.S. 805, 821-22, 110 S.Ct. 3139, 111 L.Ed.2d 638 (9th Cir.1990), cannot override the right of confrontation. Thus, the Nevada Supreme Court’s decision was “contrary to” established Supreme Court precedent in Crawford, as made retroactive under Teague and Sum-merlin.
It bears noting that Crawford requires not only cross-examination but unavailability. We have long acknowledged the sensitive situation presented by the child witness. See, e.g., Tome v. United States, 513 U.S. 150, 166, 115 S.Ct. 696, 130 L.Ed.2d 574 (1995). Indeed, prosecutions for child abuse often rely heavily on such testimony. Here, the trial court’s inquiry into Autumn’s unavailability was truncated and conclusory at best. When she refused to cooperate, the court simply declared her unavailable. Athough this issue is troubling, it is not necessary to reach this question because of the constitutional er[1022]*1022ror related to the lack of cross-examination.
The final question is whether admission of Autumn’s statement is harmless beyond a reasonable doubt. See Neder v. United States, 527 U.S. 1, 15, 119 S.Ct. 1827, 144 L.Ed.2d 35 (1999). The detective’s testimony regarding Autumn’s interview Was a critical piece of evidence, particularly in view of Autumn’s inconsistent testimony at the preliminary hearing. Even if her statement to the mother was, for argument’s sake, considered admissible, the detective’s description of Autumn’s interview was so significant that the error could have materially affected the verdict. Thus, admitting Autumn’s statement was not harmless beyond reasonable doubt. •
III. CONCLUSION
Because a majority ' concludes that Crawford must be applied in this pending habeas case, Bockting’s petition for a writ of habeas corpus is GRANTED.3