dissenting:
On June 2, 2001, Carlisle Hall, the owner of the Ah Wee Travel Agency on Utica Avenue in Brooklyn, was shot twice in the chest at his place of business, resulting in his death. He was killed by Rasheen Harry, a 25-year old who had been released from jail just days before and who had never heard of Hall until the day of the murder. Harry undisputedly did the killing, but based on the evidence at Ellis Wood’s trial, Harry did not instigate this crime. It was the 41-year old Wood, Harry’s neighbor from Lincoln Road, who commissioned it. Wood offered to pay Harry (with whom he “[hung] out on the block” and drank Hennessy during the time they both lived on Lincoln Road) $500 for doing the deed, falsely informing Harry that he, Wood, wanted Hall dead because Hall had raped Wood’s girlfriend, Nisha Bernard, a former employee of the Ah Wee Travel Agency. It was Wood who drove Harry to the travel agency in Wood’s green Lexus and who sat in the car outside the agency with his young associate, pointing out the victim as Hall strolled down the street with a colleague, George Radix. Finally, it was Wood who provided *100Harry with the gun that he carried as he walked inside the travel agency. Moments later, the 59-year-old Carlisle Hall, the principal at the Ah Wee Travel Agency for almost 20 years, had been fatally wounded.
The majority characterizes the case against Wood — which it analyzes, I believe, too casually and without regard to the record as a whole — as “not strong.” I disagree. Harry, testifying pursuant to a cooperation agreement that nevertheless provided for his incarceration for 20 years to life, identified Wood as the person who set this crime in motion — providing him with a gun and identifying Hall as the victim' — and testified to receiving money for his role after the fact at “Juanchi’s” store — also known as the One Love Culture Shop in Brooklyn. This accomplice testimony was not just supported but fully corroborated by Wood’s girlfriend, Bernard, who testified, in substance, that Wood admitted to her that he had arranged for the murder to be committed— and to be committed by an acquaintance from Lincoln Road.
There is no evidence — none—that Bernard ever met or even knew of Rasheen Harry prior to the crime. Crucially, moreover, Bernard, who reached out to law enforcement authorities to report Wood’s involvement soon after he threatened during a fight that “he could get the same person that killed [Hall] to do the same thing to [Bernard] or [her] family,” knew details about the killing for which the only explanation in this record is the one she provided — namely, that Wood related them to her. Thus, Bernard, who was not present at the scene, knew the identity of the man, George Radix, with whom Hall walked into his store immediately prior to being shot. She knew, as well, that Hall was wearing brown pants and a burgundy shirt, that he was shot twice, and that, just as Harry testified, Harry and Wood had, prior to the murder, observed Hall and Radix from inside Wood’s green Lexus, one of two differently colored Lexus vehicles that Wood possessed. Finally, although Bernard had never met Harry, she was able to identify him in a line up after seeing him in the One Love Culture Shop several weeks after the murder, where Wood, meeting Harry in response to a request for money, specifically pointed him out to Bernard as Hall’s shooter.
The majority holds that the continued questioning of Ellis Wood following his request for a lawyer violated the prophylactic rule set out by the Supreme Court in Edwards v. Arizona, 451 U.S. 477, 101 S.Ct. 1880, 68 L.Ed.2d 378 (1981). I do not take issue with this determination. The majority further concludes, however, that the admission at trial of Wood’s exculpatory statement made pursuant to this continued questioning had a “substantial and injurious effect” on the jury’s verdict — such that the New York Supreme Court, Appellate Division, Second Department, was objectively unreasonable in determining, to the contrary, that this error was harmless. I emphatically disagree with this second conclusion.
We are now the third court to assess whether the introduction of Wood’s videotaped statement was harmless error. The Appellate Division ruled, as the majority does here, that it was error to admit the disputed statement. However, it went on to find that “[this] error was harmless beyond a reasonable doubt in light of the overwhelming evidence of the defendant’s guilt.” People v. Wood, 40 A.D.3d 663, 835 N.Y.S.2d 414, 415 (2d Dep’t 2007). The district court in this case agreed that the admission of the statement was error; applying the harmless error standard set out for use in habeas review by the Supreme Court in Brecht v. Abrahamson, 507 U.S. 619, 113 S.Ct. 1710, 123 L.Ed.2d 353 *101(1998), the court relied principally on the strength of the other evidence against Wood to conclude, as well, that the error was harmless. My own review of the record leads me to agree with the unanimous conclusion of the previous courts to consider this issue — both that the other evidence of Wood’s guilt in this case was very strong and that, assessing this factor in light of other relevant indicia of an error’s harmlessness, the introduction of the disputed exculpatory statement here was harmless. I would therefore deny Wood’s petition for a writ of habeas corpus.
