I. Introduction
We review and affirm the court of appeals’ decision in People v. Edwards, 101 P.3d 1118 (Colo.App.2004), which held that Crawford v. Washington, 541 U.S. 36, 124 S.Ct. 1354, 158 L.Ed.2d 177 (2004), does not apply retroactively to cases involving postconviction proceedings that concern convictions finalized prior to Crawford.
Edwards initiated postconviction proceedings to vacate his felony conviction, which was finalized before the United States Supreme Court decided Crawford. The Court rendered the Crawford decision while Edwards’s postconviction proceedings were pending. This case established that testimonial 1 out-of-court statements are a violation of the Confrontation Clause unless the witness is unavailable to testify at trial and the defendant had a previous opportunity to adequately cross-examine the witness. Because Edwards had had no opportunity to cross-examine a key witness whose statements were admitted at trial under various hearsay exceptions, he seeks to vacate his conviction on grounds that include the violation of his confrontation rights under Crawford.2
Generally, new rules of criminal procedure do not apply retroactively to cases on collateral review. However, the United States Supreme Court case of Teague v. Lane, 489 U.S. 288, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989), outlines two exceptions to this general rule. We hold that Crawford is a new rale of criminal procedure and adopt the Teague test. We analyze Crawford under the second Teague exception, which allows for retroactive application of a rale if it constitutes a “watershed” rule of criminal procedure. Id. at 311, 109 S.Ct. 1060. To be considered watershed, a rale must meet two criteria: (1) *979“[i]nfringement of the rule must seriously diminish the likelihood of obtaining an accurate conviction;” and (2) “the rule must alter our understanding of the bedrock procedural elements essential to the fairness of a proceeding.” Tyler v. Cain, 533 U.S. 656, 665, 121 S.Ct. 2478, 150 L.Ed.2d 632 (2001) (citations and internal quotation marks omitted).
The United States Supreme Court has identified only one rule as watershed: the holding in Gideon v. Wainwrighi, 372 U.S. 335, 83 S.Ct. 792, 9 L.Ed.2d 799 (1963), which recognizes a criminal defendant’s right to counsel in cases that involve a possible prison sentence. Because the Crawford holding does not alter fundamental due process rights to the extent that the Gideon guarantee of right to counsel does, we hold that, under current United States Supreme Court precedent, Crawford does not qualify as a watershed rule.
We therefore affirm the court of appeals’ decision that Crawford does not apply retroactively to cases involving posteonviction proceedings that concern convictions finalized prior to Crawford.
II. Facts and Proceedings Below
Police forced William Edwards, petitioner and defendant below, to halt his vehicle when he failed to pull over for a traffic stop. As the police addressed Edwards, a passenger, who appeared to have been beaten, emerged from Edwards’s car yelling, “He beat me. He beat me.” Later, during the victim’s treatment for her injuries, she told attending medical personnel that she had been assaulted.
The prosecution could not locate the victim to testify at trial; but the trial court admitted her statements to the police under the excited utterance exception to the hearsay rule, CRE 803(2). The trial court also admitted the victim’s statements to medical personnel, relying on the medical diagnosis and business records exceptions to the hearsay rule — CRE 803(4) and (6), respectively. A jury convicted Edwards of first-degree assault with a deadly weapon, and a judge subsequently convicted him of habitual criminal counts.
Edwards’s conviction was affirmed in a direct appeal on issues unrelated to our inquiry, and this court denied certiorari on the case. People v. Edwards, 971 P.2d 1080 (Colo.App.1998), cert. denied, May 10, 1999. Edwards then filed a Crina. P. 35(e) motion to vacate judgment of conviction, in which he argued, among other issues, that he was denied his constitutional right to confront the victim as a witness against him. The trial court denied his motion. On appeal, in an unpublished opinion, the court of appeals remanded the case with directions for the trial court to address the confrontation issue. The trial court ruled that Edwards’s constitutional rights of confrontation were protected through the hearsay rule exceptions. While Edwards’s appeal of that ruling was pending, the United States Supreme Court decided Craivford, which held that testimonial out-of-court statements are a violation of the Confrontation Clause unless the witness is unavailable to testify at trial and the defendant had a previous opportunity to adequately cross-examine the witness. 541 U.S. at 68,124 S.Ct. 1354.
The court of appeals, relying on Teague, held that Crawford did not apply retroactively to Edwards’s case. Edwards, 101 P.3d at 1124. In reaching this conclusion, the court of appeals reasoned that the holding in Crawford announced a new constitutional rule of criminal procedure, and thus was not applicable retroactively unless it fit one of two exceptions outlined in Teague. Id. at 1121-22. The court of appeals analyzed the rule under Teague’s second exception — the watershed exception. The court of appeals held that the Crawford rule is not a watershed rule and, therefore, does not apply to Edwards. Id. at 1124.
