BATCHELDER, J., delivered the opinion of the court, in which NORRIS, J., joined. BRIGHT, J. (p. 378), delivered a separate opinion concurring in the result.
BATCHELDER, Circuit Judge.Petitioner-Appellee Larry Nevers and Walter Budzyn, Detroit Police Officers, were convicted of second degree murder in a Michigan state court in the beating death of Malice Green. They were tried jointly but by separate juries.1 In their consolidated appeal as of right they raised claims concerning, inter alia, extrinsic influences on the jury and pre-trial publicity. People v. Budzyn, Nos. 170477, 170478, slip op. (Mich.Ct. App. March 22, 1995) (unpublished per curiam). The Michigan Court of Appeals affirmed their convictions. Id. The Michigan Supreme Court affirmed the court of appeals on the basis of the court’s opinion except for the extrinsic jury influences issue, finding as to that issue “a real and substantial possibility that these external influences together could have .affected the juries’ verdicts.” People v. Budzyn, 456 Mich. 77, 566 N.W.2d 229, 240 (1997). After performing a harmless error analysis, however, the court affirmed Nevers’s conviction by finding overwhelming evidence of guilt, but reversed Budzyn’s conviction and remanded his case for a new trial. Id. at 240-43.
Nevers petitioned for a writ of habeas corpus. The district court issued the writ, holding both that the trial court’s decision not to grant a change of venue due to pervasive pre-trial publicity was “manifest error,” Nevers v. Killinger, 990 F.Supp. 844, 855-64 (E.D.Mich.1997), and that the extraneous influences on the jury constituted constitutional error which was not “harmless,” id. at 864-74. For the reasons stated below, we affirm the judgment of the district court granting the writ. We affirm the reasoning of the district court granting the writ on the claim that Nevers was denied a fair trial because of extraneous influences on the jury; and we reverse the reasoning and conclusions of the district court on the claim that Nevers was denied a fair trial because of pretrial publicity.
BACKGROUND
The facts of the beating of Malice Green are set out in the published opinions of both the federal district court and the Michigan Supreme Court,2 and we need not exhaustively detail them again. Essential to the petition for habeas corpus before us on appeal are the facts that follow.
Nevers and his partner, Budzyn, both of whom are white, were on plainclothes duty after dark in an unmarked car when they observed Malice Green, an African-American, driving a bullet-riddled car carrying one passenger, pull up in front of a house known to the officers as one used for drug activity. *355Nevers and Budzyn initiated an investigation of the car and its occupants, and Nevers asked Green for his driver’s license. Green did not respond to Nevers’s request but instead went around to the passenger’s side of the car, sat down in the car with his legs hanging out of the open passenger door, and began to rummage around in the glove compartment. When Budzyn shined his flashlight on Green and repeated Nevers’s request to see Green’s driver’s license, Green reached down and grabbed something that had apparently fallen from the glove compartment onto the floor of the car. Budzyn, suspecting that the object was cocaine, asked Green what he was holding and asked him to relinquish it. Green neither responded nor complied. At this point Budzyn and Nevers undertook to force Green to give up what he was holding, and in the ensuing struggle, Nevers struck Green repeatedly on the hands and head with a flashlight.
During the course of the incident a number of other police officers and two EMS crews arrived. According to the first two technicians to arrive on the scene, they were driving by and saw Green hanging out of the driver’s side of the car, blood streaming from his head and puddling on the ground. Nev-ers was holding Green with one hand and holding a flashlight with the other, ordering Green to hold still and open his hand. These two technicians agreed that Green looked dazed, that he was squirming and moving around but not attempting to fight Nevers off, that Green did not comply with Nevers’s orders to open his hand, and that Nevers struck Green on the head with the flashlight. Their accounts of the number of blows to the head varied; one technician said he saw Nev-ers deliver four blows; the other testified to five or six. One of these technicians testified that as soon as he arrived at the scene he approached Nevers and asked him what had happened, and that Nevers replied, “I hit him. And if he doesn’t quit it, I’m gonna hit him again.” The testimony of the second EMS crew, who arrived somewhat later, confirms the general picture painted by the first crew; one of the second crew, however, estimated that Nevers delivered ten blows to Green’s head.
Whether Green attempted to resist being handcuffed, and who did what to Green during the process of handcuffing him, are the subject of considerable variations in the testimony. At least one witness, however, testified that he saw car keys in Green’s hand, and all of the testimony is consistent that Green was repeatedly ordered to drop whatever he held in his hand but refused to do so. Various of the witnesses testified that Green was struck, punched and kicked during and after the handcuffing, and some of those blows were attributed to Nevers. Finally, all of the EMS technicians agreed that shortly after the officers succeeded in handcuffing Green, he had a seizure or seizures and, despite the ministrations by the medical technicians, he died at the scene.
Nevers testified in his own defense. According to his testimony, he was talking with one of the civilians at the scene when Budzyn began to struggle with Green. Nevers ran to the passenger side of the car with his flashlight in hand. Hearing from Budzyn that Green had something in his hand, Nevers pried Green’s clenched fist open. Something which Nevers thought was a rock of cocaine fell out, and Green closed his fist again. Green began to bring his knees up toward Nevers’s chest to prevent Nevers from again opening Green’s hand, and Nevers struck him on his knees “a couple of times” in an attempt to keep Green from kneeing him. Nevers then grabbed Green’s hand and began to strike it, with each strike telling Green to open his hand. Worried that the gathering crowd might jump into the fray, Nevers told the crowd they could all leave. At this point, Budzyn said that Green was trying to get out of the car, so Nevers ran around to the driver’s side, getting there as the door was beginning to open. Nevers pulled the door open and Green’s head and torso fell out of the door. Nevers was holding Green by his clothing, when Green grabbed the handle of Nevers’s holstered gun, and Nevers hit him on the head. Nev-ers testified that on an earlier occasion his gun had been taken from him by a suspect he was attempting to apprehend and Nevers did not intend ever to be in that position again. After Green let go of his gun, Nevers said, he did not hit Green again. Seeing an ap*356proaching EMS truck, Nevers signaled it to stop. Green then began flailing his left arm around inside the car and when Nevers grabbed it, Green began swinging at him with his clenched right fist, which held something shiny between the fingers. Nevers explained that he feared that the shiny object in Green’s hand might have been a knife or razor blade, and he therefore responded by striking Green again on the head, and telling him to open his hand, drop the object, and stop struggling. Nevers said that he struck Green in the head a total of five or six times over the course of the entire incident, and that two of those blows were struck after the first EMS crew arrived.
At trial, the state introduced the testimony of Dr. Kalil Jiraki, an assistant Wayne County medical examiner, who testified that Green had suffered at least fourteen separate blunt force blows to the head, and that the cause of death was blunt force trauma to the head. Dr. Jiraki acknowledged that Green was under the influence of cocaine when he died, but stated that the amount of cocaine in his blood probably did not contribute to his death.
Nevers introduced his own experts to contradict Dr. Jiraki. One of them, Dr. L.G. Dragovic, testified “that [the] blunt force trauma that was sustained to the brain was not and cannot be taken as a sole cause of death in [this] case.” J.A. at 1116-17 (Dr. Dragovic Test.). Rather, Dr. Dragovic maintained that without the cocaine and the chemical produced when alcohol and cocaine are in the body together, both of which were found in Green’s bloodstream, Green would not have suffered a life-threatening seizure due to the blunt force injuries to his head. Dr. Dragovic also testified that there were only eleven blunt force injuries to the brain, rather than fourteen.
A barrage of media publicity began with the first reports of Green’s death. Because this incident occurred shortly after the Los Angeles riots that followed the acquittal of the police officers who had beaten Rodney King, the media reports in Detroit compared Green’s death to the Rodney King beating and were critical of the police department generally and these police officers specifically-
Budzyn and Nevers were charged with second degree murder. Before the commencement of their trial (which began seven months after Green’s beating and death), the Detroit Police Department, without any investigation or trial, fired all of the officers who had appeared at the scene, even those who merely responded to the Officer in Distress call that had gone out. During this time period “[t]he City of Detroit also agreed to a multimillion dollar settlement with Green’s estate. In response to some criticisms of the settlement, a city attorney stated that a generous settlement might spare the city the riotous violence that racked Los Angeles after the acquittal of the police officers.” Budzyn, 566 N.W.2d at 234.
Because of the pretrial publicity, Nevers sought but was denied change of venue. Instead, the trial court permitted extensive voir dire regarding the potential jurors’ familiarity with the case, including whether they were biased or prejudiced or had any preconceived notions as to guilt or innocence. This voir dire began on June 2, 1993, and lasted at least through June 15.
Trial began on June 18, 1993, and lasted approximately seven weeks. Near the end of the trial, approximately one week before the juries began deliberations, the trial court provided the juries with several film videos with which to entertain themselves during a period when they were not in court but were required to be in the courthouse. One of the movies provided was “Malcolm X,” which begins with a video of the Rodney Kang beating accompanied by a voice-over of a racially provocative and highly inflammatory speech by Malcolm X charging the “white man” with being one of the greatest murderers in history. Defendants asked for mistrial on this basis but their motion was denied by the independent judge to whom it was referred.3
*357At the close of trial the Nevers jury deliberated nine days before returning a guilty verdict. Nevers received a sentence of twelve to twenty years imprisonment.4 After the verdict was rendered, Nevers and Bud-zyn presented to the trial court affidavits from several jurors containing those jurors’ statements that in addition to the Malcolm X video, other extraneous information had reached the jury during the trial and jury deliberations, including information that Nevers and Budzyn had allegedly been involved in a police undercover unit called STRESS, reputed to have engaged in harassment of young black men. The trial court refused to hold an evidentiary hearing with regard to the claims of Nevers and Budzyn that the jury had been tainted by this information. The defendant’s motion for a new trial was denied. .
I. THE ANTITERRORISM AND EFFECTIVE DEATH PENALTY ACT OF 1996
The Antiterrorism and Effective Death Penalty Act of 1996, Pub.L. No. 104-132, 110 Stat. 1214 (1996) (“AEDPA”), applies to this-case because Nevers filed his petition for habeas relief on October 8, 1997, well after the act’s effective date of April 26, 1996. See Harpster v. Ohio, 128 F.3d 322, 326 (6th Cir.1997), cert. denied, — U.S. -, 118 S.Ct. 1044, 140 L.Ed.2d 109 (1998). The AEDPA amended federal habeas law by, among other things, changing § 2254(d) of the habeas statute to provide as follows:
(d) An application for a writ of habeas corpus on behalf of a person in custody pursuant to the judgment of a State court shall not be granted with respect to any claim that was adjudicated on the merits in State court proceedings unless the adjudication of the claim—
(1) resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States; or
(2) resulted in a decision that 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).