As an initial matter, the majority uses this case to resolve the appropriate standard of review to apply in reviewing state court harmless error determinations, a question left open by this Court in Perkins v. Herbert, 596 F.3d 161, 176-77 (2d Cir.2010). I take no issue with the majority’s conclusion that Fry v. Pliler, 551 U.S. 112, 127 S.Ct. 2321, 168 L.Ed.2d 16 (2007), essentially “settled the debate” we referenced in Perkins, Maj. Op. at 93. Nor do I quibble with its holding that a federal court conducting habeas review of a state court judgment may appropriately apply Brecht’s “substantial and injurious effect” test in determining whether a constitutional violation was harmless, and need not additionally apply the standard articulated after passage of the Antiterrorism and Effective Death Penalty Act of 1996 (“AEDPA”) — namely, the so-called “AEBPA/Chapman” test, which requires a federal court to defer to a state court’s conclusion that a constitutional violation was harmless beyond a reasonable doubt pursuant to Chapman v. California, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967), so long as that conclusion was objectively reasonable, see, e.g., Mitchell v. Esparza, 540 U.S. 12, 18, 124 S.Ct. 7, 157 L.Ed.2d 263 (2003).
At the same time — in a case in which the majority rejects the considered view of four justices of the Appellate Division that the error in this case was harmless beyond a reasonable doubt — it is worth noting why the Supreme Court in Fry concluded that Brecht’s harmless error standard survived passage of AEDPA, and why we may appropriately apply it alone in circumstances such as these. The Fry Court concluded that the AEDPAJChapman standard was “more liberal” to habeas petitioners than the Brecht standard, and that it was implausible to assert that AEDPA, which otherwise “limited rather than expanded the availability of habeas relief,” had sub silentio replaced Brecht with a more liberal harmless error standard. Fry, 551 U.S. at 119-20, 127 S.Ct. 2321. The Court reiterated the concerns expressed in Brecht regarding the finality of state court judgments and the difficulty of retrying defendants years after the crime and found these concerns still pertinent post-AEDPA. Id. at 120 n. 2, 127 S.Ct. 2321. It also repeated an important caution from Brecht itself: that “ ‘[s]tate courts are fully qualified to identify constitutional error and evaluate its prejudicial effect on the trial process ... and ... often occupy a superior vantage point from which to evaluate the effect of trial error.’ ” Id. at 118, 127 S.Ct. 2321 (quoting Brecht, 507 U.S. at 636, 113 S.Ct. 1710).
AEDPA directs that a writ of habeas corpus may issue in circumstances like those here only when the state decision “was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States.” 28 U.S.C. § 2254(d)(1). The propriety of employing the Brecht harmless error test alone following AEDPA’s enactment— without consideration of the AEDPAJChapman standard — thus depends on *102the conclusion that the Brecht test “obviously subsumes” assessment of whether the state court’s application of Chapman was objectively reasonable. Fry, 551 U.S. at 120, 127 S.Ct. 2321. As applied by the majority, it does not. The majority enunciates our Brecht-related precedent properly' — affirming that in evaluating an error’s impact, we look both to the overall strength of the prosecution’s case and to various factors indicative of the importance of the wrongly admitted evidence, including “the prosecutor’s conduct with respect to the ... evidence,” Maj. Op. at 94 (alteration in original) (quoting Zappulla v. New York, 391 F.3d 462, 468 (2d Cir.2004)), “whether the evidence ‘bore on an issue ... plainly critical to the jury’s decision,’ and ‘whether [it] was material to the establishment of the critical fact or whether it was instead corroborated and cumulative,’ ” id. (alterations in original) (quoting Wray v. Johnson, 202 F.3d 515, 526 (2d Cir.2000)). Its flawed evaluation of the evidence, however, which it fails to consider as a whole, leads it to the wrong result in this case and to a result not in keeping with Brecht’s animating principle — that it is “the historic meaning of habeas corpus ... to afford relief to those whom society has grievously wronged,” 507 U.S. at 637, 113 S.Ct. 1710 (internal quotation marks omitted), rather than “to relitigate state trials,” id. at 633, 113 S.Ct. 1710. I address each of the factors discussed by the majority and conclude, here, that the admission of Wood’s videotaped exculpatory statement was harmless.