Edwards now petitions this court on certio-rari to reverse the judgment of the court of appeals. He argues that Crawford should apply retroactively to his conviction for several reasons: (1) Crawford did not announce a new rule of criminal procedure; (2) the court of appeals erred in applying the Teague retroactivity test because it has not yet been adopted by this court; and (3) if Teague is the applicable test, Crawford nonetheless *980constitutes a watershed rule of criminal procedure and thus is retroactively applicable.
III. Analysis
The sole issue we address in this case is whether Crawford applies retroactively to convictions finalized before that decision. Generally, new constitutional rules of criminal procedure do not apply retroactively to cases on collateral review. Schriro v. Sum-merlin, 542 U.S. 348, 352, 124 S.Ct. 2519, 159 L.Ed.2d 442 (2004). However, the United States Supreme Court has set forth exceptions to this general rule. Teague, 489 U.S. at 311, 109 S.Ct. 1060. In order to ascertain whether Crawford applies retroactively, we begin our analysis with an overview of Crawford and its constitutional underpinnings. We then establish which test controls the decision whether the rule announced in Crawford applies retroactively. Finally, we analyze the Crawford rule under the applicable test. To do so, we must make a series of determinations: (1) whether Crawford announces a procedural rule; (2) whether that rule is new; and (3) if the Crawford holding does amount to a new rule of criminal procedure, whether it constitutes a “watershed” rule. We turn now to the Sixth Amendment.
A. The Confrontation Clause of the Sixth Amendment
The Confrontation Clause of the Sixth Amendment of the Constitution, applicable to the states through the Fourteenth Amendment, provides “[i]n all criminal prosecutions, the accused shall enjoy the right ... to be confronted with the witnesses against him.” U.S. Const. amend. VI; Gonsoir v. People, 793 P.2d 1165, 1165 n. 2 (Colo.1990) (citing Ohio v. Roberts, 448 U.S. 56, 62, 100 S.Ct. 2531, 65 L.Ed.2d 597 (1980)).3 In Crawford, the U.S. Supreme Court explained that the Confrontation Clause is principally directed toward controlling the admission of “ex parte examinations as evidence against the accused.” 541 U.S. at 50, 124 S.Ct. 1354. In order to discuss Crawford’s impact on Confrontation Clause jurisprudence, we must first consider its predecessor.
Prior to Crawford, Ohio v. Roberts, 448 U.S. 56, 100 S.Ct. 2531, 65 L.Edüd 597 (1980), was the controlling case outlining the elements needed to satisfy the Confrontation Clause. In Roberts, the United States Supreme Court held the Confrontation Clause required that, in order for a witness’s prior statement to be admitted at trial, the witness must be deemed unavailable to testify and the statement must bear “adequate indicia of reliability.” Id. at 66, 100 S.Ct. 2531 (internal quotation marks omitted). The Court determined that hearsay evidence could be considered adequately reliable if it “falls within a firmly rooted hearsay exception.” Id.
In Crawford, the Supreme Court revised the criteria under which testimonial out-of-court statements may be admitted at trial when the witness who made the statements does not testify. Crawford involved a wife who refused to testify against her husband under marital privilege. 541 U.S. at 40, 124 S.Ct. 1354. As a result, the prosecution sought to have her earlier statement to police admitted at trial under a hearsay exception. Id. The Court held that the Confrontation Clause requires out-of-court testimonial statements by witnesses to be barred from use at trial unless the witness “was unavailable to testify, and the defendant had had a prior opportunity for cross-examination.” Id. at 53-54,124 S.Ct. 1354. Referencing the Roberts requirement for adequate reliability, the Court explained that the only constitutionally adequate indication of reliability for testimonial statements is confrontation, as the Constitution requires. Id. at 68-69, 100 S.Ct. 2531. Thus, by adding the requirement that there be an opportunity for cross-examination, Crawford abrogates Roberts.
However, we reiterate the distinction we noted in our recent decision of Compan v. People, 121 P.3d 876, 882 (Colo.2005). In that decision we held that, though Crawford controls the admission of testimonial state*981ments under the federal Confrontation Clause, the Roberts test still applies to non-testimonial evidence.4 Hence, as we held in Compan, Crawford abrogates Roberts only in cases where testimonial statements are at issue.5
B. Retroactivity of New Rules of Criminal Procedure to Post-conviction Proceedings
Having outlined the constitutional principles involved in this issue, we now consider which test we will use to determine whether Crawford applies retroactively to Edwards’s case.
In determining whether new rules of constitutional criminal procedure apply retroactively, we have previously used the test the United States Supreme Court articulated in Linkletter v. Walker, 381 U.S. 618, 636, 85 S.Ct. 1731, 14 L.Ed.2d 601 (1965), which we adopted in People v. Walker, 666 P.2d 113, 117 (Colo.1983).6 Because we have not faced a question of the retroactivity of a constitutional rule of criminal procedure since the Supreme Court announced the Teague test, Linkletter arguably remained precedent in Colorado at the time the court of appeals rendered its decision. However, in deciding this case below, our court of appeals used the newer Teague test. Edwards, 101 P.3d at 1121-24. The court of appeals did so without referencing the Linkletter test or explicitly adopting the Teague test.