The district court applied the AEDPA and concluded that the Michigan Supreme Court’s decision affirming Nevers’s conviction “resulted in a decision that ... involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States” on two separate bases. First, it found the trial so permeated by prejudicial pretrial publicity that the trial court’s denial of a change of venue violated the principles enunciated in the Supreme Court’s pretrial publicity line of cases. E.g., Irvin v. Dowd, 366 U.S. 717, 81 S.Ct. 1639, 6 L.Ed.2d 751 (1961); Rideau v. Louisiana, 373 U.S. 723, 83 S.Ct. 1417, 10 L.Ed.2d 663 (1963); Estes v. Texas, 381 U.S. 532, 85 S.Ct. 1628, 14 L.Ed.2d 543 (1965); Sheppard v. Maxwell, 384 U.S. 333, 86 S.Ct. 1507, 16 L.Ed.2d 600 (1966); Murphy v. Florida, 421 U.S. 794, 95 S.Ct. 2031, 44 L.Ed.2d 589 (1975); and Patton v. Yount, 467 U.S. 1025, 104 S.Ct. 2885, 81 L.Ed.2d 847 (1984). See Nevers, 990 F.Supp. at 853-64. Second, it agreed with the Michigan Supreme Court that the extraneous influences on the jury resulted in constitutional error and that such errors were subject to harmless error analysis, but concluded that the evidence of Nevers’s guilt was not so overwhelming as to justify a finding of harmless error. See Nevers, 990 F.Supp. at 864-74. We review de novo the legal conclusions involved in the district court’s decision to grant or deny the writ under § 2254, but review for clear error its findings of fact. DeLisle v. Rivers, 161 F.3d 370, 380, 1998 WL 817815 (6th Cir.1998) (en banc).
As the First Circuit recently recognized, “AEDPA is hardly a model of clarity, ... *358and its standard of review provision is far from self-explicating.” O’Brien v. Dubois, 145 F.3d 16, 20 (1st Cir.1998) (citation omitted). The truth of this observation is amply demonstrated by the attempts of the circuits to articulate the standards the courts should use in applying the amended § 2254(d).
The Fifth Circuit reads § 2254(d) as addressing three distinct challenges to a state court’s decision. See, e.g., Drinkard v. Johnson, 97 F.3d 751, 767 (5th Cir.1996), cert. denied, 520 U.S. 1107, 117 S.Ct. 1114, 137 L.Ed.2d 315 (1997), overruled on other grounds by Lindh v. Murphy, 521 U.S. 320, 117 S.Ct. 2059, 138 L.Ed.2d 481 (1997). The “contrary to” clause in § 2254(d)(1) addresses questions of pure law; the “unreasonable application” clause in § 2254(d)(1) addresses mixed questions of law and fact; and § 2254(d)(2) addresses questions of pure fact. See id. at 767-68. The Seventh Circuit has articulated an approach very similar to that of the Fifth Circuit. See Lindh v. Murphy, 96 F.3d 856, 870 (7th Cir.1996) (en banc), rev’d on other grounds, 521 U.S. 320, 117 S.Ct. 2059,138 L.Ed.2d 481 (1997).5
In Neelley v. Nagle, 138 F.3d 917 (11th Cir.1998), the Eleventh Circuit developed an approach similar to that of the Fifth and Seventh Circuits. It first “determine^] the ‘clearly established’ law at the relevant time,” id. at 922, finding helpful guidance for this determination in Teague v. Lane, 489 U.S. 288, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989) (new rules should not be applied retroactively in reviewing habeas petitions) and its progeny, id. at 922-23.6 It then applies either the “contrary to” or “unreasonable application” clause in § 2254(d), the former to “errors of pure law,” id. at 923, and the latter “[b]y its very language ... to mixed questions of law and fact,” id. at 924. According to Neelley, there are two instances in which a state court decision would be “contrary to” clearly established Supreme Court precedent under that phrase’s plain meaning: “when a state court faces a set of facts that is essentially the same as those the Supreme Court has faced earlier” but reaches a different result; and when a state court fails to apply the correct legal principles, as enunciated by Supreme Court case law, to decide a ease. Id. at 923.
The Fourth Circuit, in Green v. French, 143 F.3d 865 (4th Cir.1998), enunciated an approach similar to that of the Fifth, Seventh, and Eleventh Circuits:
[A] decision is “contrary to” precedent only when, either through a decision of pure law or the application of law to facts indistinguishable in any material way from those on the basis of which the precedent was decided, that decision reaches a legal conclusion or a result opposite to and irreconcilable with that reached in the precedent that addresses the identical issue. In contrast, a decision represents an “unreasonable application of’ precedent, only when that decision applies a precedent in a context different from the one in which the precedent was decided and one to which extension of the legal principle of the precedent is not reasonable, when that decision fails to apply the principle of a precedent in a context where such failure is unreasonable, or when that decision recognizes the correct principle from the higher court’s precedent, but unreasonably applies that principle to the facts before it (assuming the facts are insufficiently different from those that gave rise to the precedent as to constitute a new context for consideration of the principle’s applicability).
Id. at 870. Green then announced the following standard:
*359Defining the terms in this manner, respectively, captures, we believe, the obvious common sense of the statute: If a state court decision is in square conflict with a precedent (supreme court) which is controlling as to law and fact, then the writ of habeas corpus should issue; if no such controlling decision exists, the writ should issue only if the state court’s resolution of a question of pure law rests upon an objectively unreasonable derivation of legal principles from the relevant supreme court precedents, or if its decision rests upon an objectively unreasonable application of established principles to new facts.
Id. The Fourth Circuit affirmed that its approach is, in form, very similar to Drinkard, Lindh, and Neelley. See id. at 870-73. Unlike Neelley, however, Green declined to view the term “clearly established Federal law” as a codification of the Teague doctrine. But Green did not provide insight into how that term alternatively should be understood and applied. See id. at 873-74.
The First Circuit, in O’Brien v. Dubois, 145 F.3d 16 (1st Cir.1998), criticized the “bifurcated standard” enunciated by the Fifth and Eleventh Circuits because it engrafts into the statute words — “questions of law,” and “mixed questions of law and fact”— which Congress did not place there. See id. at 22-23. Accordingly, the First Circuit came up with its own approach:
A federal habeas court charged to weigh a state court decision must undertake an independent two-step analysis of that decision. First, the habeas court asks whether the Supreme Court has prescribed a rule that governs the petitioner’s claim. If so, the habeas court gauges whether the state court decision is “contrary to” the governing rule. In the absence of a governing rule, the “contrary to” language clause drops from the equation and the habeas court takes the second step. At this stage, the habeas court determines whether the state court’s use of (or failure to use) existing law in deciding the petitioner’s claim involved an “unreasonable application” of Supreme Court precedent.
Id. at 24.
To help in the determination, “of how specific a rule must be to qualify as dispositive,” O’Brien looked to Teague and its progeny, and even referenced Neelley in citation. O’Brien states:
Drawing on Teague, we hold that an affirmative answer to the first section 2254(d)(1) inquiry — whether the Supreme Court has prescribed a rule that governs the petitioner’s claim — requires something more than a recognition that the Supreme Court has articulated a general standard that covers the claim. To obtain relief at this stage, a habeas petitioner must show that Supreme Court precedent requires an outcome contrary to that reached by the relevant state court. Cf. Neelley, 138 F.3d at 923-24.
O’Brien, 145 F.3d at 24-25. O’Brien went on to qualify this requirement, noting that “[a] petitioner need not point a habeas court to a factually identical precedent,” id. at 25, because “[ojftentimes, Supreme Court holdings are ‘general’ in the sense that they erect a framework specifically intended for application to variant factual situations,” id. (citing, in footnote, Waller v. Georgia, 467 U.S. 39, 104 S.Ct. 2210, 81 L.Ed.2d 31 (1984) (“violation of the right to public trial” test); Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984) (“ineffective assistance of counsel” test); and Jackson v. Virginia, 443 U.S. 307, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979) (“insufficiency of the evidence” test)). Thus, according to O’Brien, the proper question for determining when a state court decision is “contrary to” clearly established Supreme Court precedent is
whether a Supreme Court rule — by virtue of its factual similarity (though not necessarily identicality) or its distillation of general federal law precepts into a channeled mode of analysis specifically intended for application to variant factual situations— can fairly be said to require a particular result in a particular case.
Id. at 25.
Finally, O’Brien explained, if no Supreme Court “rule” can “fairly be said to require a particular result in a particular case,” then the “unreasonable application” test applies. Under this second step, inferior federal court decisions which have decided factually similar *360cases may help “in assessing the reasonableness vel non of the state court’s treatment of the contested issue,” id.7
This Circuit was confronted with interpreting § 2254(d) in Harpster v. Ohio, 128 F.3d 322 (6th Cir.1997), cert. denied, — U.S. -, 118 S.Ct. 1044, 140 L.Ed.2d 109 (1998). Harpster involved a “heavily fact intensive” question (whether the trial judge should have declared a mistrial), which was both “a mixed question of law and fact” and an issue on which “the Supreme Court has specifically avoided prescribing rigid rules that compel outcomes.” Harpster, 128 F.3d at 327. After comparing the similar approaches of the Fifth and Seventh Circuits to that of the First Circuit,8 we declined to adopt either, finding that the differences between the two approaches would not affect the result in that case because under either approach the panel was called upon to apply the “unreasonable application” clause of § 2254(d)(1). See id. at 326-27. We have not since articulated where we stand on the issue.
The issues in the case before us today involve whether pretrial publicity necessitated a change of venue in order for the defendants to receive a fair trial, and whether the errors surrounding the extraneous influences on the jury were harmless. The former issue, like the issue in Harpster, is a heavily fact-intensive mixed question of law and fact for which there is no clear “rule” enunciated by the Supreme Court requiring a certain result. Therefore, on the question of pretrial publicity, “under either [the Fifth and Seventh Circuits’ or the First Circuit’s] approach we must decide whether the state court [decision] ... ‘involved an unreasonable application of[ ] clearly established Federal law, as determined by the Supreme Court.’ ” Id. at 327 (quoting 28 U.S.C. § 2254(d)(1)). The extraneous jury influence issue also may be viewed as a “mixed question” case which, under the Fifth and Seventh Circuits’ approaches, necessitates review under the “unreasonable application” clause.9 Under the First Circuit’s test, the extraneous jury influences issue boils down to a question of “harmless error,” an issue on which the Supreme Court has adopted a specific rule, namely the Chapman rule, see generally Chapman v. California, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967), which governs all state court harmless error review of “trial” errors involving federal constitutional rights. See Brecht v. Abrahamson, 507 U.S. 619, 630, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993). However, because “harmless error” cases inevitably are extremely fact-intensive and case-specific, we cannot say that the Supreme Court’s Chapman rule is one which “can fairly be said to require a particular result in a particular case,” O’Brien, 145 F.3d at 25, and thus this issue under O’Brien would also be governed by the “unreasonable application” clause of § 2254(d)(1).
Likewise, neither Neelley nor Green counsels consideration of the issues in this case under anything other than the “unreasonable application” clause in § 2254(d)(1). This case involves neither an instance in which “a state court [has] face[d] a set of facts that is essentially the same as those the Supreme Court has faced earlier,” but reaches a differ*361ent result, or in which a state court has failed to apply the correct legal principles, as enunciated by the Supreme Court, to decide a case, the circumstances of which necessitate application of the “contrary to” clause under Neelley. Neelley, 138 F.3d at 923. Similarly, this case does not involve “a decision of pure law or the application of law to facts indistinguishable in any material way from those on the basis of which the precedent was decided----” See Green, 143 F.3d at 870. Accordingly, as in Harpster, we need not decide today which specific approach we find most persuasive, as they all point to the “unreasonable application” clause in § 2254(d)(1).
The lack of agreement among the circuits is even sharper with regard to the meaning and application of the term “unreasonable application.” While all the cases appear to agree that “the ‘unreasonable application’ clause does not empower a habeas court to grant the writ merely because it disagrees with the state court’s decision, or because, left to its own devices, it would have reached a different result,” O’Brien, 145 F.3d at 25,10 they do not agree on what the district court must find in order to issue the writ due to an “unreasonable application of[] clearly established Federal law,” § 2254(d)(1). The disagreement is not about whether the AEDPA requires a high degree of deference to the state court’s judgment; rather, it is about how to gauge the degree of deference necessary. The Fifth Circuit holds
that an application of law to facts is unreasonable only when it can be said that reasonable jurists considering the question would be of one view that the state court decision is incorrect. In other words, [the court will] ... grant habeas relief only if a state court decision is so clearly incorrect that it would not be debatable among reasonable jurists.
