404 F.3d 1006
6th Cir.
2005

Lead Opinion

COLE, J., delivered the opinion of the court, in which CLAY, J., joined. SILER, J. (pp. 1018-19), delivered a separate dissenting opinion.

OPINION

COLE, Circuit Judge.

Petitioner-Appellee Chad Ruimveld was charged with poisoning a prison guard while he was an inmate at a Michigan prison. During his trial, which took place at a special courtroom inside the prison, Ruimveld was shackled in view of the jury, despite the fact that he did not pose any special risk of flight or violence. In affirming Ruimveld’s conviction, the Michigan state courts found that this shackling was improper, but that it was harmless error. Ruimveld then filed a petition for a writ of habeas corpus in federal district court. The district court granted his petition, on the ground that the shackling was highly prejudicial to his trial. The State now appeals the grant of the writ. Because the presumption of innocence is a central tenet of the American justice system, because Supreme Court precedent clearly holds that shackling a defendant in view of the jury causes significant harm to this presumption, and because the Michigan courts unreasonably concluded that this harm did not substantially influence the jury’s decision in Ruimveld’s case, we hereby AFFIRM the district court’s decision to grant a writ of habeas corpus.

I.

On December 23, 1998, Deanne Snyder, a corrections officer at the Baraga Maximum Security Prison, drank from a cup of coffee and immediately became ill. She had poured the cup approximately five minutes earlier, but then had left it sitting unattended in a staff break room at the prison while shuttling some prisoners to the showers. Upon later forensic investigation, it was found that a cleaning agent used in the prison had been added to Snyder’s coffee.

At the time of Snyder’s poisoning, two inmates were out of their cells performing cleaning duties with the cleaning agent at issue: Billy Sirrene and Petitioner-Appel-lee Chad Ruimveld. After another guard noted that Ruimveld had been given permission to enter the break room to get some gloves for use while cleaning, Ruim-veld was charged with poisoning, under Mich. Comp. Laws § 750.436(1).

Ruimveld’s trial was held in a special courtroom inside the prison that was regularly used to avoid transporting prisoners to an outside court. The evidence against Ruimveld was entirely circumstantial; no one actually saw him (or anyone else) pour the cleaning agent into the coffee. Two corrections officers testified that they saw Ruimveld enter the break room, but neither could be certain as to whether this was before or after Snyder poured her coffee. Further, neither officer could say with any certainty that Sirrene had not entered the break room. Sirrene testified that he had not placed the cleaning agent *1009in Snyder’s cup. Two other inmates, called as witnesses by Ruimveld, testified to Sirrene’s reputation for lying. Ruim-veld also testified, denying having placed the cleaning agent in the coffee. Ruimveld also testified that he had never interacted with Snyder nor had he entered the break room near the time at issue.

For the entirety of the trial, Ruimveld was brought into the courtroom in leg shackles, belly chains, and handcuffs. Pri- or to trial, Ruimveld’s counsel moved to remove the shackles,1 but the judge denied both this motion and a subsequent motion to cover the shackles so that they could not be seen by the jury. As a result, not only was Ruimveld shackled as he entered the courtroom, but was shackled at all times in front of the jury, even while testifying. See J.A. at 310 (“Sir, to the extent you can, would you raise your right hand for me?” (emphasis added)). Jury deliberations took more than three hours, and included a requested instruction from the judge regarding the “presumption of innocence, [the] burden of proof, and reasonable doubt.” The jury returned a verdict of guilty.

Ruimveld subsequently moved for a new trial on the basis of newly discovered evidence, because another inmate later swore that he had seen Sirrene enter the break room with a bottle of the cleaning agent immediately after Snyder had left the break room to escort other prisoners to the shower. Ruimveld’s motion was denied.

Ruimveld then appealed on various grounds. The Michigan Court of Appeals agreed unanimously that the use of the shackles was improper, and that “the trial court abused its discretion in ordering defendant to remain shackled during trial.” People v. Ruimveld, No. 227793, 2001 WL 793872, at *3 (Mich.Ct.App. July 13, 2001). However, two of the three judges agreed that this error was harmless. The state appellate court thus affirmed his conviction and sentence. Id. But see id., at *4-7 (Whitbeck, J., dissenting, on the grounds that this error was not harmless). The Michigan Supreme Court summarily declined to review the case. People v. Ruimveld, 465 Mich. 955, 639 N.W.2d 812 (Mich.2002).