1. The Strength of the Prosecution’s Case
As the majority recognizes, the overall strength of the case against a convicted habeas petitioner is generally “the single most critical factor in determining whether error was harmless.” Wray, 202 F.3d at 526 (quoting Latine v. Mann, 25 F.3d 1162, 1167-68 (2d Cir.1994)). Given this recognition, it is striking that the majority, in rejecting the considered assessment of the strength of the prosecution’s case by two previous courts and five judges, concludes that the ease was not strong with but a passing nod to the mutually corroborative character of the testimony provided by Harry, the shooter, and by Bernard, Wood’s live-in girlfriend at the time of the murder and the mother of two children with him. It is precisely the interlocking and mutually corroborative character of this testimony, however, that belies the majority’s claim that “the prosecution’s case was not strong,” Maj. Op. at 94.
At trial, Harry testified that on June 2, 2001, the day of the murder, he was approached by Wood, someone he knew from the time they both had spent living in the Lincoln Road neighborhood in Brooklyn. Wood asked him if he would like to make some money in exchange for shooting a man that Wood claimed had assaulted his girlfriend. Harry testified that he agreed to Wood’s offer, and that the two drove in a green Lexus to the Ah Wee Travel Agency. Once there, Harry asked for $20 to get something to eat, which he was given, and Wood left the car to purchase marijuana. Wood returned and identified the victim, who was walking into the travel agency with another man, as the person who had assaulted his girlfriend. Harry, having discussed with Wood a price of $500 to shoot the victim (albeit without reaching agreement on the final sum), testified that he entered the travel agency and shot Hall twice in the chest, using a revolver that Wood had provided. Following the murder, Harry gave the gun to a friend named Damion and then used the subway to leave the crime scene. He talked to Wood again after the murder, asking for money, and was told to meet a man named Juanchi Hildago at his store, *103where Harry borrowed $10. He testified that he talked to Wood from time to time about receiving additional money for killing Hall, obtaining a total of $20 following the shooting.
Harry’s account was then corroborated at trial by Bernard, who related to the jury a series of conversations she had with her boyfriend, Wood, following the murder of Hall, Bernard’s former employer. The first such conversation occurred two to three weeks after the murder. Bernard testified that her mother telephoned the apartment that Wood and Bernard shared and that, when Wood answered, her mother called him a murderer. Wood and Bernard subsequently had a fight, during which he told her that “he could get the same person that killed [Bernard’s] friend to do the same thing to [her] or [her] family.” After this conversation, Bernard and her children moved out of the apartment to stay with her mother.
On July 4, 2001, Bernard brought her children back to the apartment to see their father, and again talked with Wood. She testified that Wood told her to “[l]ook [at] what he [has] done for me, for love, and I have never done anything for him.” She asked him what he meant, and he proceeded to give details of his role in procuring Hall’s murder.1 Consistent with Harry’s testimony, Bernard testified that Wood told her that he went to the travel agency with the person who actually shot Hall and pointed out Hall, who was walking into the office with “an Indian guy with a pony tail.” From this description, she recognized the other person to be George Radix, with whom she had worked.2 Bernard also testified that Wood identified the clothes Hall had been wearing the day he was shot — brown pants and a burgundy shirt — and that Wood said he was parked near the travel agency in his green Lexus while the shooting took place. Further corroborating Harry’s account, Bernard testified that Wood said the shooter shot Hall twice and that the shooter was from Lincoln Road in Brooklyn.
