GREGORY, Circuit Judge,
concurring in the judgment on the McKoy issue:
During the sentencing phase of Allen’s trial, the judge instructed the jury that it must unanimously find the existence of any mitigating circumstance before considering that circumstance in determining whether Allen should be sentenced to life imprisonment or death. This was the same instruction that the Supreme Court found unconstitutional in McKoy v. North Carolina, 494 U.S. 433, 110 S.Ct. 1227, 108 L.Ed.2d 369 (1990). In McKoy, the Court held this jury instruction unconstitutional because the “unanimity requirement ... prevents] the sentencer from considering all mitigating evidence.” 494 U.S. at 435, 110 S.Ct. 1227. Thus, on direct review in this case, the Supreme Court vacated Allen’s sentence and remanded the case to the North Carolina Supreme Court in light of McKoy. Allen v. North Carolina, 494 U.S. 1021, 1021, 110 S.Ct. 1463, 108 L.Ed.2d 601 (1990).
On remand, the North Carolina Supreme Court held that the unanimity instruction was indeed unconstitutional under McKoy. State v. Allen, 331 N.C. 746, 417 S.E.2d 227, 228 (1992). The court nonetheless affirmed Allen’s death sentence finding the error “harmless beyond a reasonable doubt” based solely on the jurors’ responses to a post-verdict poll. Id. at 228.
Allen then filed this motion for habeas relief, challenging the North Carolina Supreme Court’s finding that the concededly unconstitutional instruction constituted harmless error. The district court, although eventually rejecting Allen’s claim, granted a certificate of appealability with respect to this issue. Thus, the district court recognized that Allen had made a substantial showing of the denial of a constitutional right. See Slack, 529 U.S. at 483, 120 S.Ct. 1595. Because Allen has made this showing, we, the plurality, proceed directly to the merits of his claim.
As noted above, we may grant Allen habeas relief only if his claim meets the criteria for relief as detailed in § 2254(d)(1) of the AEDPA. Pursuant to that statute, the state court decision must either be contrary to, or involve an unreasonable application of, clearly established federal law as determined by the Supreme Court. See Williams, 529 U.S. at 411, 120 S.Ct. 1495. As the Court has recently reiterated, “ § 2254(d)(l)’s ‘contrary to’ and ‘unreasonable application’ clauses have independent meaning,” and provide independent bases for habeas relief. See Bell v. Cone, 535 U.S. 685, 694, 122 S.Ct. 1843, 152 L.Ed.2d 914 (2002). A federal habeas court may grant relief under the “contrary to” clause if “the state court applies a rule different from the governing law set forth in [Supreme Court precedent], or if it decides a case differently than [the Supreme Court has] done on a set of materially indistinguishable facts.” Id. Alternatively, a writ can issue under the “unreasonable application” clause “if the state court correctly identifies the governing legal principle from [Supreme Court precedent] but unreasonably applies it to the facts of the particular case.” Id. A writ cannot issue merely because a “state-court decision applied clearly established federal law erroneously or incorrectly. Rather that appli[341]*341cation must also be unreasonable.” Williams, 529 U.S. at 411, 120 S.Ct. 1495.
In this case, Allen is not entitled to relief on his jury instruction claim under the “contrary to” clause. The state court properly identified the governing law— both the holding in McKoy that North Carolina law imposing a unanimity requirement on mitigating circumstances is unconstitutional, and the proper harmless error standard.1 Moreover, Allen does not contend the state court unreasonably applied the governing law with respect to McKoy itself. After all, the state court held that the unanimity instruction employed in Allen’s sentencing proceeding was unconstitutional. Allen does contend, however, that he is entitled to relief under the “unreasonable application” clause because the state court unreasonably applied the harmless error standard, as articulated in Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967), to the facts of his case.
The North Carolina Supreme Court found the McKoy error to be “harmless beyond a reasonable doubt” based solely on the results of a post-verdict poll of the jury. Allen, 417 S.E.2d at 228. The court recounted:
The clerk then polled the jurors by stating to each of them each mitigating circumstance and whether it was found or not. The clerk asked each juror whether these were the answers to “your issues,” whether these were still the answers to the issues and whether he or she still assented thereto. Each juror answered in the affirmative.
It appears from this poll that the jury was unanimous as to each of the mitigating circumstances which the jury failed to find.