Ruimveld then filed a pro se petition for a writ of habeas corpus in federal district court, alleging that the shackling violated his constitutional right to a fair trial. Ru-imveld also alleged that his counsel was ineffective, that the evidence against him was insufficient, and that the trial’s venue inside the prison was improper. The district court granted the writ, solely on the grounds that, because of his shackling, Ru-imveld “did not receive a fair trial in which his constitutional right to the presumption of innocence was upheld.” The State, on behalf of Respondent-Appellant Warden Thomas Birkett, timely appealed the district court’s grant of the writ.

II.

A. Standard of Review

Because Ruimveld’s petition was filed under 28 U.S.C. § 2254, both we and the district court are bound by the standard of review set forth in the Antiterrorism and Effective Death Penalty Act of 1996 (“AEDPA”), Pub.L. No. 104-132, § 104(d), 110 Stat. 1214, 1219 (codified at 28 U.S.C. § 2254(d)). This section limits grants of writs of habeas corpus for prisoners tried in state court to the case in which the state adjudication procedures “resulted in a decision that was contrary to, or an unreasonable application of, clearly established Federal law, as determined *1010by the Supreme Court of the United States .... ” Id. In making a determination as to what constitutes “clearly established Federal law,” lower courts are restricted to reviewing the law as established by the Supreme Court. See, e.g., Williams v. Taylor, 529 U.S. 362, 412, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000). Importantly, such law can only be determined through an examination of the holdings of the Court, as opposed to its dicta. Id.; see also Yarborough v. Alvarado, 541 U.S. 652, 124 S.Ct. 2140, 2147, 158 L.Ed.2d 938 (2004) (noting that the “holdings” of the Supreme Court that must be considered are the “governing legal principle or principles set forth by the Supreme Court as of the time of the relevant state-court decision” (citation and quotation omitted)). However, “the lack of an explicit statement” of a particular rule by the Supreme Court “is not determinative” of clearly established law, since “[t]he Court has made clear that its relevant precedents include not only bright-line rules but also the legal principles and standards flowing from precedent.” Taylor v. Withrow, 288 F.3d 846, 852 (6th Cir.2002).

A state court decision is “contrary to” clearly established Federal law “if the state court arrives at a conclusion opposite to that reached by [the Supreme Court] on a question of law,” or “if the state court confronts facts that are materially indistinguishable from a relevant Supreme Court precedent” and arrives at a different result. Williams, 529 U.S. at 405, 120 S.Ct. 1495. A state court decision is an “unreasonable application of’ clearly established Federal law “if the state court correctly identifies the correct governing legal rule from [the Supreme Court’s] cases but unreasonably applies it to the facts of the particular state prisoner’s case.” Id. at 407, 120 S.Ct. 1495. An “unreasonable application” can also occur where “the state court either unreasonably extends a legal principle from [the Supreme Court’s] precedent to a new context where it should not apply or unreasonably refuses to extend that principle to a new context where it should apply.” Id. It should also be noted that the “unreasonable” requirement is a high bar; “a federal habeas court may not issue a writ under the unreasonable application clause simply because that court concludes in its independent judgment that the relevant state-court decision applied clearly established Federal law erroneously or incorrectly.” Bell v. Cone, 535 U.S. 685, 694, 122 S.Ct. 1843, 152 L.Ed.2d 914 (2002) (citation and quotation marks omitted).

Finally, we review the decision of a district court to grant or deny a writ of habeas corpus de novo, but review factual findings by that court for clear error, except where the district court has made factual determinations based on its review of trial transcripts and other court records; in such eases we review such findings de novo. See, e.g., Wolfe v. Brigano, 232 F.3d 499, 501 (6th Cir.2000).

B. Ruimveld’s Shackles

1. Review of Supreme Court Law

Warden Birkett first argues that the district court’s decision to grant the writ of habeas corpus was improper because “[although the [Supreme] Court has commented on the constitutionality of restraining criminal defendants during trial, there is no Supreme Court case directly addressing a case in which a criminal defendant was tried while wearing visible leg irons.” A brief review of Supreme Court cases related to prejudicial trappings at trial is thus warranted.