Drinkard, 97 F.3d at 769. The Fourth and Eleventh Circuits have expressly adopted this test as their own. See Green, 143 F.3d at 870, 873; Neelley, 138 F.3d at 924. However, the First Circuit has specifically declined to adopt the Drinkard test and instead has stated that “for the writ to issue, the state court decision must be so offensive to existing precedent, so devoid of record support, or so arbitrary, as to indicate that it is outside the universe of plausible, credible outcomes.” O’Brien, 145 F.3d at 25 & n. 7. The Seventh Circuit has articulated divergent standards. See Lindh v. Murphy, 96 F.3d at 871 (stating that a state court’s “ ‘reasonable’ decision ... must be honored,” but failing to articulate what that might be other than to say that a federal court must “take into account the care with which the state court considered the subject”); Hennon, 109 F.3d at 335 (specifically rejecting the “care-with-which-the-state-court-considered-the-subject” language in Lindh and stating instead that the question of “unreasonable application” turns on “whether the determination is at least minimally consistent with 'the facts and circumstances of the case”); Hall v. Washington, 106 F.3d 742, 748-49 (7th Cir.) (“The statutory ‘unreasonableness’ standard allows the state court’s conclusion to stand if it is one of several equally plausible outcomes.”), cert. denied, - U.S. -, 118 S.Ct. 264, 139 L.Ed.2d 190 (1997).
Since we heard oral argument in this case this Court, in Herbert v. Billy, 160 F.3d 1131, 1998 WL 801319 (6th Cir.1998), has stated its agreement with the “reasonable jurist” approach taken by the Fifth Circuit in Drinkard, 97 F.3d at 769. Today we state as well our agreement with the standard enunci*362ated by the First Circuit, namely, that “for the writ to issue, the state court decision must be so offensive to existing precedent, so devoid of record support, or so arbitrary, as to indicate that it is outside the universe of plausible, credible outcomes.” O’Brien, 145 F.3d at 25; accord Hall, 106 F.3d at 748-49 (“The statutory ‘unreasonableness’ standard allows the state court’s conclusion to stand if it is one of several equally plausible outcomes.”). We recognize that the First Circuit specifically rejected Drinkard’s “reasonable jurist” test when articulating its own standard, O’Brien, 145 F.3d at 25 n. 7. We think, however, that the two are not mutually exclusive; rather, both standards can be employed to aid in arriving at the correct answer to the question of “unreasonableness.” The deference to the state courts’ judgments required by the AEDPA is achieved by adopting the rule that the unreasonableness of a state court’s application of clearly established Supreme Court precedent will not be “debatable among reasonable jurists,” Drinkard, 97 F.3d at 769, if it is “so offensive to existing precedent, so devoid of record support, or so arbitrary, as to indicate that it is outside the universe of plausible, credible outcomes,” O’Brien, 145 F.3d at 25.
II. PRE-TRIAL PUBLICITY
Petitioner argued that he should have received a change of venue due to pretrial publicity. The district court agreed, first finding that “[t]he Supreme Court has established two standards to guide courts, the ‘actual prejudice’ standard and the ‘presumed prejudice’ standard,” Nevers, 990 F.Supp. at 854, and second determining that the Michigan Court of Appeals11 had unreasonably applied the “presumed prejudice” standard to Petitioner’s case. It did not address the “actual prejudice” standard (the standard applied by the Michigan Court of Appeals) because it found the “presumed prejudice” standard dispositive. Id.
1. Clearly Established Supreme Court Precedent?
The first question that must be addressed is whether the distinction recognized by the district court as existing between presumptive and actual prejudice is “clearly established Federal law, as determined by the Supreme Court of the United States.” 28 U.S.C. § 2254(d)(1). Review of relevant Supreme Court precedent indicates that there is such a distinction; it is best articulated by Murphy v. Florida, 421 U.S. 794, 95 S.Ct. 2031, 44 L.Ed.2d 589 (1975). Murphy categorized the prejudice dealt with in the Court’s previous cases, in which the Court “overturned a state-court conviction obtained in a trial atmosphere that had been utterly corrupted by press coverage,” id. at 798, 95 S.Ct. 2031, into two types — “actual” prejudice and “presumed” prejudice, both of which require reversal for refusal to change venue. Tvsto years later these two distinct kinds of prejudice were specifically discussed in Dobbert v. Florida, 432 U.S. 282, 97 S.Ct. 2290, 53 L.Ed.2d 344 (1977), in which the Court wrote, “We concluded that the petitioner in Murphy had failed to show that the trial setting was inherently prejudicial or that the jury selection process permitted an inference ofachial prejudice. 421 U.S. at 802, 95 S.Ct. at 2037.” Dobbert, 432 U.S. at 303, 97 S.Ct. 2290 (emphasis added).
“Trial Setting Inherently Prejudicial.”
One type of prejudice is represented by Rideau v. Louisiana, 373 U.S. 723, 83 S.Ct. 1417, 10 L.Ed.2d 663 (1963), Estes v. Texas, 381 U.S. 532, 85 S.Ct. 1628, 14 L.Ed.2d 543 (1965), and Sheppard v. Maxwell, 384 U.S. 333, 86 S.Ct. 1507, 16 L.Ed.2d 600 (1966), in which the Court concluded that the “circumstances under which the trials ... were held” were such that inherent prejudice to the venire should be presumed. Murphy, 421 U.S. at 798-99, 95 S.Ct. 2031. Rideau involved the televising of an in-jail 20 minute interrogation of the defendant by the police in which the defendant confessed to the murder for which he was subsequently convicted. The Court concluded that Rideau presumptively could not have received a fair trial in that parish “because it considered the trial under review ‘but a hollow formality’' — -the real trial had occurred when tens of thou*363sands [12] of people, in a community of 150,-000, had seen and heard the defendant admit his guilt before the cameras.” Id. at 799, 95 S.Ct. 2031. “The trial in Estes had been conducted in a circus atmosphere, due in large part to the intrusions of the press, which was allowed to sit within the bar of the court and to overrun it with television equipment.” Id. Sheppard contained “not only a background of extremely inflammatory publicity but also a courthouse given over to accommodate the public appetite for carnival.” Id. Regarding these latter two cases, the Court stated:
The proceedings in these cases were entirely lacking in the solemnity and sobriety to which a defendant is entitled in a system that subscribes to any notion of fairness and rejects the verdict of a mob. They cannot be made to stand for the proposition that juror exposure to information about a state defendant’s prior convictions or to news accounts of the crime with which he is charged alone presumptively deprives the defendant of due process.
Id.
“Jury Selection Process Permitted Inference of Actual Prejudice.” The other type of prejudice, represented by Irvin v. Dowd, 366 U.S. 717, 81 S.Ct. 1639, 6 L.Ed.2d 751 (1961), stands for the proposition that absent the “televised confession amounting to a trial” or “carnival atmosphere” situations, pretrial publicity that would inherently prejudice the jury pool can be discerned only by reviewing both the extent and nature of the publicity and the responses of the prospective jurors in voir dire. Irvin involved a capital case in which the defendant was tried in a small community that was widely aware of his prior convictions, his confession to 24 burglaries and six murders (including the one for which he stood trial) and his unaccepted offer to plead guilty in order to avoid a death sentence. See Irvin, 366 U.S. at 725-28, 81 S.Ct. 1639. In addition to this prejudicial information before the trial began eight of the 12 empaneled jurors had already formed the opinion that the defendant was guilty; some went “so far as to say that it would take evidence to overcome their belief’ in the defendant’s guilt. Id. at 728, 81 S.Ct. 1639. Also, some 90% of the prospective jurors examined on the point indicated some degree of belief in the accused’s guilt; for this reason 268 of the 430 veniremen were excused for cause. Id. at 727, 81 S.Ct. 1639. The Murphy Court, after reviewing the facts of Irvin, noted, “In these circumstances, the [Irvin ] Court readily found actual prejudice against the petitioner to a degree that rendered a fair trial impossible.” Murphy, 421 U.S. at 798, 95 S.Ct. 2031. The Court went on to observe:
In a community where most veniremen will admit to a disqualifying prejudice, the reliability of the others’ protestations may be drawn into question; for it is then more probable that they are part of a community deeply hostile to the accused, and more likely that they may unwittingly have been influenced by it.
Id. at 803, 95 S.Ct. 2031. Comparing the facts in Murphy (petitioner claimed that the jury had learned of a prior conviction as well as certain facts about the crime at issue) to those in Irvin, the Murphy Court found no comparable community-based animus against Murphy, and thus affirmed the Court of Appeals’ conclusion that the petitioner had not been denied his right to a fair trial.
Since Murphy, the Supreme Court has not applied the “trial setting inherently prejudicial” standard of Rideau, Estes, and Sheppard, presumably because the “televised interrogation/confession in a smaller community” and the “tabloid-esque, carnival atmosphere” instances have not arisen. We think that the holdings in those cases are, therefore, to be limited to their facts. The benchmarks for measuring pretrial publicity appear to be Irvin on the one end and Murphy on the other, as exemplified in Dobbert v. Florida, 432 U.S. 282, 97 S.Ct. 2290, 53 L.Ed.2d 344 (1977), decided two years after Murphy. In Dobbert the petitioner, who was accused of murdering his children, pointed to extensive pretrial publicity surrounding his case. After discussing Murphy the Court concluded that Dobbert’s case was *364more like Murphy than Irvin (the others were not even mentioned):
Petitioner’s argument that the extensive coverage by the media denied him a fair trial rests almost entirely upon the quantum of publicity which the events received. He has directed us to no specific portions of the record, in particular the voir dire examination of the jurors, which would require a finding of constitutional unfairness as to the method of jury selection or as to the character of the jurors actually selected. But under Murphy, extensive knowledge in the community of either the crimes or the putative criminal is not sufficient by itself to render a trial constitutionally unfair. Petitioner in this case has simply shown that the community was made well aware of the charges against him and asks us on that basis to presume unfairness of constitutional magnitude at his trial. This we will not do in the absence of a “trial atmosphere ... utterly corrupted by press coverage,” Murphy v. Florida, supra, 421 U.S. at 798, 95 S.Ct. at 2035. One who is reasonably suspected of murdering his children cannot expect to remain anonymous. Petitioner has failed to convince us that under the “totality of the circumstances,” Mmphy, supra, the Florida Supreme Court was wrong in finding no constitutional violation with respect to the pretrial publicity.
Dobbert, 432 U.S. at 303, 97 S.Ct. 2290.
Seven years after Dobbert the Court again addressed a “trial atmosphere utterly corrupted by press coverage” in Patton v. Yount, 467 U.S. 1025, 104 S.Ct. 2885, 81 L.Ed.2d 847 (1984), which involved a high school teacher accused of raping and murdering one of his female students. The Court discussed only Irvin, see id. at 1032-34, 104 S.Ct. 2885, and Murphy, see id. at 1033-35, 104 S.Ct. 2885,13 in its review of the extent and nature of the pretrial publicity and the testimony of the jurors at voir dire. Concluding that “the jurors at Yount’s trial had [not had] such fixed opinions that they could not judge impartially the guilt of the defendant,” id. at 1035, 104 S.Ct. 2885, the Court reversed the Fourth Circuit’s grant of the writ.