Edwards asserts that the court of appeals erred in applying Teague rather than Link-letter. He bases his conclusion on two arguments. First, he contends that Teague is not binding on state courts, thus the court of appeals should have applied the Linkletter test as dictated by the precedent of this court. Second, Edwards argues Teague is not an appropriate test for determining whether judicial decisions announcing new rules of criminal procedure apply to claims raised in Crim. P. 35(c) motions. We address each argument in turn.
To support his assertion that Teague is not binding on state courts, Edwards contends that, though states are required to follow United States Supreme Court precedent that interprets provisions of the federal Constitution, the Teague test is not such an interpretation. We agree that states are not bound to follow Supreme Court precedent in all circumstances. See Dickerson v. United States, 530 U.S. 428, 438-39, 120 S.Ct. 2326, 147 L.Ed.2d 405 (2000). But we are not as easily persuaded that Teague is not the type of constitutional decision that we are bound to apply.
The United States Supreme Court has often recognized that its authority in state courts is limited: “ ‘[f]ederal judges ... may not require the observance of any special procedures’ in state courts ‘except when necessary to assure compliance with the dictates of the Federal Constitution.’ ” Id. (quoting Harris v. Rivera, 454 U.S. 339, 344-45, 102 S.Ct. 460, 70 L.Ed.2d 530 (1981)); see also Mu’Min v. Virginia, 500 U.S. 415, 422, 111 S.Ct. 1899, 114 L.Ed.2d 493 (1991) (“[I]n state courts ... our authority is limited to enforcing the commands of the United States Constitution.”)
Teague mandates that, “[ujnless 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.” Teague, 489 U.S. at 310, 109 5.Ct. 1060. However, the federal Constitution itself does not require or prohibit retroactive application of new constitutional rules. Moore v. People, 707 P.2d 990, 996 (Colo. 1985). Thus, though Teague regulates the application of constitutional rules, we ac*982knowledge the possibility that it does not interpret a constitutional rule itself, and that we are therefore not required to enforce it. Mandatory compliance with Teague would not necessarily “assure compliance with the dictates of the Federal Constitution,” especially since Teague often precludes the application of new constitutional rules. Dickerson, 530 U.S. at 438-39, 120 S.Ct. 2326. Perhaps if Teague required constitutional rules to be applied retroactively in all situations, we could definitively conclude that it assures compliance with the Constitution. But we leave this debate for another day because, even if we are not bound to follow Teague, we have the discretion to do so. See People v. Timmons, 690 P.2d 213, 215 (Colo. 1984). This observation leads us to Edwards’s second argument against the court of appeals’ application of Teague and an analysis of whether we should adopt the Teague test.
In Edwards’s second argument supporting his assertion that the court of appeals erred in applying Teague rather than Linkletter, he asserts that Teague is not an appropriate test for determining whether judicial decisions announcing new rules of criminal procedure apply to claims raised in Crim. P. 35(c) motions. Edwards reasons that Teague, which was decided in the context of federal habeas corpus review, puts forth a test more appropriate for that purpose than for Crim. P. 35(c) motions. Edwards relies on the reasoning of Justice Harlan in Mackey v. United States, 401 U.S. 667, 91 S.Ct. 1160, 28 L.Ed.2d 404 (1971), upon which the Teague Court bases its holding, to conclude that the purpose of the Teague test is for federal courts to respect comity and the finality of state court judgments. In contrast, Edwards argues, the purpose of Crim. P. 35(c) motions is to prevent criminal injustice. Thus, Edwards concludes, the more flexible Linkletter test is better suited to meet the purpose of state Crim. P. 35(c) motions.
We do not read Teague so narrowly. While the decision was rendered in the context of federal habeas corpus review of a criminal state court conviction under 28 U.S.C. § 2254, the Teague Court’s rationale — and that of Justice Harlan upon which it is based — addresses the broader category of cases on collateral review, which includes Crim. P. 35(c) motions. See Teague, 489 U.S. at 301, 305, 109 S.Ct. 1060 (modifying the Supreme Court’s approach to “retroactivity for cases on collateral review”); Mackey, 401 U.S. at 691-92, 91 S.Ct. 1160. Rule 35(c) “affords every person convicted of a crime the right to seek postconviction review upon the grounds that the conviction was obtained in violation of the Constitution or laws of the United States or the constitution or laws of this state.” Robbins v. People, 107 P.3d 384, 387 (Colo.2005) (citing People v. Hubbard, 184 Colo. 243, 247, 519 P.2d 945, 947 (1974)). Within our criminal justice system, “postcon-viction proceedings have a dual purpose: to prevent constitutional injustice and to bring finality to judgment.” People v. Rodriguez, 914 P.2d 230, 252 (Colo.1996) (citing People v. Hampton, 187 Colo. 131, 133, 528 P.2d 1311, 1312 (1974)). The Teague test meets both of these goals.