Id. (emphasis added). Thus, on the basis of each juror’s responses to a routine post-verdict jury poll, N.C. Gen.Stat. § ISA-1238 (providing that after return of a criminal verdict, upon motion of either party or judge’s own motion, each juror be polled as to whether the verdict announced is his or her verdict), the court reasoned that giving the unconstitutional jury instruction constituted harmless error beyond a reasonable doubt. Allen, 417 S.E.2d at 228. The North Carolina Supreme Court concluded that resentencing was not necessary because this poll established that no individual juror voted for any of the seven mitigating circumstances that the collective jury had not unanimously found.
In doing so, the North Carolina Supreme Court placed inordinate reliance on a poll that was never designed to, nor did it, cure the McKoy error. Rather, as is evident from even a cursory review of its substance, the poll merely confirmed that each juror followed the trial court’s unconstitutional instruction in sentencing Allen to death.
When the clerk polled each juror, she read the entire verdict form, which not only specifically asked, “Do you unanimously find from the evidence the existence of one or more of the following mitigating circumstances?”; but also stated, [342]*342“In the space after each mitigating circumstance, write ‘yes/ if you unanimously find that mitigating circumstance by a preponderance of the evidence. Write, ‘No,’ if you do not unanimously find that mitigating evidence.” (J.A. 137) (emphasis added). After reading through all of the proffered mitigating circumstances, the clerk queried the jurors if there were “[a]ny other circumstance or circumstances arising from the evidence which you, the jury, deem to have mitigating value.” (J.A. 106) (emphasis added). That is, in the verdict form, the term “you” was defined in the collective, referring to “you, the jury.” Importantly, no juror was ever asked how he or she individually voted on any of the mitigating circumstances. Instead, after reading the verdict form along with the jury’s collective responses, the clerk asked each juror, “Are these the answers to your issues?,” “Are these still the answers to the issues?,” and “[D]o you still assent thereto?” (J.A. 107).
The clerk’s questions imply that jurors were not being asked for their individual votes on the mitigating circumstances, but rather, that they were being asked if they still assented to the jury’s collective decisions. This is a vital distinction. A juror might recognize that he would have found the existence of a mitigating circumstance had he been permitted to consider the question on his own. However, he would also recognize that the jury did not unanimously find the existence of that circumstance, and so he would be compelled to “still assent” to the jury’s collective decision not to consider that circumstance, regardless of how he individually voted on the matter.
As the Supreme Court has explained, the unanimity requirement is unconstitutional because it “allows one holdout juror to prevent the others from giving effect to evidence that they believe calls for a sentence less than death.” McKoy, 494 U.S. at 439, 110 S.Ct. 1227 (internal quotation marks omitted). “Moreover, even if all 12 jurors agree that there are some mitigating circumstances, ... [a unanimity requirement] prevents them from giving effect to evidence supporting any of those circumstances in their deliberations ... unless they unanimously find the existence of the same circumstance.” Id. The damage done by imposing a unanimity requirement is that it prevents a juror who found the existence of a mitigating circumstance from fully participating in the jury’s deliberations as to death. The requirement bars him or her from bringing his or her appreciation of that mitigating circumstance to bear on the jury’s collective sentencing decision.
It is at this critical stage — where we must attempt to understand the effects of the unanimity requirement on the jury’s internal deliberations — that the post-verdict poll fails to provide any meaningful insight. The poll did not, and could not, shed light on the critical issue of whether any individual juror voted for a given mitigating factor. In short, the post-verdict poll reveals nothing about the individual votes by the jurors on the seven mitigating circumstances that the jury did not unanimously find. As Justice Blackmun observed, such a poll “tells us nothing about how the juror would have voted — either on a particular mitigating circumstance or on the ultimate life-or-death question — had he been instructed that he could give effect to all the mitigating evidence, as the Constitution requires.” Price v. North Carolina, 512 U.S. 1249, 1251, 114 S.Ct. 2777, 129 L.Ed.2d 888 (1994) (Blackmun, J., concurring in the grant of certiorari).