Because we conclude that there is clearly established Supreme Court precedent distinguishing between cases involving presumed prejudice — when the “setting of the trial [is] inherently prejudicial,” Murphy, 421 U.S. at 803, 95 S.Ct. 2031 — and actual prejudice— when review of both the jury voir dire testimony and the extent and nature of the media coverage indicates “a fair trial [was] impossible,” id. at 798, 95 S.Ct. 2031 — we turn to the question of whether the district court erred in determining that the Michigan Court of Appeals unreasonably applied “clearly established” precedent.
2. Unreasonable Application?
The Michigan Court of Appeals opinion states:
Although this case involved extensive media publicity and expressions of community sentiment, all of the seated jurors indicated under oath that they could hear defendants’ cases fairly and impartially, despite their exposure to media publicity and community sentiment. Thus, there is a presumption of impartiality which may be rebutted only by demonstrating facts sufficient to indicate actual opinion or bias on the part of seated jurors or by showing that the nature and extent of the media publicity and community sentiment was such that prejudice must be presumed. Murphy v. Florida, 421 U.S. 794, 95 S.Ct. 2031, 44 L.Ed.2d 589 (1975); People v. DeLisle, 202 Mich.App. 658, 509 N.W.2d 885 (1993).
As in Mmphy and DeLisle, less than one-third of the prospective jurors for these cases were disqualified on account of bias, and all the jurors chosen denied having fixed opinions as to defendants’ guilt. Moreover, in Mmphy and DeLisle, many or .all of the seated jurors had prior knowledge of matters arguably more damaging to the defense than the content of the publicity surrounding these cases, such as facts regarding the accused’s prior criminal record or pretrial confession to the *365charged offense. Admittedly, the instant cases are unique to the extent that jurors were exposed to predictions or concerns that local rioting may occur in the event of acquittal, as had previously occurred in Los Angeles when police officers were acquitted of state charges in the Rodney King beating ease. However, we do not find this sufficient to establish presumptive prejudice. As for defendants’ claims of actual juror bias or prejudice based upon certain selected responses given on voir dire, we find that the jurors’ remarks, when viewed in full context, do not indicate partiality, fixed opinion, or inability to decide the case solely on the evidence presented.
People v. Budzyn, Nos. 170477, 170478, slip op. at 6-7 (Mich.Ct.App. March 22, 1995) (unpublished per curiam).
The district court found that the Michigan court’s decision was an unreasonable application of clearly established Supreme Court precedent because it “failed to address the presumed prejudice standard of Rideau and Murphy, [and] ... instead relied upon juror statements that they will be impartial,” Nevers, 990 F.Supp. at 864. In particular the district court relied on Rideau, see Nevers, 990 F.Supp. at 854. Rideau, however, is not applicable to this case because Rideau is limited to its facts: a televised jail-house interrogation of a defendant, in which he confessed to a robbery/kidnapping/murder that he had committed only the previous night, seen by over one-third of the community’s residents two weeks before his arraignment, Rideau, 373 U.S. at 724, 83 S.Ct. 1417, and less than two months before his trial, id. at 729, 83 S.Ct. 1417 (Clark, J., dissenting). As a result, the trial was but a “hollow formality” because Rideau’s “real trial” had occurred when those in the community watched the televised interrogation and heard him admit his guilt. Murphy, 421 U.S. at 799, 95 S.Ct. 2031. Rideau must be understood as limited to its facts because the Supreme Court has not discussed it in a pretrial publicity case since Murphy, and even in Murphy it was discussed strictly on its facts.14 In the case before us, the only thing close to a pretrial admission by Nevers appears in an article published two days after the incident (and seven months before his trial) in which Nevers is quoted as saying, “I must have done something wrong, a guy died.” Nevers, 990 F.Supp. at 855. This is hardly the equivalent of a televised confession to murder.
Neither do the facts of this case evince the carnival atmosphere at issue in Sheppard and Estes. To be sure, there was extensive media coverage in the case before us,15 most of which was newspaper rather than television, Nevers, 990 F.Supp. at 855, and which we discuss more extensively infra. The content and extent of the newspaper articles in this ease are similar to, but much milder than, those found in Sheppard. Sheppard, however, involved out-of-control television coverage both before16 and during the trial that makes the media hype surrounding the O.J. Simpson trial appear tame, and, as the Court noted in Murphy, it is this television circus, in addition to the newspaper coverage, for which Sheppard is significant.
Sheppard arose from a trial infected not only by a background of extremely inflammatory publicity but also by a courthouse given over to accommodate the public ap*366petite for carnival. The proceedings in ... \Sheppard and Estes ] were entirely lacking in the solemnity and sobriety to which a defendant is entitled in a system that subscribes to any notion of fairness and rejects the verdict of a mob.
Murphy, 421 U.S. at 799, 95 S.Ct. 2031.17
Because this case did not involve a televised confession as in Rideau or the carnival atmosphere displayed in Estes and Sheppard, the clearly established law we must apply here is that of Irvin, where the question was whether “adverse pretrial publicity ... created] such a presumption of prejudice in a community that the jurors’ claims that they can be impartial should not be believed.” Patton, 467 U.S. at 1031, 104 S.Ct. 2885 (discussing Irvin). The analysis under Irvin differs from that under Rideau, Estes, and Sheppard, in that we must review not only the media coverage itself but the substance of the jurors’ statements at voir dire in determining whether there was a “community-wide sentiment” against the defendant.
To be sure, there were many in the community enraged by the beating death of Malice Green, as shown by the several newspaper articles and editorials contained in the record and discussed thoroughly in the district court’s opinion. See Nevers, 990 F.Supp. at 855-63.18 The district court’s opinion correctly summarizes the most problematic examples of the pretrial publicity, see Nevers, 990 F.Supp. at 862-64: prejudicial statements to the media by the mayor and police chief, especially the statement by the mayor to NBC Nightly News, “A young man who was under arrest was literally murdered by police;” the rapid settlement by the city of a civil suit lodged by Green’s family for over $5 million; the reports of riot preparation by the Detroit Police Department in the event of an acquittal; and the widely publicized facts concerning Nevers’s reputation as a violent rogue cop, his involvement in the shooting deaths of previous suspects, the brutality complaints filed against him, and his former involvement in STRESS,19 id. at 863.
However, while the information contained in the media coverage was certainly prejudicial, it was not nearly so prejudicial as that found in Irvin, in which the
barrage of newspaper headlines, articles, cartoons and pictures ... unleashed against ... [the defendant] during the six or seven months preceding his trial ... announced his confession to the six murders and the fact of his indictment for four of them in Indiana[;] ... [his] offer to plead guilty if promised a 99-years sentence, but also the determination, on the part of the prosecutor to secure the death penalty[;] and ... [his] confession] to 24 burglaries (the modus operandi of these robberies was compared to that of the murders and the similarity noted). One story ... characterized petitioner as re*367morseless and without conscience but also as having been found sane by a court-appointed panel of doctors. In many of the stories petitioner was described as the “confessed slayer of six,” a parole violator and fraudulent-check artist____ On the day before the trial the newspapers carried the story that Irvin had orally admitted the murder of Kerr (the victim in this case) as well as the “robbery-murder of Mrs. Mary Holland; the murder of Mrs. Wilhelmina Sailer in Posey County, and the slaughter of three members of the Duncan family in Henderson County, Ky.”
Irvin, 366 U.S. at 725-26, 81 S.Ct. 1639.
More importantly, Irvin discussed information regarding the jury panel which the Murphy Court found extremely important. As noted previously, in Irvin “90% of those examined on the point ... entertained some opinion as to guilt — ranging in intensity from mere suspicion to absolute certainty.” Id. at 727, 81 S.Ct. 1639. “[Ejight of the 12 [empaneled] jurors had formed an opinion that the defendant was guilty before the trial began; some went ‘so far as to say that it would take evidence to overcome their belief in [the petitioner’s] guilt.” Murphy, 421 U.S. at 798, 95 S.Ct. 2031. “The panel consisted of 430 persons[, and] [t]he court itself excused 268 of those on challenges for cause as having fixed opinions as to the guilt of the petitioner....” Irvin, 366 U.S. at 727, 81 S.Ct. 1639. The Murphy Court, after noting these facts, stated:
In the present case, by contrast, 20 of the 78 persons questioned were excused because they indicated an opinion as to petitioner’s guilt. This may indeed be 20 more than would occur in the trial of a totally obscure person, but it by no means suggests a community with sentiment so poisoned against petitioner as to impeach the indifference of jurors who displayed no animus of their own.
Murphy, 421 U.S. at 803, 95 S.Ct. 2031 (footnote omitted).
In the ease before us, the Appellant’s Brief states, without citation to the record, that of the 105 prospective jurors dismissed, only 15 were dismissed for expressing bias, (Appellant’s Br. at 25 n. 5), a fact not contested in Appellee’s Brief or at oral argument. Also, it appears certain that all potential jurors had heard about the case, as the prosecutor conceded as much during voir dire, see J.A. at 502 (Voir Dire Tr.); but this, of course, is not enough to justify a change of venue, as even the “existence of a[] preconceived notion as to the guilt or innocence of an accused, without more, is [not] sufficient to rebut the presumption of a prospective juror’s impartiality,” Irvin, 366 U.S. at 723, 81 S.Ct. 1639.
The Joint Appendix submitted on appeal contains almost 500 pages of jury voir dire, which, in the main, is the voir dire testimony of the empaneled jurors, with scatterings of excerpts from the voir dire testimony of other potential jurors who did not become members of the panel. We attach as an appendix a summary of our review of these pages of the transcript.
Our extensive review of the transcript of the jury voir dire leads us to conclude that none of the jurors actually empaneled had any fixed opinion as to Nevers’s guilt before the trial began, in contrast to the situation in Irvin, where eight of the twelve jurors had such opinions. These transcript pages disclose only two jurors eventually seated who were challenged for cause; only as to one of those, namely Juror # 5, does the trial court’s denial of the challenge for cause appear to have been questionable. Cf. Beck v. Washington, 369 U.S. 541, 557-58, 82 S.Ct. 955, 8 L.Ed.2d 98 (1962) (“The fact that petitioner did not challenge for cause any of the jurors so selected is strong evidence that he was convinced the jurors were not biased and had not formed any opinions as to his guilt.”). As for the jurors’ statements about “community sentiment” and the opinions of their friends and co-workers, the voir dire transcript indicates that there were many in Detroit who had the opinion that Nevers likely was not guilty, although those believing Nevers guilty were the most vocal and were probably in the majority. It appears, however, that those jurors who were actually empaneled took seriously the obligation they were about to undertake and did not have preformulated opinions as to Nevers’s guilt. Cf. Gentile v. State Bar of Nevada, 501 U.S. *3681030, 1054-55, 111 S.Ct. 2720, 115 L.Ed.2d 888 (1991) (involving an attorney’s ability to speak about a trial to the media and stating, “Only the occasional case presents a danger of prejudice from pretrial publicity. Empirical research suggests that in the few instances when jurors have been exposed to extensive and prejudicial publicity, they are able to disregard it and base their verdict upon the evidence presented in court.”) (citing law review articles). It also appears, however, that the news media was, in a layman’s eyes, painting Nevers as guilty, and that at least one juror was concerned about having his name publicly disclosed for fear of harassment by the extremists on the side adversely affected by the jury’s verdict.