The Court in Teague emphasizes finality as an underlying consideration for its decision. But the Court also acknowledges that a balance must be struck between honoring finality and preventing injustice: “[t]he fact that life and liberty are at 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.” Teague, 489 U.S. at 309, 109 S.Ct. 1060 (citations and internal quotation marks omitted). Thus, while the Teague test underscores the preservation of finality, it allows for the prevention of injustice in the most egregious instances through its exceptions to the general rule that new constitutional rules of criminal procedure do not apply retroactively to cases on collateral review.
We have also recognized that the concept of finality is an important landmark on the Colorado criminal justice landscape. E.g., People v. Wiedemer, 852 P.2d 424, 434 (Colo. 1993). And we have noted its enhanced significance in the context of Crim. P. 35(c) proceedings. In Waits v. People, we declined to retroactively apply a constitutional rule of criminal procedure, stating that “[t]he fact that this is a collateral attack ... under *983Crim. P. 35(c) serves as an additional reason not to apply [the rule] retrospectively.” 724 P.2d 1329, 1336 (Colo.1986). Hence, given Crim. P. 35(e) postconviction proceedings’ twin purposes of finality and prevention of injustice, and given Teague’s balance between these two goals, we conclude that the Teague test is appropriate for determining whether new rules of criminal procedure apply to Crim. P. 35(c) claims.
Finally, we note that, though before today we had yet to explicitly adopt the Teague test, Colorado jurisprudence has indicated an implicit acceptance of the test. See Martin Marietta Corp. v. Lorenz, 823 P.2d 100, 119 n. 2 (Colo.1992) (Erickson, J., concurring in part and dissenting in part) (“The Linkletter approach has been rejected in criminal law. New rules must be applied to all cases pending on direct review ... but not necessarily to cases pending on collateral review.”) (citing Teague, 489 U.S. at 310, 109 S.Ct. 1060); see also People v. Bradbury, 68 P.3d 494, 497 (Colo.App.2002) (“[U]nless they fall within one of the exceptions recognized in Teague v. Lane ... new constitutional rules of criminal procedure are not applicable to cases that have become final before the new rules are announced.”), cert, denied, Apr. 28, 2003.
We have consistently followed the lead of the United States Supreme Court when determining whether a rule of criminal procedure is retroactive. Timmons, 690 P.2d at 215. Thus, for reasons of uniformity and compliance with current Supreme Court precedent, and because Teague meets the underlying goals of Crim. P. 35(c) collateral attacks, we adopt the Teague test with respect to federal constitutional determinations of retroactivity and apply it here. In doing so, we join the ranks of a majority of states. See Windom v. State, 886 So.2d 915, 943 (Fla.2004) (noting that twenty-eight state supreme courts and the District of Columbia had adopted the Teague test at that point in time).
Having adopted the Teague test to determine whether constitutional rules of criminal procedure apply retroactively to cases on collateral review, we now apply the test. To analyze Crawford under Teague, we employ yet another test, which the Supreme Court has outlined for determining whether a constitutional rule of criminal procedure applies to postconviction proceedings. Beard v. Banks, 542 U.S. 406, 411, 124 S.Ct. 2504, 159 L.Ed.2d 494 (2004). The Banks test incorporates Teague and requires a three-part examination: (1) whether the defendant’s conviction is final; (2) whether the rule in question is in fact new; and (3) if the rule is new, whether it meets either of the two Teague exceptions to the general bar on ret-roactivity. Id. We address each of these three factors in turn.
1. The Finality of Edwards’s Conviction
The first step in the Banks retroac-tivity test calls for us to determine whether Edwards’s conviction is final — a question with which we can easily dispense. Convictions in state courts are “final ‘for purposes of retroactivity analysis when the availability of direct appeal to the state courts has been exhausted and the time for filing a petition for a writ of certiorari has elapsed.’ ” Banks, 542 U.S. at 411, 124 S.Ct. 2504 (quoting Caspari v. Bohlen, 510 U.S. 383, 390, 114 S.Ct. 948, 127 L.Ed.2d 236 (1994)); see also People v. Hampton, 876 P.2d 1236, 1239 (Colo.1994) (“[A] conviction is not final and has no legal force until after appeals have been exhausted.”). We denied certiorari on Edwards’s direct appeal on May 10, 1999, thereby ending the process of direct appeals. Thus, Edwards’s conviction is final, satisfying the first prong of the retroactivity test.
2. Whether Crawford Announces a New Rule of Criminal Procedure
We now turn to the second prong of the Banks retroactivity test and consider whether the Crawford holding constitutes a new rule of criminal procedure. Edwards does not address whether the rule is procedural, but he contends that it is not new. We first address whether the rule is procedural, then analyze whether it is new.
We deemed the Crawford holding to be a procedural rule in People v. Fry, 92 P.3d 970, 976 (Colo.2004). In that opinion, we explained that the United States Supreme Court’s rendering of the Confrontation Clause in Crawford “provides a procedural, *984not a substantive, guarantee” because Crawford “commands, not that evidence be reliable, but that reliability be assessed in a particular manner: by testing in the crucible of cross-examination.” Id. (quoting Crawford, 541 U.S. at 61,124 S.Ct. 1354).