In Illinois v. Allen, 397 U.S. 337, 90 S.Ct. 1057, 25 L.Ed.2d 353 (1970), the Court held that where a prisoner was intentionally disruptive at trial, and where he was informed that he could remain in the courtroom as long as he ceased his disruptions, his constitutional right to be *1011present throughout his trial was waived when he chose to continue disrupting the proceedings. In so holding, Justice Black noted for the Court that:

[t]rying a defendant for a crime while he sits bound and gagged before the judge and jury would to an extent comply with that part of the Sixth Amendment’s purposes that accords the defendant an opportunity to confront the witnesses at trial. But even to contemplate such a technique, much less see it, arouses a feeling that no defendant should be tried while shackled and gagged except as a last resort. Not only is it possible that the sight of shackles and gags might have a significant effect on the jury’s feelings about the defendant, but the use of this technique is itself something of an affront to the very dignity and decorum of judicial proceedings that the judge is seeking to uphold.

Id. at 344, 90 S.Ct. 1057. Nonetheless, the Court held that in some situations, binding and gagging a defendant might be appropriate, due to safety or security concerns for all parties involved in the trial. Id.

In Estelle v. Williams, 425 U.S. 501, 96 S.Ct. 1691, 48 L.Ed.2d 126 (1976), the Court noted that when a defendant who was not incarcerated on some other charge was forced to appear before a jury in prison clothing, the “clothing is so likely to be a continuing influence throughout the trial that ... an unacceptable risk is presented of impermissible factors coming into play.” Id. at 504, 96 S.Ct. 1691. However, because the defendant had failed to object at trial to being forced to wear prison garb, the Court held that “although the State cannot, consistently with the Fourteenth Amendment, compel an accused to stand trial before a jury while dressed in identifiable prison clothes, the failure to make an objection to the court ... is sufficient to negate the presence of a compulsion necessary to establish a constitutional violation.” Id. at 512-13, 96 S.Ct. 1691. Again, the Court implied how prejudicial it believed the defendant’s appearance to be, but its holding cannot fairly be said to involve specifically the binding of a defendant. But see McKaskle v. Wiggins, 465 U.S. 168, 178, 104 S.Ct. 944, 79 L.Ed.2d 122 (1984) (describing Estelle as recognizing “that [defendants] may not normally be forced to appear in court in shackles or prison garb”).

Finally, in Holbrook v. Flynn, 475 U.S. 560, 106 S.Ct. 1340, 89 L.Ed.2d 525 (1986), the Court considered “whether the conspicuous, or at least noticeable, deployment of security personnel in a courtroom during trial is the sort of inherently prejudicial practice that, like shackling, should be permitted only where justified by an essential state interest specific to each trial.” Id. at 568-69, 106 S.Ct. 1340. While considering the stationing of numerous security officers near the front of a courtroom during a criminal trial, the Court held, in a pre-AEDPA context, that:

[a]ll a federal court may do in such a situation is look at the scene presented to jurors and determine whether what they saw was so inherently prejudicial as to pose an unacceptable threat to defendant’s right to a fair trial; if the challenged practice is not found inherently prejudicial and if the defendant fails to show actual prejudice, the inquiry is over.”

Id. at 572, 106 S.Ct. 1340. The Court thus affirmed the conviction, finding that the defendant had not shown either that the row of security officers was prejudicial or that his case was actually prejudiced thereby.

These are the only three Supreme Court cases that discuss directly how shackling or other “indicia of guilt” at trial could potentially prejudice a defendant’s case.2 *1012The Michigan state appellate court clearly-found that “[fjreedom from shackling is an important component of a fair trial,” and that “there is no record evidence that defendant presented any of the risks that shackling is intended to prevent.” Ruimveld, 2001 WL 793872, at *3. Accordingly, it concluded that “the trial court abused its discretion by ordering [Ruimveld] to remain shackled during trial.” Id. However, it affirmed Ruimveld’s conviction on the grounds that “any error in the present case regarding defendant’s shackling was ... harmless” because “[t]he jury was aware ... that defendant was not only the accused in a criminal trial, but was currently incarcerated ... on an unrelated charge. Id. While this fact did not justify defendant’s shackling,” the court held that “precisely because of this fact, the jury did not place any unfair importance on defendant’s being shackled during trial.” Id.