While this certainly is not a clear-cut case, we are unable to say, after carefully reviewing the voluminous Joint Appendix submitted by the parties, that these facts place this case in the same league as Irvin in terms of community bias and actual prejudice. And this is a federal habeas action, in which we must review the decisions of the state courts under the deferential standard imposed by the AEDPA. Even if we would reach a different conclusion if this case were before us on diréct appeal from a criminal trial in a federal district court, we are unable to say that the Michigan Court of Appeals’s application of Supreme Court precedent resulted in a decision whose unreasonableness “could not be debatable among reasonable jurists,” Drinkard, 97 F.3d at 769, because it is “so offensive to existing precedent, so devoid of record support, or so arbitrary, as to indicate that it is outside the universe of plausible, credible outcomes,” O’Brien, 145 F.3d at 25. Accordingly, because we cannot say that the Michigan Court of Appeals’s decision “resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States,” 28 U.S.C. § 2254(d)(1), we hold that the district court erred in granting the writ on this issue.
III. EXTRANEOUS INFLUENCES ON THE JURY
Nevers and Budzyn claimed in their appeals before the Michigan appellate courts that extraneous influences on the jury deprived them of fair trials. The Michigan Supreme Court, citing Hughes v. Borg, 898 F.2d 695, 700 (9th Cir.1990), and other Ninth Circuit precedent, articulated the following test for considering their allegations:
In order to establish that the extrinsic influence was error requiring reversal, the defendant must initially prove two points. First, the defendant must prove that the jury was exposed to extraneous influences. Second, the defendant must establish that these extraneous influences created a real and substantial possibility that they could have affected the jury’s verdict. Generally, in proving this second point, the defendant will demonstrate that the extraneous influence is substantially related to a material aspect of the case and that there is a direct connection between the extrinsic material and the adverse verdict. If the defendant establishes this initial burden, the burden shifts to the people to demonstrate that the error was harmless beyond a reasonable doubt. We examine the error to determine if it is harmless beyond a reasonable doubt because the error is constitutional in nature. The people may do so by proving that either the extraneous influence was duplicative of evidence produced at trial or the evidence of guilt was overwhelming.
Budzyn, 566 N.W.2d at 235 (citations and footnotes omitted). The court then noted that Budzyn and Nevers had submitted four affidavits from jurors, three of whom had sat on Nevers’s jury and one of whom had sat on Budzyn’s, as proof. After determining that three of the affidavits were admissible evidence available for their consideration,20 the *369court concluded that during the trial, three extraneous matters had reached some of the jurors or the jury as a whole, namely: (1) the jury had viewed (at least in part) the film “Malcolm X”; (2) a member of Nevers’s jury had learned from news reports about the city’s preparing for a potential riot in the event of an acquittal; and (3) the jurors had been exposed to and allegedly had considered the supposed fact that Nevers and Budzyn had been members of STRESS. In a footnote, the court noted that the trial court had denied a request from Budzyn and Nevers for an evidentiary hearing regarding these affidavits, and instead had accepted as true that the extraneous information had come into the jury’s possession and had been considered. Because the court explicitly found a real and substantial possibility that these matters could have influenced the juries’ verdicts, see 566 N.W.2d at 236-40, it conducted a harmless error review. The court concluded that the extraneous matters were harmless as to Nevers because of the strength of the evidence against him, namely the consistency and credibility of the EMS workers’ testimonies, id. at 240-41, but that the extraneous matters were not harmless as to Bud-zyn because the testimony of the civilian witnesses was not as credible, did not conclusively show him striking Green on the head, and was conflicting. See id. at 241-43.
On habeas review, the district court agreed with the Michigan Supreme Court that the jury’s receipt of the extraneous information amounted to constitutional error. Nevers, 990 F.Supp. at 867. The district court then determined that the error was “trial error,” rather than “structural error.” In reaching this conclusion, the district court noted that the Supreme Court has found “structural errors” in only limited circumstances, see Nevers, 990 F.Supp. at 867 (citing Gideon v. Wainwright, 372 U.S. 335, 83 S.Ct. 792, 9 L.Ed.2d 799 (1963) (total deprivation of the right to counsel); Tumey v. Ohio, 273 U.S. 510, 47 S.Ct. 437, 71 L.Ed. 749 (1927) (lack of impartial trial judge); Vasquez v. Hillery, 474 U.S. 254, 106 S.Ct. 617, 88 L.Ed.2d 598 (1986) (unlawful exclusion of grand jurors of defendant’s race); McKaskle v. Wiggins, 465 U.S. 168, 104 S.Ct. 944, 79 L.Ed.2d 122 (1984) (the right to self-representation at trial); Waller v. Georgia, 467 U.S. 39, 104 S.Ct. 2210, 81 L.Ed.2d 31 (1984) (right to public trial); Sullivan v. Louisiana, 508 U.S. 275, 113 S.Ct. 2078, 124 L.Ed.2d 182 (1993) (erroneous reasonable doubt instruction)); the extrinsic information reached the jury only after the trial had begun rather than permeating the entire process and thus was more closely akin to errors which occurred during the presentation of the case to the jury, see id. at 868; and prior Supreme Court extraneous jury influence cases have treated the issue as subject to harmless error review, see id. (citing Smith v. Phillips, 455 U.S. 209, 217, 102 S.Ct. 940, 71 L.Ed.2d 78 (1982); Remmer v. United States, 347 U.S. 227, 229, 74 S.Ct. 450, 98 L.Ed. 654 (1954)). Finally, the district court concluded that because the error was of the “trial” and not the “structural” variety, it was subject to “harmless error” analysis. See generally Brecht v. Abrahamson, 507 U.S. 619, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993).
Whether Nevers has shown that he is entitled under § 2254(d) to issuance of the writ on his claim of jury taint appears to us to have two separate aspects. The first and obvious aspect is his claim that the Michigan Supreme Court’s application of harmless error precedent was unreasonable. The second aspect, less obvious but clearly implied from the first, is the claim that the state courts’ failure to conduct any inquiry into the actual effect of the extraneous information the jury acquired resulted in a decision that was contrary to clearly established law. It may be that Nevers has waived his right to argue this second claim, since the state courts’ opinions do not squarely address this issue and it was not explicitly raised in the habeas petition or before us on appeal. Our review of this case, however, persuades us that where the state court has expressly ruled in a manner that prevents a defendant in a criminal proceeding from having the opportunity to obtain the information he must have to pursue a claim of constitutional error, whether that ruling was contrary to law is inextricable from the question of whether the constitutional error was harm*370less. Accordingly, we will address both aspects of Nevers’s claim of jury taint.
We agree with the district court — for the reasons mentioned by that court — that Nevers’s claim that the jury was influenced by extraneous information was amenable to harmless error analysis. Additionally, we note that at least four other circuit courts have found extrinsic juror influence claims to be subject to “harmless error” analysis. E.g., Pyles v. Johnson, 136 F.3d 986, 993 (5th Cir.), cert. denied, — U.S. -, 118 S.Ct. 2338, 141 L.Ed.2d 707 (1998); Jeffries v. Wood, 114 F.3d 1484, 1489 (9th Cir.), cert. denied, — U.S. -, 118 S.Ct. 586, 139 L.Ed.2d 423 (1997); Sherman v. Smith, 89 F.3d 1134, 1137 (4th Cir.1996), cert. denied, 519 U.S. 1091, 117 S.Ct. 765, 136 L.Ed.2d 712 (1997); Bibbins v. Dalsheim, 21 F.3d 13, 16 (2d Cir.) (per curiam), cert. denied sub nom., Bibbins v. New York, 513 U.S. 901, 115 S.Ct. 261, 130 L.Ed.2d 181 (1994).
We must therefore determine what standard we must use in reviewing the Michigan Supreme Court’s application of harmless error review. Pre-AEDPA Supreme Court precedent originally held that constitutional errors occurring in the state trial court require reversal unless the government could prove the error “harmless beyond a reasonable doubt.” See generally Chapman v. California, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967). In 1993 the Supreme Court in Brecht v. Abrahamson specifically held that Chapman no longer applied to federal habeas review of state court harmless error decisions, although state courts would still be required to apply that standard on direct review. Instead, on habeas review, federal courts were to grant the writ where eonstitu-tional “trial” errors, as opposed to “structural” errors not amenable to harmless error review, were demonstrated to have “‘had substantial and injurious effect or influence in determining the jury’s verdict.’ ” Brecht, 507 U.S. at 638, 113 S.Ct. 1710 (quoting Kotteakos v. United States, 328 U.S. 750, 776, 66 S.Ct. 1239, 90 L.Ed. 1557 (1946)). The Brecht standard was further qualified in O’Neal v. McAninch, 513 U.S. 432, 115 S.Ct. 992, 130 L.Ed.2d 947 (1995), which held that “where the record is so evenly balanced [on the issue of whether the error had a substantial and injurious effect] that a conscientious judge is in grave doubt as to the harmlessness of an error,” id. at 437, 115 S.Ct. 992, “the petitioner must win,” id. at 436, 115 S.Ct. 992. Thus, until the AEDPA the “substantial and injurious effect” test governed habeas review of constitutional “trial-type” errors.
The AEDPA, however, amended § 2254(d) to decree that the writ shall not issue unless the state court’s affirmance of the conviction amounted to an “unreasonable application” of clearly established Supreme Court precedent. On direct review, the state appellate court would determine whether the claimed error was harmless under Chapman’s “unreasonable beyond a reasonable doubt” standard.21 The question federal habeas courts must now address when assessing the harmlessness of constitutional error is what the standard is for determining whether the state court’s decision resulted from an “unreasonable application” of Chapman.
The district court, recognizing the conundrum, reviewed the Michigan Supreme Court’s determination under both the Chap*371man and the Brecht /O’Neal standards and found both that the extraneous information had a substantial and injurious effect on the jury’s verdict, and that the Michigan Supreme Court had unreasonably applied Supreme Court precedent by concluding that the error had been harmless. One month after the district court issued its opinion in this case, a different judge of the Eastern District of Michigan concluded that the AEDPA had abrogated the Brecht /O’Neal standard and required the habeas court to assess whether the state court’s application of the Chapman standard was unreasonable. See Barker v. Yukins, 993 F.Supp. 592, 602 n. 7 (E.D.Mich.1998). We have found no opinions from our sister circuits addressing the appropriate standard to now apply when faced with this question.22 Accordingly, we write on a clean slate.
The clear language of § 2254(d) permits us to review the Michigan Supreme Court’s adjudication of harmless error only to determine whether it is contrary to or resulted from an unreasonable application of clearly established federal law — specifically, the harmless error standard of Chapman v. California,23 The threshhold question confronting us is how we are to make that determination. We think that the amendments to § 2254(d) do not vitiate in any way the premise set out in Brecht v. Abrahamson, that
[sjtate courts are fully qualified to identify constitutional error and evaluate its prejudicial effect on the trial process under Chapman, and state courts often occupy a superior vantage point from which to evaluate the effect of trial error. For these reasons, it scarcely seems logical to require federal habeas courts to engage in the identical approach to harmless-error review that Chapman requires state courts to engage in on direct review.
Brecht, 507 U.S. at 636,113 S.Ct. 1710 (internal citations omitted). As the Court pointed out, if federal habeas courts are required simply to apply a Chapman analysis, the result will be to overturn convictions because there is a “reasonable possibility” that trial error contributed to the verdict, a result at odds with the historic meaning of habeas corpus, which is .to afford relief to those whom society has “grievously wronged.” Id. at 637, 113 S.Ct. 1710.