Therefore, we find that in determining that this post-verdict poll rendered the McKoy error harmless, the North Carolina Supreme Court’s application of [343]*343Chapman’s governing legal principles was objectively unreasonable, resulting “in a decision that ... involved an unreasonable application of clearly established Federal law.” 28 U.S.C. § 2254(d)(1).2 Accordingly, “the requirements of § 2254(d) ... pose no bar to granting petitioner habeas relief.” Wiggins, 123 S.Ct. at 2539.3
Although the jury poll was the sole basis for the state court’s (unreasonable) determination that the McKoy error was harmless, and it was also the only argument raised by the government on appeal to support this (unreasonable) determination, we may affirm on any ground supported by the record, even if it has not been raised by the parties. See In re A.H. Robins Co., 880 F.2d 709, 748 (4th Cir.1989), abrogated on other grounds by Amchem Prods., Inc. v. Windsor, 521 U.S. 591, 117 S.Ct. 2231, 138 L.Ed.2d 689 (1997). It is therefore appropriate for us to engage in the additional analysis mandated by Brecht v. Abrahamson, 507 U.S. 619, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993).
Under Brecht, 507 U.S. 619, 113 S.Ct. 1710, a habeas petitioner is still not entitled to relief, despite having satisfied AEDPA criteria, unless the state court’s unreasonable application of federal law had a “substantial and injurious effect or influence in determining the jury’s verdict.” Id. at 637, 113 S.Ct. 1710.4 The Brecht [344]*344standard, adopted from and first explicated in Kotteakos v. United States, 328 U.S. 750, 66 S.Ct. 1239, 90 L.Ed. 1557 (1946), requires a greater showing of prejudice than Chapman, Brecht, 507 U.S. at 637, 113 S.Ct. 1710, but a lesser showing than the “reasonable probability” standard. Kyles v. Whitley, 514 U.S. 419, 435-36, 115 S.Ct. 1555, 131 L.Ed.2d 490 (1995). Under the Brecht standard, the inquiry cannot be merely whether sufficient evidence “support[s] the result,” or whether the jurors were “right in their judgment, regardless of the error or its effect upon the verdict,” Kotteakos, 328 U.S. at 764-65, 66 S.Ct. 1239; rather the proper inquiry must focus on whether the error had a “substantial and injurious effect or influence in determining the jury’s verdict.” Brecht, 507 U.S. at 631-39, 113 S.Ct. 1710. If a court “cannot say, with fair assurance, after pondering all that happened without stripping the erroneous action from the whole, that the judgment was not substantially swayed by the error, it is impossible to conclude that substantial rights were not affected.” Kotteakos, 328 U.S. at 765, 66 S.Ct. 1239.
Neither Brecht nor Kotteakos was a capital case, and thus they only addressed the question of the defendant’s guilt or innocence, not the question of whether the defendant should be executed. Accordingly, we recognize that the Brechh-Kotteakos standard must be somewhat modified when applied in circumstances like those at issue here — the sentencing phase of a capital case. In this context, that standard requires a reviewing court to determine whether it can say “with fair assurance” that an error did not “substantially sway[]” the response of the jury to the question put before it, ie., should the defendant receive the death penalty. Kotteakos, 328 U.S. at 765, 66 S.Ct. 1239. Moreover, in this context a reviewing court must undertake the Brecht analysis mindful of the Supreme Court’s recognition that an appellate court faces certain difficulties when “determining sentencing questions in the first instance.” See Clemons v. Mississippi, 494 U.S. 738, 754, 110 S.Ct. 1441, 108 L.Ed.2d 725 (1990); accord Caldwell v. Mississippi, 472 U.S. 320, 330, 105 S.Ct. 2633, 86 L.Ed.2d 231 (1985).
With these principles in mind, we turn to the application of the standard required by Brecht and Kotteakos to the facts at hand. We note at the outset that the jury poll at issue in this case was so woefully inadequate, for all the reasons outlined above, that it alone clearly did not render the McKoy error harmless even under the Brecht standard. We next examine whether, based on a more robust review of the record evidence beyond the jury poll, the McKoy error had a “substantial and injurious effect or influence” on the jury’s sentencing recommendation in Allen’s case. In particular, we ask whether we can conclude, with “fair assurance,” that had the jurors been permitted to independently assess the additional mitigating evidence presented at Allen’s sentencing proceeding, and weigh that evidence, along with the mitigating factors unanimously found, against the aggravating circumstances, they would nevertheless have voted for the death penalty. We cannot.