Bernard related to the jury a third episode with Wood that took place shortly thereafter. She was with Wood when he received a call on his cell phone from a person Wood expressly identified as the shooter, asking for money. Bernard then went with Wood to the One Love Culture Shop, a store owned by the mother of Juanchi Hildago, a friend of Wood’s. Wood pointed the shooter out to Bernard while they were both in the car, and then left to go into the store. After this episode, Bernard contacted a federal law enforcement officer — the same one who had asked her questions in June 2000, over a year earlier, about stolen credit cards — - “because [she] was frightened and scared ... [and] wanted advice from him what to do.” Bernard was later able to identify *104Harry in a line-up at the police station as the man Wood pointed out that day.
In the face of the detailed and mutually corroborative account of Wood’s role orchestrating Hall’s murder presented by Harry and Bernard, the majority supports its conclusion that the evidence against Wood was nevertheless not strong by noting that the strength of this case “rested squarely on [the witnesses’] credibility.” Maj. Op. at 94. But the majority ignores the degree to which each witness’s credibility was enhanced by the other’s testimony, and by the further fact that the record is devoid of explanation — let alone evidence to support any such explanation — as to how Bernard knew details of the crime and was able to identify Harry as the killer except as a result of Wood’s admissions.
The majority asserts that “both of these witnesses were heavily impeached.” Id. Harry’s past misconduct, however, simply fails seriously to undermine the strength of this case, given that his testimony was so powerfully corroborated by Bernard and by her knowledge of facts for which the only explanation at trial was that they were related by Wood. Moreover, even conceding that Harry is “the archetypal miscreant” described by the majority, id. at 1672, it bears observing that the most seriously damaging behavior to which the young Harry admitted at trial was the very conduct at the center of this case: Harry’s agreeing to participate in the older Wood’s scheme and to kill a man for no motive other than Wood’s promise of some money. The majority notes that Harry testified pursuant to a plea agreement that capped his prison exposure at 20 years to life (giving him some chance of not spending almost his entire adult life in jail). It suggests that “[u]nder these circumstances, a jury would be unlikely to credit Harry’s version of events absent strong corroboration.” Id. at 95. The majority ignores the reality, however, that the government frequently bases successful prosecutions on the strength of cooperating witness testimony that is considerably less fully corroborated than the testimony here. Moreover, the majority neglects to mention that Harry first admitted to killing Hall and implicated Wood in this murder-for-hire scheme on the very day of his arrest, at a time when the State could still seek the maximum penalty against him.3 Cf. Zappulla, 391 F.3d at 469 (noting, in finding a particular witness’s testimony unreliable for the purposes of its harmless error analysis, that the witness “was potentially the other murder suspect” in the case but that his account exonerated him of any guilt in the killing).
As for Bernard, I simply do not understand the majority’s remarkable suggestion that her credibility was compromised because she was not an eyewitness, but *105could only report what Wood told her. I also do not share the majority’s conclusion that it is particularly pertinent to the strength of this case: (1) that Bernard did not at first believe that the father of her children had arranged for someone’s murder; (2) that she was drawn into a fraudulent credit card scheme orchestrated by Wood; (3) that she often argued with Wood; and (4) that she did not tell Wood that she had advised police of his involvement in Hall’s murder, but rather visited him in jail during the period before trial. Indeed, the latter point seems a particularly contrived reason to deem the force of her testimony “blunted,” Maj. Op. at 96, given Bernard’s reasonable explanation that she was “afraid [for] my life, because, at once, Mr. Wood did threaten me and told me, you know, if I give any information, that he was going to get the same person that killed [Hall] to do the same thing to me.” At bottom, none of this purported impeachment is ultimately telling in the absence of an explanation as to how Bernard came to know details of the crime and the identity of the killer except through Wood. No such explanation exists on this record, and no evidence was provided to support one. I therefore agree with the trial court judge who heard Bernard testify: “She [was] a very damaging witness.”