Before our Court, the State does not dispute that Ruimveld’s shackling was error, but instead argues, as did the majority on the state appellate court, that any such error was harmless. The parties do agree that Ruimveld had neither a criminal history nor a prison disciplinary record that would warrant any additional or unusual protections in the courtroom during trial. As a result, we need not determine whether Ruimveld’s shackling constituted constitutional error, since the state court clearly found that it did. Ruimveld, 2001 WL 793872, at *3 (applying Michigan state cases which state conclusions effectively equivalent to those in Allen, regarding when shackling is permitted); cf. Kennedy v. Cardwell, 487 F.2d 101, 110-11 (6th Cir.1973) (detailing factors to consider when determining whether or not shackling constitutes constitutional error); United States v. Waagner, 104 Fed.Appx. 521, 526-27 (6th Cir.2004) (unpublished opinion) (applying Kennedy factors in recent case).

Rather, this Court effectively must answer two questions, based on its review of applicable Supreme Court precedent. First, we must determine whether constitutionally erroneous shackling is so inherently prejudicial that it could never be found to be harmless. If so, then the state court’s conclusion that Ruimveld’s shackling constituted harmless error was contrary to established Supreme Court law, and the district court’s grant of the writ should be upheld. However, if shackling can sometimes constitute harmless error, we must then determine whether the state court’s determination that Ruimveld’s shackling constituted harmless error was unreasonable. If clearly established Federal law as enunciated by the holdings of the Supreme Court would have us answer “yes” to either of these questions, this Court must affirm the district court’s grant of the writ.

2. Is Shackling Always Prejudicial?

Ruimveld first argues, based on dicta from Allen, that an erroneously shackled defendant always suffers prejudice so extreme that it could never be found to be harmless. As the State notes, the majority of constitutional errors at trial are subject to harmless error analysis. See, e.g., Penry v. Johnson, 532 U.S. 782, 795, 121 S.Ct. 1910, 150 L.Ed.2d 9 (2001); Brecht v. Abrahamson, 507 U.S. 619, 638, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993); see also Fed.R.Crim.P. 52(a). However, it is *1013also clear that some constitutional violations are so inherently prejudicial that they can never be deemed harmless. See, e.g., Kimmelman v. Morrison, 477 U.S. 365, 395 n. 2, 106 S.Ct. 2574, 91 L.Ed.2d 305 (1986) (stating that the complete denial of counsel at trial could never result in error that could be considered harmless); Tumey v. Ohio, 273 U.S. 510, 47 S.Ct. 437, 71 L.Ed. 749 (1927) (holding same regarding a biased adjudicator). In these cases, a reviewing court need not apply the harmless error test because any such error could never be found to be harmless. Id.

Here, it cannot be said that the Supreme Court has held squarely that shackling is a practice so prejudicial as to preclude all harmless error review. While the Court did state in dicta that “no defendant should be tried while shackled and gagged except as a last resort,” Allen, 397 U.S. at 344, 90 S.Ct. 1057, and that forcing a prisoner to wear prison garb is “the sort of inherently prejudicial practice that, like shackling, should be permitted only where justified by an essential state interest specific to each trial,” Holbrook, 475 U.S. at 568-69, 106 S.Ct. 1340, the question of whether shackling could result in harmless error was not before the Court in those cases, each of which was solely concerned with whether constitutional error had occurred at all. Nor do any of the Court’s other shackling-related cases discuss whether the error of shackling a defendant can be harmless in light of other factors such as overwhelming evidence. As a result, there does not appear to be a clearly announced rule, in a Supreme Court holding, that harmless error analysis should not be applied to shackling cases.

Further, there is Supreme Court precedent holding that harmless error analysis should apply in cases where the courtroom atmosphere hints at a defendant’s dangerousness or guilt. In Holbrook, the Court described the analysis that it would require federal courts to undertake to determine if reversal was required due to prejudicial trappings at trial (in that case, when numerous security guards were stationed at the front of the courtroom):

All a federal court may do in such a situation is look at the scene presented to jurors and determine whether what they saw was so inherently prejudicial as to pose an unacceptable threat to defendant’s right to a fair trial; if the challenged practice is not found inherently prejudicial and if the defendant fails to show actual prejudice, the inquiry is over.”