In this case, the trial judge found sufficient evidence to submit two aggravating factors and ten mitigating factors to the jury.5 The jurors unanimously found the [345]*345existence of two aggravating factors: (1) murder committed for the purpose of avoiding a lawful arrest and (2) murder of a law enforcement officer while engaged in the performance of his official duties. They also unanimously found three mitigating factors to be present: (1) Allen had no history of crimes involving deadly weapons, (2) he expressed remorse for the victim’s death, and (3) he earned a GED. However, because of the unconstitutional McKoy instruction, each juror was deprived of the ability to individually weigh the other seven mitigating factors.
North Carolina statutory law requires jurors to determine: (1) whether any sufficient aggravating circumstances exist; (2) whether any sufficient mitigators exist; and (3) after weighing these considerations, whether the defendant should be sentenced to life imprisonment or death. N.C. Gen.Stat. § 15A-2000(b); see also State v. Hunt, 357 N.C. 257, 582 S.E.2d 593, 598 (2003); State v. Barfield, 298 N.C. 306, 259 S.E.2d 510, 542 (1979). This sentencing scheme expressly prohibits the jury from considering any aggravating factors not listed in the statute (which does not contain a nonstatutory category or a catch-all provision), and limits the jurors’ deliberations to weighing the relevant aggravating and mitigating factors. See N.C. Gen.Stat. § 15A-2000(b), (e); see also Hunt, 582 S.E.2d at 598; Barfield, 259 S.E.2d at 542. Our duty under Brecht is to determine what effect the admitted error had on the jury’s decision-making process. In doing so, we recognize that the jury’s process was carefully circumscribed by statutory dictates, limiting deliberation to two aggravating factors — that the murder was committed for the purpose of avoiding a lawful arrest and that the murder was of a law enforcement officer while engaged in the performance of his official duties — thus there is no legal basis to weigh the overall circumstances of the crime outside these two factors.
So we proceed, as we must, under the North Carolina statutory scheme, and in doing so further recognize that a North Carolina court may not impose death if a single juror votes in favor of life imprisonment. Indeed, “[i]f the jury cannot, within a reasonable time, unanimously agree to its sentence recommendation, the judge shall impose a sentence of life imprisonment .... ” N.C. Gen.Stat. § 15A-2000(b). Accordingly, we must now assess whether we can say “with fair assurance,” Kottea-kos, 328 U.S. at 765, 66 S.Ct. 1239, that not a single resolute juror would have voted for a life sentence.
In addition to the mitigating factors unanimously found by the jury (that Allen earned a GED degree, had no history of crimes involving deadly weapons, and expressed remorse for the victim’s death) the defense offered evidence of seven other mitigating factors: (1) Allen’s age, (2) his lack of significant criminal history, (3) his rearing in a single-parent home, (4) his parenting of three children, (5) his history of gainful employment, (6) his status as a supporting parent to a child, and (7) a catch-all category for other circumstances that the jury deemed to have mitigating value.
During Allen’s sentencing proceeding, substantial evidence was presented to support several of these additional mitigating factors. However, we do not believe that [346]*346any reasonable juror could have concluded that Allen presented sufficient evidence that his age — thirty at the time of the murder — should count as a sufficient mitigating factor. After all, a person of thirty cannot reasonably be characterized as too young to appreciate the seriousness of his crime. Further, although Allen’s criminal history — convictions for shoplifting, breaking and entering, and burglary — does not involve convictions for violent offenses, and so in that respect is arguably “not significant” and thus could be considered mitigating, we do not believe any reasonable juror could conclude that this evidence added enough to the mitigating factors unanimously found (given that the jury had already found Allen’s lack of history of crimes involving deadly weapons to be a mitigating factor) to outweigh the aggravating circumstances. Thus, if the only additional mitigating evidence Allen offered pertained to his age and prior criminal history, we believe the McKoy error here would be harmless.
But of course that is not the situation here. Allen offered evidence that from the time he was a young boy he was raised in a single-parent home, and although he had more frequent contact with his father later in life, he only saw his father three or four times a year until he entered sixth grade. Further, Allen remained employed when he was not incarcerated, a fact that the dissent does not contest. And, in addition to the indisputable fact that Allen fathered three children, there was abundant evidence that he maintained an ongoing, supportive relationship with his eldest child, Timothy Jr. Indeed, Timothy Jr. was present at his father’s sentencing hearing, at which his mother, Allen’s long divorced ex-wife, unequivocally testified that Allen had a “very good” relationship with their son.