The majority, rejecting the view of the trial judge and the five judges who have, since trial, evaluated the strength of the evidence in this case, accuses me of substituting my own assessment of the credibility of the witnesses’ accounts for the hypothetical credibility assessment that would have been performed by a jury hearing their testimony in the absence of the erroneously admitted statement: “It was for the jury to decide the credibility of the witnesses, and absent the impact of the statement, they may very well not have seen that issue in the same light as does the dissent.” Maj. Op. at 96. With respect, I have not substituted my own judgment for that of a jury. I have instead faithfully followed this Court’s admonition that in assessing the likelihood that improperly admitted evidence affected a verdict, we must undertake an objective assessment of the strength of the remaining evidence of guilt. See, e.g., Perkins, 596 F.3d at 177 (“Although harmless error analysis originally focused on whether the error had affected the jury, over the years our focus has shifted from the impact of the error on the jury’s analysis to an assessment of the strength of the remaining evidence of guilt.” (quoting United States v. Mejia, 545 F.3d 179, 199 n. 5 (2d Cir.2008))); see also United States v. Ramirez, 609 F.3d 495, 501 (2d Cir.2010) (“We have frequently stated that the strength of the government’s case is the most critical factor in assessing whether error was harmless.”).
Based on that assessment, I have no difficulty concluding that a jury would easily reject the strained and speculative leaps that are required (and that the majority embraces) in an effort to identify doubt of Wood’s guilt in the face of the testimony of Bernard and Harry. On the majority’s proposed account, Bernard in July 2001 (before anyone had been arrested for committing this crime) armed herself with details of Hall’s murder from neighborhood gossip — gossip that was so accurate and so detailed that she came to know that a green Lexus of the sort driven by Wood was, coincidentally, at the scene of the crime, and that the killer hailed from Lincoln Road. The intrepid Bernard then chose to thrust herself into a murder investigation — contacting the police herself — in order falsely to implicate Wood in the hope that in aiding in the wrongful conviction of her live-in boyfriend, she could put to rest any lingering fears of *106prosecution arising out of an arrest for credit card fraud that had occurred almost a year earlier, for which the charges had been dismissed. The majority neglects to mention in this regard that both Wood and Harry were arrested only after Bernard came forward with her account of Wood’s admissions. This detail is important, because it opens the curtain on the true acrobatics of the majority’s theory of the case. For as the majority would have it, Bernard, piecing together neighborhood gossip, was able not only persuasively (but falsely) to implicate Wood to police. She also managed to identify the Lincoln Road killer — a person there is no evidence she ever met — in a line up.4 That killer then paid her the truly remarkable courtesy of confessing and, within hours of his arrest, also falsely implicating Wood (a person he conveniently happened to know), right down to asserting that the pair drove to the scene in the green Lexus Bernard had already described to police.
With respect, these contortions amount to rejecting the strength of the government’s case, not evaluating it, as our case law demands. See United States v. Reifler, 446 F.3d 65, 87 (2d Cir.2006) (noting that of all the factors used to evaluate whether a particular error was harmless, “[t]he strength of the prosecution’s case is probably the single most critical factor” (alteration in original) (quoting Latine, 25 F.3d at 1167-68)). Wood had every incentive at trial to raise any possible avenue of impeachment in attempting to blunt the force of the combined testimony of Harry and Bernard. He failed to do so in any meaningful way because he could not — and that remains the case whether his erroneously admitted exculpatory statement is in evidence or not.
The majority suggests, finally, that we may presume the case was weak — or at least that the jurors were in disagreement — because the jury deliberated into a third day. Suffice it to say that I do not find this claim persuasive. Even when a trial results in a hung jury (a fact conclusive on the subject of dissension among jurors) we have cautioned against drawing broad inferences about the strength of the case, noting that “[a] jury may hang for any number of reasons, including the idiosyncratic views of a single juror.” United States v. Newton, 369 F.3d 659, 680 (2d Cir.2004). The same holds for the length of jury deliberations. The majority makes much of the fact that the jurors asked to see Wood’s statement replayed. But the jurors also asked to hear Bernard’s testimony during their deliberations, along with selected portions of Harry’s, as well as to view Harry’s videotaped confession and his two other statements to police (but not his Wood-dictated “Affidavit in Support of Statement”). This suggests that jurors took time because they were paying scrupulous attention to the totality of the evidence in this case — evidence that powerfully implicated Wood, taking no account of his erroneously admitted denial of the crime.