Id. at 572, 106 S.Ct. 1340. The Court thus made clear that a particular trial practice ought to be examined as to whether it prejudiced the defendant’s case. This is in line with the majority of other constitutional errors considered by the Supreme Court. See, e.g., Brecht, 507 U.S. at 638, 113 S.Ct. 1710. In addition, while Hol-brook was decided before AEDPA, there is no reason to expect that the Supreme Court would require any different analysis by a state appellate court; neither any relevant provision of the Constitution nor any relevant trial-related constitutional doctrine has changed since then. As a result, and despite the apparently contrary statements in dicta in both Holbrook and Allen that shackles are inherently prejudicial, this Court cannot fairly say that the Michigan appellate court’s decision to employ harmless error analysis in this case constituted a decision that was either contrary to clearly established Federal law or an unreasonable application thereof. Indeed, our Court has previously utilized harmless error analysis in a case similar to Ruimveld’s. See Bell v. Hurley, 97 Fed.Appx. 11, 16-17 (6th Cir.2004) (unpublished opinion) (utilizing harmless error analysis in concluding both that no hearing on a defendant’s shackling was required due to his obvious dangerousness and that even if the shackling had constituted error, *1014no prejudice resulted to the defendant from being shackled during trial because of the overwhelming evidence of guilt against him).

3. Was the Michigan Court’s Harmless Error Analysis Unreasonable?

That said, the state court’s application of the harmless error test to this case could still be unreasonable. “[Ejrrors that undermine confidence in the fundamental fairness of the state adjudication certainly justify the issuance of the federal writ.” Williams, 529 U.S. at 375, 120 S.Ct. 1495 (citing Teague v. Lane, 489 U.S. 288, 311-14, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989), among others). However, the Supreme Court has clearly stated that habeas corpus is only warranted where errors “had substantial and injurious effect or influence in determining the jury’s verdict.” Brecht, at 637-38, 113 S.Ct. 1710 (citing Kotteakos v. United States, 328 U.S. 750, 776, 66 S.Ct. 1239, 90 L.Ed. 1557 (1946)). While the burden of persuasion on harmless error review falls on a petitioner, he need not prove that the error was outcome-determinative; instead he must merely remove any assurances that the error did not affect the outcome. See Beck v. Haik, 377 F.3d 624, 635 (6th Cir.2004); see also Brecht, 507 U.S. at 637, 113 S.Ct. 1710. Accordingly, we must review the state appellate court’s decision to determine if it properly considered whether Ruimveld successfully countered any assurances that the jury’s verdict was not harmed by his shackling.

There can be little question that the holdings of the Supreme Court reflect a strong concern that trial practices such as shackling can have substantial or injurious influences on jury verdicts. The holdings in Estelle and Holbrook, especially, effectively state that indicia of guilt, such as shackles, impose a significant level of harm upon the presumption of innocence, and that the injurious influence of such harm can only be overcome by a confident finding that the indicia were warranted or that the outcome of the defendant’s case was not affected. Estelle, 425 U.S. at 512-13, 96 S.Ct. 1691; Holbrook, 475 U.S. at 568-69, 572, 106 S.Ct. 1340; see also McKaskle, 465 U.S. at 178, 104 S.Ct. 944. Cf. Taylor, 288 F.3d at 853 (noting that for AEDPA purposes, “clearly established Federal law” includes “legal principles and standards flowing from precedent”).3 A wide array of lower courts have also reached the conclusion that the Supreme Court requires a strong showing that the prejudicial effects of indicia of guilt imposed upon a defendant were outweighed by other factors rendering the imposition of such indicia harmless. See, e.g., Kennedy v. Cardwell, 487 F.2d 101, 105 (6th Cir.1973) (citing many such cases);4 see also Szu-*1015chon v. Lehman, 273 F.3d 299, 314-15 (3d Cir.2001); United States v. Zuber, 118 F.3d 101, 103-04 (2d Cir.1997); United States v. Childress, 58 F.3d 693, 704 (D.C.Cir.1995) (“[CJriminal defendants do have a right to be free of court-imposed physical appearances that are unfairly suggestive of their guilt”). As the Ninth and Fourth Circuits have noted:

Basic to American jurisprudence is the principle that an accused, despite his previous record or the nature of the pending charges, is presumed innocent until his guilt is established beyond a reasonable doubt by competent evidence. It follows that he is also entitled to the indicia of innocence. In the presence of the jury, he is ordinarily entitled to be relieved of handcuffs, or other unusual restraints, so as not to mark him as an obviously bad man or to suggest that the fact of his guilt is a foregone conclusion.