Furthermore, the jury heard a good deal of additional testimony that could qualify as another “circumstance ... deem[ed] to have mitigating value.” N.C. Gen.Stat. § 15A-2000(f)(9). For example, Allen presented substantial evidence that his parents engaged in physical fights in his presence, which petrified him. In fact, their ultimate separation and divorce resulted from one such fight that occurred during Allen’s third birthday party in which the police were called.6 Similarly, the jury heard evidence that Allen was a shy, timid person, bullied by others, who did not use firearms, and was unwilling or unable to fight back. And although Allen was never physically abused as a child, from an early age he lost himself in alcohol and drugs.
Given this evidence, we cannot say “with fair assurance” that no reasonable juror could have found these circumstances to have mitigating value. Nor can we conclude that precluding the jurors from individually weighing these last five mitigating factors plus the three unanimously found mitigators against the two aggravating circumstances did not “substantially sway[ ]” their decision to recommend that Allen be executed. Kotteakos, 328 U.S. at 765, 66 S.Ct. 1239. Rather, there is a sufficient likelihood that at least one juror, after hearing this evidence and properly weighing all eight colorable mitigating factors against the two aggravating factors, would have been persuaded to spare Allen’s life. This is not to suggest that Allen did not commit a horrible crime, nor that jurors would not properly consider the aggravating circumstances — murder of a law enforcement officer performing his duties for [347]*347the purpose of avoiding arrest — to be exacerbating. Nevertheless, we cannot say with any “fair assurance” that a reasonable juror would not have concluded, after properly weighing the aggravating factors against all eight mitigating factors, that Allen’s life should be spared because, despite the terrible crime he committed, Allen exhibited some redeeming qualities— including efforts to overcome a less than ideal childhood and lead a productive life by obtaining a GED, finding employment, and providing ongoing emotional support for a son. There is a sufficient likelihood that these efforts, particularly given Allen’s remorse for Trooper Worley’s death, the loyalty Allen inspired in his ex-wife and son, demonstrated by his ex-wife’s supportive testimony on his behalf and the presence of his son at the sentencing hearing, and the uncontradicted testimony that Allen had never before used a firearm to commit a crime, could well have swayed at least one of the twelve jurors to determine that Allen should be imprisoned for the rest of his life, rather than executed.
As the Supreme Court recognized in Mills v. Maryland, 486 U.S. 367, 374, 108 S.Ct. 1860, 100 L.Ed.2d 384 (1988), even if all twelve jurors agreed that some of the non-unanimous mitigating circumstances were present, and further agreed that those mitigating circumstances (combined with the unanimously found mitigators) outweighed the aggravating factors, if the jurors disagreed as to the particular mitigating circumstances because of the unconstitutional instruction, they would never be permitted to “engage in the [proper] weighing process or any deliberation on the appropriateness of the death penalty.” Id. Indeed, even if eleven jurors agreed that five non-unanimous mitigating factors were present, under the unconstitutional jury instruction they could have found no additional mitigating circumstances. Id. Thus, instead of those eleven jurors weighing eight mitigating circumstances against the two aggravating factors, they were only permitted to weigh the three mitigating circumstances on which they were unanimous. Recognizing that all aggravating and mitigating factors do not warrant the same weight, and that jury deliberations are an inexact science at best, this means that a possible fifty-five additional votes (eleven jurors times five factors) could have been cast in favor of mitigation. Any one of these fifty-five possible votes could, in turn, have formed the basis for a decision against the imposition of the death penalty. When the substantial evidence presented is evaluated in light of the broad discretion conferred on jurors in capital sentencing proceedings, we find our-selves unable to say, as we must to uphold a sentence of death, that none of the jurors would have been persuaded to vote for life imprisonment instead.7 After all, it only takes one hold-out juror to prevent the imposition of the death penalty, and in this case, we cannot say with [348]*348“fair assurance” that no juror would have been swayed by the mitigating factors that the jurors were unlawfully precluded from individually considering, including the highly discretionary catch-all factor, particularly when combined with the unanimously found mitigating factors.