In sum, after carefully reviewing Bernard’s account of Wood’s admissions to her, in combination with Harry’s own detailed description of his and Wood’s interactions as Wood solicited his help in killing Hall, I am in full agreement with the five judges who have considered this case already and have concluded that the evidence was very strong. The testimony of Bernard and Harry together constituted a fully corroborated and coherent narrative of guilt that was powerfully incriminating, *107irrespective of Wood’s statement denying culpability. The majority’s claim to the contrary disregards the interlocking character of the testimony offered by Bernard and Harry and the strong indicia of reliability created by these two independent sources of incrimination. I have no hesitation in deeming this case strong, not weak, a factor that weighs heavily in favor of a finding of harmlessness.
2. The Importance of the Improperly Admitted Evidence
The next factor to be considered is the importance of the erroneously admitted evidence. See Wray, 202 F.3d at 526. The disputed statement, as the majority notes, was not a confession, but rather an account of Hall’s murder intended by Wood to be exculpatory with respect to his own role in procuring it. In his approximately 15 minute statement (made to an Assistant District Attorney after Wood was apprised of his Miranda rights, see Miranda v. Arizona, 384 U.S. 436, 86 S.Ct. 1602, 16 L.Ed.2d 694 (1966), for at least the second time), Wood asserted that he had been at the scene of the crime, that he saw Harry and a friend in a Honda there, and that he, Wood, in making conversation, pointed out Hall, who was walking down the street, as someone with whom his girlfriend had experienced problems. A few minutes later, he heard two gunshots and then saw Harry coming around the corner. Subsequent to Hall’s murder, Wood stated that he gave Harry a little money occasionally and “a bag of weed” in response to his requests. While he did not claim to have actually seen Harry commit the murder, Wood stated that he believed Harry had done so, and that he sent Harry away from the video store he owned, calling Harry “crazy.” Wood claimed that after the crime, he “tried to don’t even have conversations with [Harry]. I didn’t trust him any more. I didn’t know if he was gonna shoot me.”
Wood’s videotaped statement was thus an account of Hall’s murder intended by him to be exculpatory, and not a confession to his own role in orchestrating the crime. To the extent it bore on “the central disputed issue,” Wray, 202 F.3d at 530, whether Wood “procured commission of [Hall’s] killing pursuant to an agreement with a person other than the intended victim” in exchange for something of value, N.Y. Penal Law § 125.27(1)(a)(vi), it constituted a denial by Wood that he committed the crime.
The majority asserts that the videotape nevertheless had a substantial, injurious influence at this trial because it “ ‘locked [Wood] into’ an implausible and highly incriminating depiction of the shooting.” Maj. Op. at 1675. The admissible evidence, however, had already put Wood in the lockbox that the majority describes. As outlined previously, to create a reasonable doubt in the mind of the jury, Wood faced the difficult task of explaining why the admitted shooter would identify him as the procurer of the murder-for-hire — at a time when in doing so he implicated himself in a first degree murder with no cooperation agreement in place — or why Wood’s then-girlfriend would corroborate this account with detailed testimony relating Wood’s admissions to her regarding his role. And, again, Wood needed some theory as to how Bernard even knew the details of the crime she related and was able to pick Harry out of a line up if her account of Wood’s admissions was in any way untrue.
Indeed, Wood’s videotaped statement was helpful to the defense to the extent it suggested an otherwise missing explanation as to how Bernard came upon her knowledge of the crime — namely, through supposedly truthful statements made by *108Wood, who was present at the scene. The defense, arguing that “[w]hen it was in [Bernard’s] best interest to get rid of Ellis Wood, she had no problem going to the authorities,” played the videotape during its own summation, using it in precisely this way:
Now, again, Ms. Bernard is telling you basically about what she thinks were some conversations she had with Ellis Wood. And she’s telling you about what she thinks were some observations that she made about Ellis Wood having some kind of meeting with Rasheen Harry.
In a couple of minutes, or at some point today before I sit down, and close to the end when I sit down, you will know when you see the video tape that ... I will not be talking to you much beyond that.
But, in that video tape, Ellis Wood speaks to Detective Arnao, and tells Detective Arnao what Ellis Wood got to say about what happened on June 2nd.
When you watch that video tape, I’m going to ask you to watch very, very carefully.... [W]e will prove to you that what my client says to you in that video tape is accurate and that it’s truthful and that you can rely on it.