We have already explained that under the highly deferential standard of AEDPA we may not find the state court’s application of federal law unreasonable unless reasonable jurists would find it so arbitrary, unsupported or offensive to existing precedent as to be outside the realm of plausible credible outcomes. We think that when the issue before the federal habeas court is the state court’s finding of harmless error, the test set out by the Supreme Court in Kotteakos and explicitly reiterated in Brecht quite precisely captures Congress’s intent as expressed in AEDPA and, therefore, continues to be applicable. That test is whether the error “ ‘had substantial and injurious effect or influence in determining the jury’s verdict,”’ Brecht, 507 U.S. at 637, 113 S.Ct. 1710; it is the habeas petitioner’s burden to demonstrate that the trial error resulted in “actual prejudice.” Id. If the petitioner is able to make that showing, he will surely have demonstrated that the state court’s finding that *372the error was harmless beyond a reasonable doubt — the Chapman standard — was outside the realm of plausible credible outcomes, and therefore resulted from an unreasonable application of Chapman.
The Michigan Supreme Court held that the jury’s possession of extraneous information was constitutional error amenable to harmless error analysis. Further, the court conceded that there was “a real and substantial possibility that these external influences together could have affected the juries’ verdicts.” The court then conducted a harmless eiTor analysis to determine whether the error was harmless beyond a reasonable doubt, which, the court said, the state could demonstrate by “proving that either the extraneous influence was duplicative of evidence produced at trial or the evidence of guilt was overwhelming.” Budzyn, 566 N.W.2d at 235 (citing Hughes v. Borg, 898 F.2d 695, 700 (9th Cir.1990)).24 The court concluded:
In the Nevers’ trial, the four EMS witnesses, who had no apparent motive to lie, provided interlocking testimony that Nev-ers repeatedly bludgeoned Malice Green in the head with his heavy police flashlight while Green was dazed and not offering significant resistance. The medical testimony of the injuries to Green’s head also substantiated this testimony. The people have proven that there was unimpeachable, compelling evidence that Nevers harbored, at the very least, an unjustified intent to commit great bodily harm against Green. Thus, under Hughes, the extraneous influences were harmless.
Id. at 240-41 (footnote omitted).
We agree with the district court that the Michigan Supreme Court applied the Chapman “harmless beyond a reasonable doubt” test. We must therefore determine whether the district court was correct in holding that the application of that test was unreasonable.
The issue at the heart of Nevers’s prosecution was not what Nevers did, but why he did it. As the Michigan Supreme Court said, the cumulative evidence that Nevers beat Malice Green was overwhelming; it is virtually undisputed that blows to the head were a cause of Green’s death; and Nevers admitted striking many of those blows. But Nevers’s defense centered around his contention that the blows he struck were intended solely to subdue Green, to prevent him from grabbing Nevers’s gun, to protect himself from Green, and to force Green to relinquish the object he clutched in his hand, which Nevers claimed to have thought was something sharp and perhaps useable as a weapon. Nevers, in short, claimed that he used force only in response to Green’s refusal to submit to the police and that the force he used was only that necessary to protect himself and subdue Green.
It was the province of the jury alone to determine Nevers’s credibility and the veracity of his testimony. The extraneous information that the jury was exposed to during the course of the trial, and particularly the information regarding STRESS, unquestionably had the potential for influencing how the jury viewed Nevers’s testimony about his motivation for beating Green, and the Michigan Supreme Court acknowledged as much. The Michigan court, however, went on to conclude that in light of the evidence against Nevers, including evidence the court labeled as “unimpeachable” regarding his motives, the extraneous influences were harmless beyond a reasonable doubt.
We cannot conclude that the Michigan court’s application of the Chapman “harmless beyond a reasonable doubt” test was reasonable. Chapman explicitly includes the Fahy “reasonable possibility that the ... [error] complained of might have contributed to the conviction” test. Chapman, 386 U.S. at 24, 87 S.Ct. 824 (internal quotation marks omitted). The Michigan Supreme Court characterized the extraneous information regarding Nevers’s supposed membership in STRESS as
the kind of concrete, factual evidence that could substantially compromise the ability of a jury to issue a fair verdict because the evidence relates directly to the past conduct of the police officers ... It suggests *373... that these officers may have been acting in accordance with their preexisting racist predisposition to target young black men for abuse when they encountered Malice Green. This evidence was never introduced at trial. It is the kind of evidence that has a direct and rational connection between it and an adverse verdict.
Budzyn, 566 N.W.2d at 239 (first emphasis in the original, second emphasis added) (footnote omitted). Of particular significance in this case is the fact that even though the affidavits containing the extraneous information included at least one juror’s statement that the STRESS information “set the tone” for the jury’s deliberations, the trial court accepted as true the substance of those affidavits and denied Nevers’s request for an evidentiary hearing with regard to them. The Michigan Supreme Court similarly accepted the affidavits. In the absence of any evidence or findings of fact regarding whether or how the extraneous information actually affected or influenced the members of the jury, and conceding that there was a “real and substantial possibility” that the extraneous material introduced into the jury room may have affected the jury’s verdict, the Michigan Supreme Court nonetheless concluded that because the evidence of Nevers’s “unjustified intent to commit great bodily harm against Green” was “unimpeachable,” extraneous and prejudicial information reinforcing the evidence of wrongful intent, was harmless. But the evidence of wrongful intent was “unimpeachable” only if the jury chose not to credit Nevers’s testimony. We think that under these circumstances, Nevers need point to nothing more than the juror’s sworn statement that the STRESS information set the tone for the jury’s deliberations and the Michigan Supreme Court’s acknowledgment that this was the kind of material that “has a direct and rational connection between it and an adverse verdict” to demonstrate that the jury’s possession of the extraneous information “‘had [a] substantial and injurious effect or influence in determining the jury’s verdict,’ ” Brecht, 507 U.S. at 637, 113 S.Ct. 1710, and resulted in actual prejudice. Id. We conclude that it would not be debatable among reasonable jurists that the Michigan Supreme Court’s finding of harmless error is so arbitrary, unsupported and offensive to existing precedent as to be outside the realm of plausible credible outcomes. Accordingly, we hold that the district court’s granting of the writ on this ground is affirmed.
Finally, had the trial court undertaken to investigate the claim of extraneous jury influence, Nevers might not have been able to claim constitutional error at all. It is a matter of clearly established Supreme Court precedent that a criminal defendant claiming implied juror bias is entitled to the opportunity to prove actual bias. See Dennis v. United States, 339 U.S. 162, 171-72, 70 S.Ct. 519, 94 L.Ed. 734 (1950) (“Preservation of the opportunity to prove actual bias is a guarantee of a defendant’s right to an impartial jury.”). When a trial court is presented with evidence that an extrinsic influence has reached the jury which has a reasonable potential for tainting that jury, due process requires that the trial court take steps to determine what the effect of such extraneous information actually was on that jury. In Smith v. Phillips, 455 U.S. 209, 102 S.Ct. 940, 71 L.Ed.2d 78 (1982), the Supreme Court made it clear that, while no court has the ability to shield jurors from every outside contact or influence that might affect their votes, nonetheless, the trial court has a duty to take steps to ensure that the jury votes solely on the basis of the evidence presented at trial. “Due process means a jury capable and willing to decide the case solely on the evidence before it, and a trial judge ever watchful to prevent prejudicial occurrences and to determine the effect of such occurrences when they happen.” Id. at 217, 102 S.Ct. 940. The Court went on to hold that a post-trial hearing is sufficient to decide juror partiality in both the state and federal courts, and that in a federal habeas proceeding, the factual finding made by the state trial court in such a hearing is entitled to the presumption of correctness. Id. at 218, 102 S.Ct. 940.
In Nevers’s case, the state trial court accepted as true the affidavits presented by Nevers and Budzyn detailing the extraneous information that had come into the jury’s possession; however, the court denied *374their request for an evidentiary hearing and made no attempt to determine what effect that information actually had on the jury’s verdict. Clearly the failure of the trial court to make such a factual determination does not comport with due process, since, in the absence of any findings of fact with regard to whether or to what extent the jury’s verdict was actually influenced by the extraneous information, Nevers had no specific facts with which to support his claim that the jury did not decide the case solely on the evidence properly before it.
The Michigan Supreme Court did not address the failure of the trial court to accord Nevers his due process right to have a factual determination made regarding the effect of the extraneous information on the jury’s deliberations, proceeding instead, as we have heretofore discussed, to its determination that the jury’s possession of the extraneous material was harmless beyond a reasonable doubt when weighed against the evidence of Nevers’s guilt. To the extent that we understand the distinction between a state court decision that is “contrary to” established Supreme Court precedent and one that involves an “unreasonable application of’ such precedent, we think that the Michigan Supreme Court’s failure even to consider the trial court’s refusal to permit Nevers any opportunity to establish actual jury bias resulted in a decision contrary to established law. This is not a case such as Dennis, 339 U.S. 162, 70 S.Ct. 519, 94 L.Ed. 734, or Chandler v. Florida, 449 U.S. 560, 101 S.Ct. 802, 66 L.Ed.2d 740 (1981), in which the Court refused to set aside the defendant’s conviction because the defendant made no attempt to demonstrate with specificity the “impaired ability of the jurors to decide the case on only the evidence before them,” id. at 581,101 S.Ct. 802. Here Nevers asked for but was denied an eviden-tiary hearing on his claim of jury bias, and was thus effectively prevented from demonstrating with specificity that the extraneous information the jury possessed did in fact impair the ability of the jury to decide the case solely on the evidence properly presented to them through the mechanism of the trial. Whether we look at the failure of the Michigan courts to accord Nevers the requested hearing on his jury bias claim as an instance of the state courts’ failing to apply the correct legal principals as enunciated by the Supreme Court, see Neelley v. Nagle, 138 F.3d at 923, or an instance of the state courts’ issuing decisions squarely in conflict with a Supreme Court precedent, see Green v. French, 143 F.3d at 870, or an instance of the state courts’ ruling in a manner contrary to a specific Supreme Court rule that governs the habeas petitioner’s claim, see O’Brien v. Dubois, 145 F.3d at 24, we must conclude that the result of the state courts’ proceedings is contrary to clearly established federal law as determined by the Supreme Court.25
Because we conclude that the Michigan Supreme Court’s decision with regard to the claim of extraneous jury influence was both contrary to and resulted from an unreasonable application of clearly established Supreme Court precedent, we AFFIRM the judgment of the district court granting the writ on this issue.
CONCLUSION
For the foregoing reasons, we hold that the district court erred in concluding that the petition for a writ of habeas corpus should be granted on the claim that Nevers was denied a fair trial because of pretrial publicity. We hold further that the district court correctly concluded that the petition for a writ of habeas corpus should be granted on the claim that Nevers was denied a fair trial because of extraneous influences on the jury. Accordingly, we AFFIRM the judgment of the district court granting the petition for a writ of habeas corpus.
APPENDIX
SUMMARY OF JURY VOIR DIRE
Juror # 1 acknowledged that his neighbors, with whom he had talked about the *375ease before he was called for jury duty, thought Nevers guilty and wanted him convicted, J.A. at 467 (Voir Dire Tr.). However, Juror # 1 also specifically stated that he had not formed any opinions concerning Mr. Nev-ers, J.A. at 473 (Voir Dire Tr.), and that, although he had read about the case, he questioned the veracity of what he had read “because the papers write anything, the news say [sic] anything, so its hard to believe,” J.A. at 464 (Voir Dire Tr.). Juror # 1 was passed for cause by defense counsel. J.A. at 475 (Voir Dire Tr.).