Judges Luttig and Williams, of course, reach a contrary conclusion. In the view of Judge Williams, the effect of the mitigating factors unlawfully kept from individual consideration by the jurors is so minimal that she can conclude, with “fair assurance,” that the judgment was not substantially swayed by the error. Post at 354. We respectfully disagree with Judge Williams’ conclusion. Her approach does not appear to adequately acknowledge the possible cumulative impact of the additional mitigating factors. Even if a juror might not have found that each of those additional factors independently outweighed the aggravators, a reasonable juror well could have concluded to the contrary when considering those factors collectively, and in addition to the three unanimously found mitigators. A reasonable juror could have determined, when the additional evidence as to Allen’s difficult childhood, his ongoing emotional support of his son, and his employment are considered along with Allen’s genuine remorse for the murder, his achievement while incarcerated, and his lack of history of crime involving deadly weapons, that those mitigating factors out-weighed the two aggravating circumstances. Cf. Caldwell, 472 U.S. at 330, 105 S.Ct. 2633 (“Whatever intangibles a jury might consider in its sentencing deliberation, few can be gleaned from an appellate record. This inability to confront and examine the individuality of the defendant would be particularly devastating to any argument for consideration of what this Court has termed ‘[those] compassionate or mitigating factors stemming from the diverse frailties of humankind.’ ”) (citation omitted).
A reviewing court’s proper role in determining an error’s harmlessness entails an individualized inquiry of the sort in which we have engaged here. The inquiry before us is “not [whether these laymen] were right in their judgment, regardless of the error or its effect upon the verdict” but “rather what effect the error had or reasonably may be taken to have had upon the jury’s decision.” Kotteakos, 328 U.S. at 764, 66 S.Ct. 1239. Long ago the Supreme Court cautioned that in assessing whether an error is harmless, an appellate judge is to determine jurors’ “reactions not by [the judge’s] own, but with allowance for how others might react and not be regarded generally as acting without reason.” Kotteakos, 328 U.S. at 764, 66 S.Ct. 1239 (emphasis added). Thus, the Court directed: “The crucial thing is the impact of the thing done wrong, on the minds of other men, not on one’s own, in the total setting.” Id.
Our assessment of the harmlessness of an error is particularly significant in this case because of the character of the proceedings at issue — a sentencing hearing held to determine whether a man lives or dies. In reviewing for harmless error, the Supreme Court has told us that a court is to consider “the character of the proceeding, what is at stake upon its outcome, and the relation of the error asserted to casting the balance for decision on the case as a whole.” Kotteakos, 328 U.S. at 762, 66 S.Ct. 1239. This does not mean that we apply a different harmless error standard in capital cases, but we must apply the standard with the utmost care, see Caldwell, 472 U.S. at 329, 105 S.Ct. 2633 (“[U]nder the Eighth Amendment the qualitative difference of death from all other punishments requires a correspondingly greater degree of scrutiny of the capital [349]*349sentencing determination.”) (internal quotation marks and citation omitted), and with due acknowledgment of the fact that the jury retains great discretion in determining whether to impose the death penalty on a capital defendant, see Twilaepa v. California, 512 U.S. 967, 979-80, 114 S.Ct. 2630, 129 L.Ed.2d 750 (1994) (stating that “the sentencer may be given unbridled discretion in determining whether the death penalty should be imposed after it has found that the defendant is a member of the class made eligible for that penalty”) (internal quotation marks and citations omitted). That the error here denied jurors the right to consider a number of mitigating factors in reaching that important, and largely discretionary, decision is particularly problematic because it deprives us of any “fair assurance” that all jurors would have nonetheless voted for death.
Having conscientiously applied the Supreme Court’s directives, we cannot say with fair assurance that the totality of the mitigation evidence weighed against the two aggravating factors would not have swayed at least one juror to spare Allen’s life. Thus, we hold that the McKoy error had a “substantial and injurious effect or influence in determining the jury’s verdict” warranting habeas relief under Brecht. At a minimum, the likelihood that a properly-instructed jury would not have voted to unanimously impose the death penalty is sufficiently great to raise “grave doubt” as to whether the McKoy error was harmless. See O’Neal v. McAninch, 513 U.S. 432, 435, 115 S.Ct. 992, 130 L.Ed.2d 947 (1995) (holding that when “grave doubt” exists in a habeas case, “[w]e conclude that the uncertain judge should treat the error, not as if it were harmless, but as if it affected the verdict (i.e., as if it had a ‘substantial and injurious effect or influence in determining the jury’s verdict’)”). Because the Brecht standard does not require even a reasonable probability that the error in question affected the outcome of the proceeding, see Kyles, 514 U.S. at 436, 115 S.Ct. 1555, we find a “substantial and injurious effect” as well. Therefore, we join the judgment on the McKoy issue and hold that Allen is entitled to habeas relief under Brecht,8