A rule is procedural rather than substantive if it regulates “only the manner of determining the defendant’s culpability.”7 Summerlin, 542 U.S. at 353, 124 S.Ct. 2519. In articulating the test to discern whether testimonial out-of-court statements can be admitted at trial, the Crawford holding regulates the admissibility of evidence, which is a manner of determining culpability and is, therefore, a procedural rule. As we noted in Fry, the language of the Crawford opinion, itself, supports this conclusion, labeling the Confrontation Clause’s ultimate goal of ensuring the reliability of evidence a procedural guarantee. Fry, 92 P.3d at 976; Crawford, 541 U.S. at 50, 124 S.Ct. 1354. Thus, we reiterate our conclusion in Fry and hold that the Crawford decision articulates a rule of criminal procedure.
Having determined that the rule announced in Crawford is procedural, we next consider whether it is a new rule. Edwards concludes that it is not, and thus should be applied to vacate his conviction. We disagree with Edwards’s analysis and hold that Crawford imparts a new rule of criminal procedure.
Edwards offers two main arguments in support of his conclusion: (1) the result in Crawford was compelled by prior precedent; and (2) Supreme Court case law prior to Crawford is consistent with Crawford’s underlying principles. In his argument, Edwards also relies upon a concurring opinion in Bockting v. Bayer, 399 F.3d 1010, 1022 (9th Cir.2005), which we address separately.
We first address Edwards’s contention that a new rule is one in which the result, rather than the rationale, was not compelled by prior precedent. To support his argument, Edwards points to language in Teague that states “a case announces a new rule if the result was not dictated by precedent existing at the time the defendant’s conviction became final.” 489 U.S. at 301, 109 S.Ct. 1060.
We read Teague as supporting the opposite conclusion that Crawford indeed announces a new rule of criminal procedure. The language quoted by Edwards emphasizes the word “dictated.” Thus, if the result in Crawford — the use of the witness’s statement at trial violated the Confrontation Clause because the defendant had not had a prior opportunity to cross-examine the witness — was not dictated by Roberts, Crawford announces a new rule. Because Roberts did not require an opportunity to cross-examine, we cannot say it dictated the result in Crawford. In fact, review of the Crawford case under Roberts did not dictate the result reached by the Supreme Court. The Washington Supreme Court decision that was reached under Roberts and overruled in Crawford found the witness’s statement was sufficiently trustworthy to be admitted at trial, a result that is obviously not dictated by Crawford. Crawford, 541 U.S. at 41, 69, 124 S.Ct. 1354.
Teague provides further support that Crawford announces a new rule. The Court begins its discussion by acknowledging the difficulty in determining whether a case announces a new rule and explaining that the opinion does not “attempt to define the spectrum of what may or may not constitute a new rule.” Teague, 489 U.S. at 301, 109 S.Ct. 1060. However, the Court gives a general description of a new rule as one that “breaks new ground or imposes a new obligation on the States or the Federal Government.” Id. By requiring that a defendant have the opportunity to cross-examine a witness against him if that witness is unavailable to testify at trial — a requirement that did not exist under the previous precedent of Roberts — Crawford imposes a new obligation on the states and federal government.
In his second argument supporting the conclusion that Crawford does not announce a new rule, Edwards relies upon the Craw*985ford Court’s observation that its prior decisions “remained faithful to the Framers’ understanding [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.” Crawford, 541 U.S. at 59, 124 S.Ct. 1354. We disagree with Edwards’s interpretation. The Court qualifies its statement with the explanation that, while the results of its decisions “have generally been faithful to the original meaning of the Confrontation Clause,” its rationales have not. Id. at 60, 124 S.Ct. 1354. This observation indicates that the outcomes, though consistent with the purposes of the Confrontation Clause, were nonetheless reached under discrepant rules. Consistent outcomes can emerge from divergent rules.
Moreover, the Crawford Court acknowledges that all its pre-Crawford precedent may not have had outcomes consistent with Crawford’s mandates. It identifies one opinion rendered under Roberts that is “arguably in tension with the rule requiring a prior opportunity for cross-examination when the proffered statement is testimonial.” Id. at 58 n. 8, 100 S.Ct. 2531 (citing White v. Illinois, 502 U.S. 346, 112 S.Ct. 736, 116 L.Ed.2d 848 (1992)).
