concurring:
The majority has joined in a thoughtful and thorough judgment in this case. I concur in that judgment. The majority has forcefully marshaled the facts and analyzed the applicable law. If this case had arisen entirely within the federal system, e.g., under 28 U.S.C. § 2255, I would fully join in the opinion and have no further comments or suggestions.
This case does not arise entirely within the federal system, however, but comes to us from state court under § 2254, and thus our review must be guided by the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA). AEDPA requires us to give deference to decisions of the state courts both as to the facts and the law, except in limited circumstances.
The majority identifies Missouri v. Seibert, 542 U.S. 600, 124 S.Ct. 2601, 159 L.Ed.2d 643 (2004), as clearly established federal law at the time the California Court of Appeal decided this case and looks to Justice Kennedy’s concurring opinion to provide the appropriate rule of decision. I agree. See United States v. Williams, 435 F.3d 1148, 1157-59 (9th Cir.2006) (analyzing Seibert and concluding that Justice Kennedy’s opinion is the narrowest holding obtaining five votes). In the majority’s view, the California Court of Appeal recognized Seibert as controlling but unreasonably interpreted it, effectively ignoring Seibert and deciding the case under the prior law set out in Oregon v. *835Elstad, 470 U.S. 298, 105 S.Ct. 1285, 84 L.Ed.2d 222 (1985). The majority concludes that, since the California court failed to apply clearly established law, we may withhold any deference and exercise our independent judgment in reviewing the California Court of Appeal’s decision de novo.
I disagree. My review of the record leads me to conclude that the California Court of Appeal’s conclusions of law are in conformity with Seibert and Justice Kennedy’s concurrence. In my view, the state court’s approach to the law is sound, but its finding of facts are unreasonable in context. I therefore, on this alternate ground, join in this Court’s judgment. My reasons are as follows.
In reaching a contrary view, the majority focuses exclusively on one part of Justice Kennedy’s concurrence. Seibert addresses what we have termed a two-step interrogation leading to a confession. The first part of the interrogation is unwarned. Once the suspect confesses, Miranda warnings are given, and the interrogation resumes. Typically, the suspect confirms his confession. In practice, the earlier confession is suppressed, but following Elstad, the subsequent post-warning statement is allowed into evidence if it is voluntary. Seibert modified Elstad in cases such as this. The opinion was fragmented. Justice Souter wrote an opinion for a plurality of four justices. Justice Kennedy separately concurred, arguably on a more' limited basis, and Justice O’Connor wrote a dissent in which three other justices joined. The majority and I agree that Justice Kennedy’s opinion provides the holding of Seibert. We disagree on how Seibert should be applied to this case.
In my view, Justice Kennedy adopts all of Justice Souter’s plurality opinion but imposes a limitation. Rightly understood, Justice Kennedy’s opinion adopts a two-part test for determining the validity of a confession where the police use a two-step approach in their interrogation of a suspect. The first prong, which I will call the Kennedy prong, asks whether the two-step procedure was chosen intentionally in order to render subsequent Miranda warnings ineffective. The second prong, which Í will call the Souter prong, asks if the two-step process, whether or not intentional, rendered the subsequent warnings ineffective. For Justice Kennedy, both prongs must be satisfied in order to create a Seibert violation and take the case out of Oregon v. Elstad. Since the test has two prongs, it is analogous to the Strickland test for determining ineffective assistance of counsel.1 Like that test, a reviewing court should be able to look to either prong first, and, if that prong is not satisfied, there is no need to address the other prong. See Pearson v. Callahan, 555 U.S. 228, 241, 129 S.Ct. 808, 172 L.Ed.2d 565 (2009). Viewed in this light, the California Court of Appeal’s decision is within the law. That court quoted Justice Souter directly for the second prong of the test and concluded that nothing in the record of the interrogation “challenged the comprehensibility and efficacy of the Miranda warnings to the point that a reasonable person in the suspect’s shoes would not have understood them to convey a message that [he] retained a choice about continuing to talk.”
Justice Kennedy’s test requires that any two-step procedure must be intentionally motivated to undermine Miranda. Justice Souter rejects reference to the police intent and focuses only on whether the process itself challenged the comprehensibility and effectiveness of the Miranda warnings. Thus, if as the California Court of *836Appeal found, Reyes understood at the time of the second stage of the interview that he retained a choice about whether to talk, his decision to talk was consistent with Seibert, and there was no need for the Court of Appeal to address the first prong of the test and determine whether the police intended to nullify the Miranda warnings.
The Court of Appeal did not separately address the first prong of the test, and so by analogy to Strickland, we may review that prong de novo and exercise independent judgment.2 See Porter v. McCollum, 558 U.S. 30, 39, 130 S.Ct. 447, 175 L.Ed.2d 398 (2009); Mann v. Ryan, 774 F.3d 1203, 1215 (9th Cir.2014) (“Because the state post-conviction court did not reach the deficiency prong of the Strickland analysis, our review of this prong is not circumscribed by AEDPA.”). I agree with the majority that, under the facts of this case, the use of the two-step procedure was a conscious effort to undermine Miranda. See Williams, 435 F.3d at 1158-60 (discussing how a court should determine whether an interrogation was deliberately used to undermine Miranda).
The second prong of the test was addressed by the Court of Appeal, and we must grant deference and may only reject the finding if it 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)(2).
The majority points out that the Court of Appeal did not explain its factual finding that the mid-stream Miranda warning was not undermined by the interrogation procedure chosen by the police. In such a case, we must look to all of the relevant evidence to determine if the fact finding is reasonably supported. See Delgado v. Lewis, 223 F.3d 976, 982 (9th Cir.2000) (“Federal habeas review is not de novo when the state court does not supply reasoning for its decision, but an independent review of the record is required to determine whether the state court clearly erred in its application of controlling federal law.”). Reyes has a heavy burden to establish that a state court fact finding is unreasonable. See Burt v. Titlow, — U.S.-, 134 S.Ct. 10, 15-16, 187 L.Ed.2d 348 (2013). I am satisfied that he has sustained that burden here.
The majority has summarized all of the relevant evidence, and it need not be repeated here. Guided by Williams regarding a determination of the effectiveness of mid-stream Miranda warnings, see Williams, 435 F.3d at 1160-62, I do not believe that there is substantial evidence that would support an inference that the warnings given to Reyes were effective at the time they were given. There is no evidence of the corrective measures identified by Justice Kennedy. Turning to the factors considered relevant by the plurality: 1) the pre-warning interrogation was complete and detailed, consuming many hours; 2) the two rounds of interrogation overlapped; 3) the two rounds of interrogation were close in time; 4) there was a continuity of police personnel; and, most importantly, 5) the interrogator’s questions treated the second round of interrogation as continuous with the first.
*837Viewed in light of the totality of the circumstances, there is no evidence that would permit a finding that the two-step interrogation in this case did not undermine Miranda and lead to an involuntary confession.
. Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984).
. It is clear that the Court of Appeal did not think that the procedure followed undermined the effectiveness of the mid-stream Miranda warnings. It is not clear that the Court of Appeal considered whether the procedure was chosen to undermine Miranda. Even if I were to assume that the Court of Appeal found no intent, sub silentio, requiring deference I would still conclude under Williams that the procedure was chosen intentionally. See Williams, 435 F.3d at 1160 (discussing how to prove a deliberate choice of procedure).