The fact of the matter is it’s something you can rely on a whole lot more than anything you’re hearing from Rasheen Harry and anything you’re hearing from Nisha Bernard.
To be clear, I do not take issue with the majority’s conclusion that the admission of Wood’s exculpatory statement was harmful to the defense in various ways — principally, by further corroborating the facts that Wood was present at the scene, that he identified Hall to Harry, and that he paid Harry after the fact. To the extent it locked Wood into an implausible story, however, the record is devoid of any indication that without the videotape, there was any more plausible tale to be told. Thus, the majority suggests that the videotape was “devastating to the defense argument that Harry was an irresponsible fabricator.” Maj. Op. at 97. But Bernard’s testimony - bolstered Harry’s credibility more than the videotape since, after all, she was the person to whom Wood, in substance, confessed. On this record, the only explanation as to how Harry and Bernard came to offer mutually corroborative accounts of a crime at which only one of them was present comes down to Wood himself — the common denominator between them. This fact renders their accounts highly persuasive and any supposed alternative defense that Wood could have presented, absent the videotape, wholly ephemeral.
The majority argues that the prosecution made substantial use of Wood’s videotaped statement in its closing argument, and that this factor weighs against a finding of harmlessness. While the prosecution’s early focus in this respect was to some degree necessitated by the defense counsel’s own, previous use of the videotaped statement in dramatically closing its summation, I do not dispute that the prosecution referred to and made use of Wood’s statement. The prosecution did not dwell on this statement alone, however, but instead wove it into the coherent narrative created by the testimony of the two principal witnesses. See Gutierrez v. McGinnis, 389 F.3d 300, 309 (2d Cir.2004) (noting, in finding harmless the improper admission of a 911 tape, that “the prosecutor highlighted the [evidence] as only one of several important pieces of evidence that the prosecutor stressed during his lengthy summation”).
Thus, while the majority claims that Wood’s statement was crucial to the establishment of Harry’s credibility, the prosecution noted numerous reasons present in *109the record for the jury to credit Harry’s testimony — from the lack of any other plausible motive for Harry to kill Hall other than the promise of payment, to the fact that Harry’s cooperation agreement still left him serving a significant sentence and that he had implicated Wood, as well as himself, well before receiving the agreement. The prosecution also observed, importantly, that “[Harry’s] testimony doesn’t stand on its own”:
[Bernard] takes her kids back to see [Wood] on the 4th of July. When she is there, he starts laying the whole thing down and, ladies and gentlemen, he provides her with details that she couldn’t have known from any other source other than a person who was there and set it up.
... [I]n July ... Mr. Wood picks her up and drives her to Juanchi’s store and says, “We’re going there because I am going to pay the man who killed Mr. Hall. He wants money....”
... Mr. Wood points out that man....
And you know that he did because, ladies and gentlemen, Nisha Bernard looked at a police line-up and out of six people that she had never seen before, she picked out Rasheen Harry accurately and correctly.
Because Rasheen Harry is the gunman. He’s the killer. How could she possibly have done that if Mr. Wood, in fact, had not pointed him out that day? There is no answer for that other than the fact that he did, he did tell her that it was the man.
And, again, the police didn’t come looking for her. She went looking for them....
The prosecution’s discussion of Wood’s videotaped statement thus occurred in the context of a summation that focused principally on the testimony of Harry and Bernard. Even if the majority is right in concluding that the videotape was given a “prominent place,” Maj. Op. at 98, moreover, the prosecution’s conduct in referring to erroneously admitted evidence simply “does not weigh as heavily in our [harmless error] analysis as the overall strength of the prosecution’s remaining case,” Perkins, 596 F.3d at 179; see also id. (concluding that even though prosecutor made numerous references to improperly admitted confession during summation, the error in admitting the confession was harmless in light of the strength of remaining evidence establishing guilt). And contrary to what the majority contends, this case was strong indeed.