Edwards also cites a concurring opinion in Bockting to support his contention that Crawford does not announce a new rule. We address that opinion separately because it was written by the Honorable John T. Noo-nan, Jr., a noted constitutional scholar and senior judge on the U.S. Court of Appeals for the Ninth Circuit. In his concurrence, Judge Noonan concludes — contrary to the majority opinion of the Ninth Circuit — that Crawford constitutes a rationale shift rather than a new rule. Bockting, 399 F.3d at 1023-24 (Noonan, J., concurring). Judge Noonan bases his determination on two reasons. First, he relies on the Supreme Court’s statement in Crawford that its earlier decisions were faithful to the Framers’ intent that testimonial statements of absent witnesses be admitted only if the declarant is unavailable and the defendant had a prior opportunity to conduct a cross-examination. Id. at 1023 (citing Crawford, 541 U.S. at 59, 124 S.Ct. 1354). Second, Judge Noonan refers to Summerlin, in which Justice Scalia, who penned Crawford during the same term, wrote that it is unlikely that any “new procedural rules without which the likelihood of an accurate conviction is seriously diminished” have yet emerged. Id. at 1023-24 (quoting Summerlin, 542 U.S. at 352, 124 S.Ct. 2519). We addressed Judge Noonan’s first argument above, so we turn to his second argument.
Judge Noonan asserts that the Court’s pronouncement in Summerlin, delivered by Justice Scalia shortly after he wrote the Crawford opinion, that the emergence of “new procedural rules without which the likelihood of an accurate conviction is seriously diminished” is unlikely, thwarts a conclusion that Crawford announces a new rule. We disagree. We read this portion of Summer-lin as addressing new watershed rules of criminal procedure, not merely new rules. In Summerlin, the Court addresses whether its decision in Ring v. Arizona, 536 U.S. 584, 122 S.Ct. 2428, 153 L.Ed.2d 556 (2002), applies retroactively. Summerlin, 542 U.S. at 349, 124 S.Ct. 2519. In making its determination, the Court analyzes whether Ring, which holds “that ‘a sentencing judge, sitting without a jury, [may not] find an aggravating circumstance necessary for imposition of the death penalty,’ ” constitutes a watershed rule of criminal procedure. Id. at 353, 124 S.Ct. 2519 (quoting Ring, 536 U.S. at 609, 122 S.Ct. 2428). Without considering whether the Ring holding constitutes a new rule, the Court in Summerlin simply states that it does. Id, at 352, 358, 124 S.Ct. 2519. The Court’s analysis instead centers on whether the rule is procedural and whether it is watershed. The language in Summerlin relied upon by Judge Noonan addresses the characteristics of watershed rules; it is not employed to analyze whether a rule is new.
We now depart from Judge Noonan’s arguments and make an additional observation in support of our holding that Crawford announces a new rule. We deem it significant that Crawford diverges from the prior test of Roberts. Under the Banks test, to determine whether the Court has announced a new rule of criminal procedure, we must *986ask “whether the rule ... was dictated by then-existing precedent.” Banks, 542 U.S. at 413, 124 S.Ct. 2504 (quoting Lambrix v. Singletary, 520 U.S. 518, 527-28, 117 S.Ct. 1517, 137 L.Ed.2d 771 (1997)). By adding the requirement that the party seeking to exclude the evidence must have had the opportunity for cross-examination, Crawford creates a test that was not dictated by the then-current precedent of Roberts, which conditioned admission of the evidence on a more broad requisite of trustworthiness. In fact, the Court concedes that the Roberts test diverges from the historical principles underlying the Confrontation Clause. Crawford, 541 U.S. at 60, 124 S.Ct. 1354. The Crawford concurrence also supports the view that the decision announces a new rule, asserting that “the Court of course overrules Ohio v. Roberts.” Id. at 75, 100 S.Ct. 2531 (Rehnquist, C.J., concurring). Our language in Fry is consistent with this analysis. In Fry, we did not definitively state that Crawford puts forth a new rule because the facts in that case did not demand such a specific conclusion. However, in explaining that Crawford abrogates Roberts, we noted that “Crawford rejects the reliability prong of the Roberts test.” Fry, 92 P.3d at 976.
Finally, we note that several courts have addressed this issue and determined that Crawford indeed sets forth a new rule. Brown v. Uphoff, 381 F.3d 1219, 1226 (10th Cir.2004); Murillo v. Frank, 402 F.3d 786, 790 (7th Cir.2005); Bockting, 399 F.3d at 1016; see Dorchy v. Jones, 398 F.3d 783, 788 (6th Cir.2005) (“Teague thus prohibits Dorchy from availing himself of the new rule articulated in Crawford.”)-, cf. Mungo v. Duncan, 393 F.3d 327, 335 (2d Cir.2004) (assuming Crawford announces a new rule in order to analyze whether it qualifies as a watershed rule).
Hence, based upon the language of Crawford and the fact that its holding satisfies the requirements for determining whether a rule is new, we hold that Crawford announces a new rule of criminal procedure.
3. The Teague Exceptions to the General Rule of Nonretroactivity
The final step in our retroactivity analysis requires us to determine whether the rule announced in Crawford fits an exception to the general tenet that a new constitutional rule of criminal procedure does not apply retroactively to cases on collateral review. Summerlin, 542 U.S. at 352, 124 S.Ct. 2519.