Juror # 3 acknowledged that he was aware of Malice Green’s death through the newspaper and television. J.A. at 477 (Voir Dire Tr.). Juror # 3 stated that from the “things stated in the media ... everybody seemed to have an opinion,” J.A. at 481 (Voir Dire Tr.), but also said, “After the initial hype of the incident was over, then it just kind of quieted down. I didn’t really' think about it anymore.” J.A. at 486 (Voir Dire Tr.). Juror # 3 admitted that he had discussed the case with his professional colleagues of various races when it first occurred, which included comparison of it to the Rodney King beating, but said that the discussion turned to violence in general, rather than this specific incident. J.A. at 493-98 (Voir Dire Tr.). Juror #3 specifically stated that he had formed no opinion on guilt or innocence, and had no opinion as to whether the encounter between Green and Nevers was “proper” or “improper.” J.A. at 499-500 (Voir Dire Tr.). Defense counsel’s challenge for cause of Juror # 3, an African-American, was denied.
Juror # 5 acknowledged that he had heard about the case from the television and newspapers. J.A. at 508-09, 531 (Voir Dire Tr.). He said that when he learned about the case the first question in his mind concerned what Malice Green had been doing to deserve being stopped by the police. J.A. at 510-11 (Voir Dire Tr.). He also wondered why it had taken “three officers to bring down one person,” J.A. at 512 (Voir Dire Tr.), and stated that he was bothered by this fact because it should not have taken that many police officers, which he thought was “too many,” J.A. at 514 (Voir Dire Tr.), to subdue Green, J.A. at 513 (Voir Dire Tr.). Juror # 5 further stated that from these concerns, he had formed an opinion that “something wrong” had occurred at the scene, but did not know whether it had been Green or the police officers that had been its cause, J.A. at 515, 516 (Voir Dire Tr.). He also repeatedly acknowledged that it was his “opinion that if Malice Green didn’t commit á crime out there, that Mr. Nevers or one of those other defendants did.” J.A. at 517, 520, 522 (Voir Dire Tr.). He acknowledged that based on what he had heard and read about the incident, he had made some fairly firm determinations concerning what had occurred and that he would be surprised if Defense counsel did not put on evidence showing that Malice Green had done something wrong. J.A. at 525-26 (Voir Dire Tr.). He then stated, however, that he could put all of that out of his mind and look only at the evidence presented at trial. J.A. at 526-28. Finally, he acknowledged that based on what he had read and seen on T.V., he believed that Mr. Nev-ers or one of the other officers had caused Malice Green’s death, but that he did not have an opinion as to how it happened, that is, whether it was murder. J.A. at 533-34. Defense counsel’s challenge for cause of Juror # 5 was denied. J.A. at 540-46 (Voir Dire Tr.).
Juror # 15 stated that had no preconceived ideas as to Nevers’s guilt, J.A. at 562 (Voir Dire Tr.), and acknowledged that Mr. Nevers “sits there in [her] eyes innocent.” J.A. at 564 (Voir Dire Tr.). .It does not appear from the transcript pages included in the Joint Appendix that Juror # 15 was challenged for cause.
Juror # 10 acknowledged that she believed one of the three defendants, or maybe all three, were guilty, but had no opinion as to Mr. Nevers’s guilt or innocence at that time. J.A. at 588-91 (Voir Dire Tr.). Definitely, however, she felt that of the three standing trial, “somebody is guilty of something.” J.A. at 591 (Voir Dire Tr.). Later on in her voir dire testimony she acknowledged that she believed that Mr. Green died “because of a senseless act by somebody.” J.A. at 601-02 (Voir Dire Tr.). She stated that she sympathized with Green and his family, but *376that she also sympathized equally with Nev-ers and his family. See J.A. at 606-07 (Voir Dire Tr.). It does not appear from the Joint Appendix that the Defense sought to have Juror # 10 struck for cause.
Juror # 13, who was passed for cause, stated she had no opinion as to guilt or innocence, J.A. at 619, 626 (Voir Dire Tr.), and specifically disagreed with Juror # 10 that one of the defendants had to be guilty of something, J.A. at 619 (Voir Dire Tr.).
Juror # 9 specifically stated that he had no opinion as to guilt or innocence, and that he did not form an opinion from the media coverage because he felt as if the “whole story” had not been told yet. J.A. at 649 (Voir Dire Tr.) He had not had any discussions with friends or co-workers about the case, other than briefly mentioning it to his girlfriend. J.A. at 648 (Voir Dire Tr.). Juror # 9 was very worried about having his identity disclosed in the media after the trial was over. See J.A. at 641-43, 658-62, 664 (Voir Dire Tr.). He was worried about being harassed by either the group protesting outside of the courthouse during voir dire or by police officers, depending upon how the verdict came out. J.A. at 661-62 (Voir Dire Tr.). He said, however, that his concern would not affect his ability to render a decision, and that if asked for a verdict right then it would have been not guilty. J.A. at 662-63 (Voir Dire Tr.). He was passed for cause. J.A. at 657, 675 (Voir Dire Tr.).
Juror # 14 specifically stated that he had not formed any opinions about what had occurred in this case because he was not very familiar with the case. J.A. at 695 (Voir Dire Tr.). He was passed for cause. J.A. at 688, 702 (Voir Dire Tr.).
Juror # 4 had heard about the case from “some rumblings in the neighborhood,” but not from newspaper or T.V., which he neither read nor watched. J.A. at 707-08 (Voir Dire Tr.). He repeatedly said that he had no opinions on the case whatsoever, J.A. at 704, 708 (Voir Dire Tr.), and that if pressed to make a decision he would have to find Nev-ers not guilty because he had not heard any evidence, J.A. at 709 (Voir Dire Tr.). He was passed for cause. J.A. at 722, 725 (Voir Dire Tr.).
Juror # 16 admitted that he had heard about the case and had discussed it with his friends, whose opinions had varied from guilty to not guilty. J.A. at 736-37 (Voir Dire Tr.). He then stated that his personal opinion was that, from the facts presented in the media, Nevers was not guilty. J.A. at 737, 738 (Voir Dire Tr.). Juror # 16 was passed for cause. J.A. at 745, 751 (Voir Dire Tr.).
Juror # 12 stated that she had heard about the case through the T.V. and radio, but had not read about it. J.A, at 763 (Voir Dire Tr.). When asked what she thought about the case when she learned of it, she replied, “I thought that when the point came for a trial that it was going to be very, very grave because people would really need to hear the facts to make sure it was a fair trial and not based on emotion or public opinion.” J.A. at 769 (Voir Dire Tr.). From the remarks she had heard she believed public opinion was divided. J.A. at 769-70 (Voir Dire Tr.). She stated that she had not formed an opinion, and wanted to wait to do so until the facts came out at trial. J.A. at 771, 776 (Voir Dire Tr.). She was passed for cause. J.A. at 781, 794 (Voir Dire Tr.).
Juror # 2 stated that she had heard about the incident before reporting for jury duty, and that she had heard that Green “had died as a result of a beating from the officer.” J.A. at 810 (Voir Dire Tr.). She also acknowledged having heard the Mayor’s comment that the police officers were “murderers,” but stated that she thought the Mayor had gone “overboard,” had “jumpfed] the gun,” and had acted irresponsibly when he said that. J.A. at 810-11 (Voir Dire Tr.). She further stated that she had not formed an opinion and that she had not really worried too much about the case. J.A. at 812-13 (Voir Dire Tr.). Finally, she stated that she presumed Nevers to be innocent, J.A. at 828 (Voir Dire Tr.), that if asked for her vote she would have to say not guilty, id., and that she had no opinion at that time concerning guilt or innocence, J.A. at 831 (Voir Dire Tr.). She was passed for cause. J.A. at 833, 845 (Voir Dire Tr.).
*377Juror # 8 stated she heard from the media that several police officers were involved in the killing of Malice Green, that the media was portraying Nevers as guilty, J.A. at 858-59 (Voir Dire Tr.), and that it was the prevailing opinion of her co-workers that Nevers was guilty, J.A. at 860-61 (Voir Dire Tr.). The media’s portrayal of the officers had bothered her, she said, because she felt those making such judgments had not heard the officers’ side of the story and thus had not given them a fail1 chance. J.A. at 859 (Voir Dire Tr.). She also stated that she did.not have an opinion as to guilt or innocence, J.A. at 849 (Voir Dire Tr.), that she had not formed an opinion because she had not had a chance to hear the officers’ sides of the story, J.A. at 859 (Voir Dire Tr.), and that it was both her legal and personal opinion that, as of that moment, Nevers was not guilty. J.A. at 862 (Voir Dire Tr.). Juror # 8 was passed for cause. J.A. at 854, 865 (Voir Dire Tr.).
Juror # 6 stated that she had read about the case in the newspapers, J.A. at 877 (Voir Dire Tr.), and that she had no opinion as to guilt or innocence, J.A. at 880 (Voir Dire Tr.). The people she worked with had all different opinions, ranging from guilty to not guilty. J.A. at 883 (Voir Dire Tr.). She was passed for cause. J.A. at 887 (Voir Dire Tr.).
Juror # 7 stated that she had heard about the ease through T.V., and admitted that her first reaction was to compare it to the Rodney King incident, not out of any notion of the beating’s having been racially motivated, but rather out of a fear that there might be a riot if the officers were acquitted. J.A. at 901-02 (Voir Dire Tr.). She further stated, however, that she no longer held that fear after talking with her co-workers, as it was their opinion that if the officers were acquitted, the people of Detroit would not react as the citizens of Los Angeles had. See J.A. at 902-03 (Voir Dire Tr.). She stated that if asked at that moment she would have to find Nevers not guilty. J.A. at 906 (Voir Dire Tr.). She was passed for cause. J.A. at 893, 911 (Voir Dire Tr.).
Juror #11 stated that he had heard very little about the case, and had not formed any opinions as to guilt or innocence at that time. See J.A. at 913, 924 (Voir Dire Tr.). He was passed for cause. J.A. at 924, 928 (Voir Dire Tr.).
The limited testimony in the Joint Appendix of jurors who were not chosen to serve on the panel tends to show that this crime and the media’s portrayal of it did affect some potential jurors strongly. One potential juror admitted that she had made up her mind about Nevers’s guilt and that she felt she could not give him a fair trial. She was excused for cause. J.A. at 888 (Voir Dire Tr.). Another potential juror, who was struck peremptorily by defense counsel, J.A. at 675 (Voir Dire Tr.), had much to say about his opinion concerning “community sentiment.” Defense counsel asked him his opinion on “what the community is saying,” and he replied, “Well, the community, basically what I am hearing is that he is guilty. That’s what I’m hearing.” Defense counsel then stated, “Okay. In other words, at least, you have heard in the community that the general consensus of the community is that he — by he you mean Mr. Nevers, is guilty?” to which the juror stated, “Right.” J.A. at 550 (Voir Dire Tr.). This potential juror also acknowledged that he had heard the Mayor say that Nevers was guilty, that he had a high regard for the mayor, and that when he heard the Mayor make that statement, he thought that the Mayor must have based it on some facts indicating Mr. Nevers’s guilt. J.A. at 551-53 (Voir Dire Tr.). Finally, this potential juror acknowledged that based on what he had heard in the community, his own initial opinion was that Nevers was guilty. J.A. at 553 (Voir Dire Tr.). Another potential juror stated that he read in the media “[tjhings basically intimating that [Nevers] is guilty.” J.A. at 547 (Voir Dire Tr.). This same individual, however, stated that he did not have an opinion concerning guilt or innocence because he did not necessarily believe what he read in the media. J.A. at 546a (Voir Dire Tr.). Another potential juror admitted that initially she had formed an opinion as to Nevers’s guilt and that the first reports had made her angry, but she thought she could set these aside because she thought Nevers deserved a fair trial. J.A. at 929-30 (Voir Dire Tr.). Another potential juror also admitted that she had formed “a slight opin*378ion” and was “upset” when she initially heard about Green’s death, but she also thought she could set that aside. J.A. at 930 (Voir Dire Tr.). Still others indicated that they had not formed opinions because they could not trust the media because the media had not been there when the alleged crime was committed. See J.A. at 934-36 (Voir Dire Tr.). One of these, however, admitted that he was angered when he first learned of Green’s death. J.A. at 935-37 (Voir Dire Tr.).