The final factor to consider is whether the videotape was corroborative and cumulative of other properly admitted evidence. The district court found that it was, as it did not establish any facts not already established by other testimony but, instead, primarily corroborated Harry’s testimony. The majority does not disagree, but claims that “where guilt rests on witness credibility, key evidence affecting credibility is not merely corroborative or cumulative: by permitting the jury to credit otherwise suspect testimony, it provides a key link in the prosecution’s case.” Maj. Op. at 99. While the majority quotes approvingly the language in Zappulla that evidence is not cumulative when it “fill[s] in a missing link,” Zappulla, 391 F.3d at 472, however, Zappulla involved an improperly admitted statement that constituted the only evidence of motive the prosecution had, see id. at 473. In contrast, here, the majority can point to nothing in Wood’s statement not otherwise before the jury and is forced instead to rely on its effect in bolstering the credibility of Harry, in which respect the statement was nevertheless still cumulative of Bernard’s *110testimony. See Perkins, 596 F.3d at 179 (finding a defendant’s recorded statement cumulative when properly admitted oral statement from defendant stated essentially the same facts); see also Newton, 369 F.3d at 679-80 (noting in finding admission of a statement harmless that the defendant “had essentially acknowledged” the same fact evidenced by the disputed statement in an admissible statement that was itself corroborated by other testimony). I agree with the district court that the statement was in large part cumulative of other evidence properly introduced at trial, and that this fact weighs against Wood’s claim that the error was not harmless.
* * *
While the factors considered above may not uniformly point to finding the error harmless, it is the decisive strength of the prosecution’s case, “the most important factor in our inquiry,” Perkins, 596 F.3d at 179, that leads me to conclude that the admission of Wood’s videotaped statement did not have a “substantial and injurious effect or influence in determining the jury’s verdict.” Brecht, 507 U.S. at 637, 113 S.Ct. 1710 (quoting Kotteakos v. United States, 328 U.S. 750, 776, 66 S.Ct. 1239, 90 L.Ed. 1557 (1946)). Even without its admission, the detailed and mutually corroborative testimony of Harry and Bernard clearly established Wood’s role in the murder-for-hire of the victim Carlisle Hall. Moreover, Wood’s statement, intended by him to be exculpatory, contained no facts not already in evidence through the other testimony presented at trial, and even its effect in strengthening the credibility of Harry was cumulative of Bernard’s testimony. The majority misapplies Brecht in this case and rejects as objectively unreasonable the eminently reasonable judgment of the Appellate Division that the error here was harmless. I respectfully dissent from the court’s grant of a writ of habeas corpus on this appeal.
. The State’s theory at trial was that Wood’s true motive was not love, but self preservation. Bernard was employed at Hall's travel agency when she first met Wood and took up with him. During her employment there, Wood regularly sent in people to purchase airline tickets with stolen credit card information in order to pocket the proceeds. Bernard was approached by federal law enforcement agents in June 2000 and was asked questions related to Wood. She was herself arrested by local authorities for her involvement in August 2000 (and on the complaint of her employer, Hall), but the charges were not pursued. With both federal and state authorities inquiring into his fraudulent scheme, Wood had an interest in ensuring that he was not further implicated by Hall.
. Kelly Fleary, another employee of the travel agency who was working on the day of the murder, further corroborated Bernard’s testimony on this point, testifying at trial that just before his murder Hall had arrived at the agency with Radix.
. Harry initially denied committing the crime, claiming first that Wood shot Hall and then that Wood forced Harry to commit the crime by threatening his family. Within about six hours of his arrest, however, he admitted to killing Hall in exchange for the promise of money from Wood. The majority points to the supposedly impeaching effect of a later affidavit in which Harry denied Wood's involvement and wrote that "Ellis paid me (Rasheen Harry) no monetary wage for said commission of murder.” But the majority neglects to mention that this affidavit was prepared while both Hariy and Wood were imprisoned together awaiting trial. Harry testified that he signed it as a result of Wood's "pressuring” him over the course of three weeks, and that the words, though written by him, were provided by Wood. At trial, in response to the question, “Is, ‘monetary wage,’ a term that you have ever used in your life?” Harry answered, "I used it once, but after once, I ain't used it no more.” A close read of Harry’s testimony strongly suggests that Harry’s "Affidavit in Support of Statement” was more likely to have bolstered his credibility than impeached it.
. To be clear, there is also no evidence that Bernard knew or had ever heard of Harry's girlfriend, a person referred to only fleetingly in Harry’s testimony and alluded to by the majority as the potential source for some of the gossip that Bernard supposedly picked up.