Teague outlines two exceptions to this general rule. The first exception allows a new rule to “be applied retroactively if it places ‘certain kinds of primary, private individual conduct beyond the power of the criminal law-making authority to proscribe.’ ” Teague, 489 U.S. at 311,109 S.Ct. 1060 (quoting Mackey, 401 U.S. at 692, 91 S.Ct. 1160). The second exception mandates the retroactive application of a new rule when it requires observance of “procedures that ... are ‘implicit in the concept of ordered liberty.’ ” Id. (quoting Mackey, 401 U.S. at 693, 91 S.Ct. 1171) (additional citation omitted). The first exception is not relevant here because the Crawford holding does not decriminalize a particular type of conduct. Thus, our analysis centers on the second — the watershed exception.
C. Watershed Rules of Criminal Procedure
The second exception applies to what the Supreme Court has called “watershed” rules of criminal procedure. Id. Edwards argues that Crawford fits this exception, and therefore is applicable to his case.
To support his conclusion that Crawford announces a watershed rule, Edwards cites a variety of cases characterizing the Confrontation Clause as a fundamental right essential to a fair trial. He bolsters his argument with language from Crawford itself, describing the long history and importance of what the Court terms a “bedrock procedural guarantee.” Crawford, 541 U.S at 42, 124 S.Ct. 1354. Before we address Edwards’s arguments, we first take a closer look at the category of watershed rules of criminal procedure.
A watershed rule is one that implicates the “fundamental fairness and accuracy of the criminal proceeding.” Summerlin, 542 U.S. at 352, 124 S.Ct. 2519 (citations omitted). However, in order to be consid*987ered watershed, a new rule of criminal procedure must be more than “fundamental”; it “must be one ‘without which the likelihood of an accurate conviction is seriously diminished.’ ” Id. (quoting Teague, 489 U.S. at 313, 109 S.Ct. 1060).
In Teague, the Court acknowledged the value of limiting the retroactive application of constitutional rules. Doing so is vital to effectuating finality, which is an essential component of the criminal justice system. Teague, 489 U.S. at 309, 109 S.Ct. 1060. Accordingly, the United States Supreme Court has created a very narrow definition of a watershed rule. To fall within the watershed exception, a new rule must fulfill two criteria: (1) “[infringement of the rule must seriously diminish the likelihood of obtaining an accurate conviction;” and (2) “the rule must alter our understanding of the bedrock procedural elements essential to the fairness of a proceeding.” Tyler, 533 U.S. at 665,121 S.Ct. 2478 (citations and internal quotations marks omitted). We recognize these watershed criteria are sufficiently general to accommodate a number of fundamental rules and that reaching the level of watershed distinction is a matter of degree. To illustrate what type of rule would qualify as watershed, the United States Supreme Court has cited the holding in Gideon, 372 U.S. 335, 83 S.Ct. 792, which recognizes a criminal defendant’s right to counsel in eases that involve a possible prison sentence. Saffle v. Parks, 494 U.S. 484, 495, 110 S.Ct. 1257, 108 L.Ed.2d 415 (1990). Indeed, the only decision the Supreme Court has identified as watershed is the Gideon holding. See Murillo, 402 F.3d at 790.
The notion of a watershed rule of criminal procedure has perhaps been better defined by the instances in which the Supreme Court has declined to apply the label. For example, the Court applied Teague in Summerlin when it considered whether Ring v. Arizona applied retroactively to cases that had already become final. Ring held that “a sentencing judge, sitting without a jury, [may not] find an aggravating circumstance necessary for imposition of the death penalty.” 536 U.S. at 609,122 S.Ct. 2428. In its analysis, the Court found that the type of judicial factfinding in question did not create an impermissible risk of injustice, and thus the Ring holding did not constitute a watershed rule of criminal procedure. Summerlin, 542 U.S. at 356,124 S.Ct. 2519.
Having outlined the analytical framework for determining whether a new rule of criminal procedure is applicable in postconviction proceedings, and bearing in mind the strictures of the watershed exception, we now consider whether Crawford constitutes a watershed rule of criminal procedure.
Edwards notes the strong language in Crawford supporting his conclusion that it announces a watershed rule. We agree that it is difficult to read Crawford and conclude otherwise. Beginning with pre-Roman times, Justice Scalia spends no fewer than six full pages surveying the history of this “bedrock procedural guarantee” and extolling its importance to a fair trial. Crawford, 541 U.S. at 42, 43-50, 124 S.Ct. 1354. Justice Scalia’s eloquent rendering of the Confrontation Clause’s fundamental role in providing a fair trial creates the impression that Crawford’s holding could be at the level of Gideon’s watershed ruling. However, Crawford read in conjunction with Summerlin appears to present an insurmountable hurdle to a determination that Crawford meets the watershed exception.
Summerlin explains that the watershed “class of rules is extremely narrow, and it is unlikely that any ... has yet to emerge.” 542 U.S. at 352, 124 S.Ct. 2519 (citations and internal quotation marks omitted). The plain language of this statement, combined with the fact that it was written by Justice Scalia during the same term in which he authored Crawford, creates a context in which we deem it impossible to hold that Crawford constitutes a watershed rule. And the Sum-merlin holding, itself, further indicates that the Court does not consider Crawford a watershed rule. The Court’s refusal in Sum-merlin to apply the watershed exception to the Ring holding, which subsumes the weighty matter of whether a defendant receives the death penalty or not, is a strong implication that Crawford, which regulates the admissibility of evidence, is not a watershed rule.