. Officer Robert Lessnau was charged with assault with intent to murder and was tried with Budzyn and Nevers, but by the bench instead of a separate jury. The trial court acquitted Less-nau.
. Nevers v. Killinger, 990 F.Supp. 844 (E.D.Mich.1997); People v. Budzyn, 456 Mich. 77, 566 N.W.2d 229 (1997).
. The Michigan Supreme Court noted:
The trial judge did not select the movies or approve the selections himself, but he took responsibility for the action taken by the employees of the court. The trial judge disqualified himself on ruling on defendants' motion. *357The Chief Judge of Recorder’s Court referred the motion to a third judge who heard the motion and denied it.
Budzyn, 566 N.W.2d at 234 n. 7.
. The Budzyn jury deliberated eight days before returning a verdict of guilty. Budzyn received a sentence of eight to eighteen years. As noted previously, Budzyn’s conviction was reversed and remanded for a new trial by the Michigan Supreme Court due to extraneous jury influences that the court held could not be found to be harmless error.
. The district court in the case sub judice chose to employ the Fifth and Seventh Circuits' “categorization” approach (although it concluded that the result would be the same under the First Circuit's approach, discussed infra), see Nevers, 990 F.Supp. at 851 & n. 4. One month after the district court’s determination, another judge in the Eastern District of Michigan also found the Fifth and Seventh Circuits’ categorization approach to be appropriate. See Barker v. Yukins, 993 F.Supp. 592, 600-01 (E.D.Mich.1998).
. In a footnote Neelley acknowledges that "[t]he overlap between the statute and Teague is not complete,” Neelley, 138 F.3d at 923 n. 3, but found “no need to determine the exact overlap between Teague and § 2254 in this case because the Supreme Court case law governing Neelley’s contentions was manifestly clear at all relevant times,” id.
. But see Herbert v. Billy, 160 F.3d 1131, 1134— 35 (6th Cir.1998) (stating that after the advent of AEDPA, "[a] district court or court of appeals no longer can look to lower federal court decisions in deciding whether the state decision is contrary to, or an unreasonable application of, clearly established federal law”) (emphasis added).
. In Harpster we discussed the First Circuit case of Martin v. Bissonette, No. 96-1856, 1997 WL 280602 (1st Cir. May 29, 1997). This case was withdrawn and replaced by Martin v. Bissonette, 118 F.3d 871 (1st Cir.1997), which does not discuss the approach initially discussed in the withdrawn opinion. However, comparison of the original Martin to O’Brien shows that the standard enunciated in Martin is the one adopted and applied in O'Brien.
.As we shall more fully address in a subsequent section of this opinion, the extraneous jury influence issue also involves an issue of whether the state courts' failure to malee any factual inquiry into the actual effect, if any, of the extraneous information on the jury’s verdict was contrary to clearly established law. This issue, however, may have been waived; it is difficult to determine whether it was raised on direct appeal, and was raised only by implication in the habeas proceedings. We address it because we view it as an inextricable aspect of the jury influence issue as specifically addressed by the state courts.
. Accord Neelley, 138 F.3d at 924 ("[T]he mere fact that a district court disagrees with a state court does not render that state court’s decision 'unreasonable'; certainly two courts can differ over the proper resolution of a close question without either viewpoint being unreasonable.”); Herman v. Cooper, 109 F.3d 330, 334 (7th Cir.), cert. denied, - U.S. -, 118 S.Ct. 72, 139 L.Ed.2d 32 (1997) ("[T]he fact that we might disagree with the state court's determination that the prosecutor's remarks did not deny Hennon due process because they are unlikely to have swayed the jury — might think the determination incorrect — would not carry the day for him.”); Drinkard, 97 F.3d at 768 (“This 'unreasonable application’ standard of review of a state court decision must mean more than that a federal court may grant habeas relief based on its simple disagreement with the state court decision; this would amount to nothing more than a de novo review.”).
. The Michigan Supreme Court affirmed this part of the court of appeals decision on the basis of that court’s opinion. Budzyn, 566 N.W.2d at 234.
. At least 53,000, and maybe as many as 97,-000, people were estimated to have seen the tape of the televised confession at least once. See gRideau, 373 U.S. at 724, 83 S.Ct. 1417.
. Rideau was cited in a footnote as a precedent "describing situations in which state procedures are inadequate to uncover bias.” Patton, 467 U.S. at 1038 n. 13, 104 S.Ct. 2885.
. Accord DeLisle v. Rivers, 161 F.3d 370, 377-78 (6th Cir.1998) (stating "it was only 'the televising of a defendant in the act of confessing to a crime' that the Rideau Court held 'was inherently invalid under the Due Process Clause of the Fourteenth Amendment.’ ”) (quoting Estes v. Texas, 381 U.S. 532, 538, 85 S.Ct. 1628, 14 L.Ed.2d 543 (1965)) (emphasis added in DeLisle).
. Indeed, the district court’s description of the media coverage alone encompasses over eight pages of the Federal Supplement. See 990 F.Supp. at 855-63.
.One pretrial incident was a three-day “inquest” broadcast live over T.V. and radio at which “Sheppard was brought into the room by police who searched him in full view of several hundred spectators," he was "questioned for five and one-half hours about his actions on the night of the murder,” his "counsel were present ... but were not permitted to participate,” and “[w]hen Sheppard’s chief counsel attempted to place some documents in the record[] he was forcibly ejected from the room by the Coroner, who received cheers, hugs, and kisses from ladies in the audience.” Sheppard, 384 U.S. at 339-40, 86 S.Ct. 1507.
. Accord DeLisle, 161 F.3d 370, 384-85 ("There is no allegation by DeLisle that his trial took place under the conditions of total chaos that prevailed in Estes and Sheppard, and a review of those cases leaves no room for doubt that it was that chaos that drove those decisions. See Murphy, 421 U.S. at 799, 95 S.Ct. 2031.”) (emphasis added).
. The articles described Nevers's and Budzyn’s reputations as "hard line police officers,” id., who often used force and intimidation “as tools of persuasion,” J.A. at 424 (Ex. M32 (Detroit News Nov. 8 Article)), thus garnering them the nicknames "Starsky and Hutch” in the neighborhood in which they patrolled, id. The articles also mentioned previous instances of police brutality alleged against Nevers, see Nevers, 990 F.Supp. at 855-57, and Nevers’s former membership in STRESS (Stop the Robberies Enjoy Safe Streets), "a controversial undercover unit that was involved in alleged harassment of young black men before it was disbanded by Former Detroit Mayor Coleman Young.” Id. at 855. Another article mentioning Nevers’s STRESS participation "specifically referred to the officers involved in STRESS as major culprits of police brutality and stated that the controversial unit was involved in the deaths of twenty people, seventeen of whom were black.” Id. at 855-56. Other articles compared the incident to the Rodney King beating, see id. at 856. Many of the articles and editorials involved criticism of the police department generally, especially concerning the way in which it had previously handled police officers with histories of violent behavior. The Mayor of Detroit even stated, on national television, "A young man who was under arrest was literally murdered by police.” The Mayor's statement was subsequently reported in the newspaper. See id. at 859.
.See supra note 18.
. The court refused to consider the fourth affidavit, which the juror who produced it had refused to sign. Nevers had attempted to remedy the lack of a signature by submitting an affidavit from an attorney who swore that the juror had verbally adopted the statements made in the affidavit but had refused to sign out of fear that it would bring renewed media and public pressure upon her. The court, however, correctly adduced that such an attesting affidavit would make the juror’s affidavit inadmissible hearsay. Budzyn, 566 N.W.2d at 236 n. 14.
. After the Michigan Supreme Court had determined that the appellants had shown the error to be constitutional error the court stated, “Because defendants have carried their initial burden, we must decide whether any error was hannless beyond a reasonable doubt." Budzyn, 566 N.W.2d at 240 (emphasis added). While the Michigan Supreme Court did not specifically cite Chapman or any other Supreme Court precedent (it cited Ninth Circuit precedent instead), we agree with the district court that it “clearly applied the proper Chapman harmless error standard on direct review,” Nevers, 990 F.Supp. at 870 (citing Budzyn, 566 N.W.2d at 240). See Clemons v. Mississippi, 494 U.S. 738, 744, 753, 110 S.Ct. 1441, 108 L.Ed.2d 725 (1990) (finding that Mississippi Supreme Court had applied Chapman standard, even though it had not cited Chapman, from its statement, “ 'We likewise are of the opinion beyond a reasonable doubt that the jury's verdict would have been the same with or without the 'especially heinous, atrocious or cruel' aggravating circumstance.' ”); Barker v. Yukins, 993 F.Supp. 592, 603 (E.D.Mich.1998) (finding that although the Michigan Supreme Court did not cite to decisions of the United States Supreme Court, it clearly applied the Chapman harmless error analysis, and thus its decision was reviewable under the "unreasonable application” clause of § 2254(d)(1)).
. The Fourth Circuit, in Green v. French, 143 F.3d 865 (4th Cir.1998), combined the two without specifically addressing the question of whether one or the other must apply. It held
that the North Carolina Supreme Court correctly concluded under Chapman that the trial court's failure to instruct the sentencing jury as to ... non-statutory mitigating factors was harmless under the standard for direct review, and, a fortiori, that the trial court’s error was harmless under the less exacting standard for federal habeas review of state court convictions under Brecht v. Abrahamson ....
Id. at 893 (citation omitted). The court then articulated its reasons for so finding and concluded by stating, "Green therefore has not met his burden of establishing that these errors were prejudicial under the Brecht standard.” Id. at 894.
. Chapman v. California, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967), held that "before a federal constitutional error can be held harmless, the court must be able to declare a belief that it was harmless beyond a reasonable doubt.” Id. at 24, 87 S.Ct. 824. In so holding, the Court said, it did "no more than adhere to the meaning of our Fahy case.” Id. Fahy v. Connecticut, 375 U.S. 85, 86-87, 84 S.Ct. 229, 11 L.Ed.2d 171 (1963) held that "the question is whether there is a reasonable possibility that the evidence complained of might have contributed to the conviction.”
. Although Hughes does not directly cite Chapman v. California, it does rely on Marino v. Vasquez, 812 F.2d 499 (9th Cir.1987), which cites to Chapman.
. We suppose that where a state court decision is contrary to clearly established federal law, the error may still be found to be harmless. In the case before us today, however, the denial of due process could not be harmless error under either the Chapman or Brecht standards because it resulted in the denial of Nevers's right to demonstrate that the admittedly prejudicial material improperly in the jury’s possession actually affected the jury’s verdict.