*988We decline to interpret the Supreme Court’s reference to the Confrontation Clause in Crawford as a “bedrock procedural guarantee” to be a designation of watershed status. Crawford, 541 U.S. at 42, 124 S.Ct. 1354. While the Clause’s guarantee of an accused’s right to confront witnesses is undoubtedly fundamental to a fair trial, the rule set forth in Crawford does not create that right. Rather, by changing its interpretation of what constitutes an adequate indicia of reliability, the Crawford Court redefines how the confrontation right is to be implemented. In contrast, the watershed rule announced in Gideon ensures that an accused will receive assistance of counsel; it does not merely define how that right must be effected. Hence, we do not view the Crawford rule that an accused must have had the opportunity to cross-examine an unavailable witness as “insuring] fundamental human rights of life and liberty” to the degree that the right to counsel does. Gideon, 372 U.S. at 343, 83 S.Ct. 792 (quoting Johnson v. Zerbst, 304 U.S. 458, 462, 58 S.Ct. 1019, 82 L.Ed. 1461 (1938)).
Edwards further argues that the fact that error under Crawford is harmless rather than structural does not preclude a conclusion that Crawford announces a watershed rule. We agree, but we note that the standard of review assigned to Confrontation Clause violations nonetheless provides some guidance in navigating this issue.
Violations of the Confrontation Clause are constitutional trial errors. Fry, 92 P.3d at 980. A constitutional trial error requires reversal only if an appellate court determines the error was not harmless beyond a reasonable doubt, that is, harmless error. Id. Constitutional errors can also be structural, meaning they affect the framework of the entire trial and require automatic reversal. Id. The United States Supreme Court has labeled total deprivation of counsel — a violation of the right guaranteed by the watershed rule in Gideon — as structural error. Johnson v. United States, 520 U.S. 461, 468-69, 117 S.Ct. 1544, 137 L.Ed.2d 718 (1997). We do not suggest that a classification of harmless error cannot coincide with watershed status. However, given the rank of Confrontation Clause violations as harmless error, it would be “difficult to conclude that the rule in Crawford alters rights fundamental to due process.” Brown, 381 F.3d at 1227.
This court has consistently followed the lead of the United States Supreme Court when determining whether a rule of criminal procedure is retroactive. Timmons, 690 P.2d at 215. Thus, we join the Tenth, Second, Sixth, and Seventh circuits in determining that Crawford is not a watershed rule of criminal procedure. Brown, 381 F.3d at 1227; Mungo, 393 F.3d at 336; Dorchy, 398 F.3d at 788; Murillo, 402 F.3d at 790. We hold that Crawford is not a watershed rule of criminal procedure and therefore does not apply retroactively to cases involving post-conviction proceedings that concern convictions finalized prior to Crawford. Accordingly, we uphold the court of appeals’ decision in People v. Edwards.
IV. Conclusion
For the reasons stated, we affirm the judgment of the court of appeals.
Justice RICE and Justice COATS do not participate.. In Crawford, the Supreme Court did not define "testimonial,” but did explain that, at a minimum, the term applies to “prior testimony at a preliminary hearing, before a grand jury, or at a former trial; and to police interrogations.” Crawford, 541 U.S. at 68, 124 S.Ct. 1354.
. We granted certiorari on the following question: Whether the court of appeals erred when it concluded that Crawford v. Washington, 541 U.S. 36, 124 S.Ct. 1354, 158 L.Ed.2d 177 (2004), does not apply to the defendant’s postconviction motion.
. Similarly, the Colorado Constitution gives an accused "the right ... to meet the witnesses against him face to face.” Colo. Const, art. II, § 16. However, because Edwards did not raise, and the court of appeals did not address, the issue of Edwards's confrontation rights under the Colorado Constitution, we do not address it here.
. We recently applied this holding in People v. Vigil, 127 P.3d 916, 2006 WL 156987 (Colo. 2006).
. Edwards argues that the victim's statements in this case were testimonial and that their admission violated his confrontation rights. But because we hold that Crawford does not apply retroactively, we need not address that issue.
. That test required an analysis of three factors: “(a) the purpose to be served by die new standards, (b) the extent of the reliance by law enforcement authorities on the old standards, and (c) the effect on the administration of justice of a retroactive application of the new standards.” Walker, 666 P.2d at 117 (quoting Stovall v. Denno, 388 U.S. 293, 297, 87 S.Ct. 1967, 18 L.Ed.2d 1199 (1967)).
. In contrast, a substantive rule is one that "alters the range of conduct or the class of persons that the law punishes." Summerlin, 542 U.S. at 353, 124 S.Ct. 2519. Generally, such rules apply retroactively. Id. at 351, 124 S.Ct. 2519.