Hamilton v. Vasquez, 882 F.2d 1469, 1471 (9th Cir.1989) (quoting United States v. Samuel, 431 F.2d 610, 614-15 (4th Cir.1970)). Further, the Supreme Court has consistently held that a defendant’s right to be presumed innocent until proven guilty is central to our system of criminal justice, and that actions that impinge upon this presumption are to be taken only when absolutely necessary. See, e.g., Estelle, 425 U.S. at 505, 96 S.Ct. 1691; see also Coffin v. United States, 156 U.S. 432, 453, 15 S.Ct. 394, 39 L.Ed. 481 (1895) (“The principle that there is a presumption of innocence in favor of the accused is the undoubted law, axiomatic and elementary, and its enforcement lies at the foundation of the administration of our criminal law.”). As a result, the Court has clearly required close scrutiny of any sort of error that would in any way harm a defendant’s right to the presumption of innocence. See, e.g., Estelle, 425 U.S. at 505, 96 S.Ct. 1691. Thus, because the Supreme Court has clearly stated that indicia of guilt such as shackles harm a defendant’s presumption of innocence, the state appellate court’s decision here was only reasonable if that court reasonably could have determined that the harmful effects of Ruimveld’s shackling in front of the jury did not have a substantial and injurious effect or influence in determining the jury’s verdict.

The state appellate court determined that “the jury did not place any unfair importance on defendant’s being shackled at trial,” because it “was aware from the facts of the case that defendant was not only the accused in a criminal trial, but was currently incarcerated in the prison on an unrelated charge.” Ruimveld, 2001 WL 793872, at *3. The state court engaged in no further analysis of this issue. As Ruimveld notes, the state court did not find there to be “overwhelming evidence” against him, nor did it find that the shackling was in any way warranted to ensure a fair, undisrupted trial. It did not examine whether other prisoner witnesses (and specifically prisoner Sirrene, whom Ruimveld claimed had actually committed the crime) appeared before the jury in shackles. It did not examine whether the jury would believe, due to the shackles, that Ruimveld was a violent and dangerous man who would be more likely to commit the crime at issue than would other inmates. In fact, it seems the court did not consider any of the other possible reasons why a jury might be influenced by viewing *1016a defendant in shackles, beyond a mere consideration of the fact that shackles might lead a jury to believe that the defendant is a prisoner or convict.

Indeed, the state court’s reliance on the fact that Ruimveld was a prisoner is also misplaced; the fact that Ruimveld still had to be shackled despite the fact that he was already securely within the walls of a maximum security prison and guarded by numerous armed guards might have further prejudiced his case in the eyes of jurors who might have believed him to be a particularly dangerous or violent person, even among inmates. Cf. Hamilton, 882 F.2d at 1471. Moreover, as the district court noted,

it is entirely possible that the indicia of innocence of the charge being tried may be more important for [a] defendant who is incarcerated on a prior charge than for a defendant who is not, because the defendant already has one “strike” against him or her by the jury knowing that he or she is a person previously convicted of committing a crime and currently imprisoned for it.

Ruimveld v. Birkett, No. 03-CV-60030-AA, slip op. at 33 (E.D.Mich.) (Opinion and Order of June 2, 2004) (emphasis in original). However, even disregarding these two possible arguments, it is unlikely that all possible prejudice to Ruimveld’s case from his shackling instantly evaporated as soon as the jurors found out he was a prisoner. Yet this is the only prejudice-mitigating factor that has ever been discussed by either the state appellate court or Warden Birkett.

Further, the Supreme Court requires the prejudicial effect of constitutional errors on habeas review to be evaluated in light of the totality of the evidence. See, e.g., Williams, 529 U.S. at 398-99, 120 S.Ct. 1495. “A verdict or conclusion only weakly supported by the record is more likely to have been affected by errors than one with overwhelming record support.” Strickland v. Washington, 466 U.S. 668, 696, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). Though the state court did not even hint at such an analysis, Ruimveld’s case was not an open-and-shut case. The evidence against him was merely circumstantial; no one saw him place the cleaning agent in the coffee, and though two witnesses testified that they saw Ruimveld enter the break room during the day of the poisoning, neither could place him there with certainty at the time of the poisoning. As Ruimveld notes, Sirrene was also in the area, handling the same cleaning agent, and though Sirrene denied having committed the crime, several witnesses testified as to Sirrene’s poor credibility. As further evidence of the closeness of the case, Ru-imveld notes that the jury deliberated for over three hours despite the simple facts, and made inquiries to the judge regarding presumptions of innocence, burdens of proof, and reasonable doubt. Given the closeness of the case, the effect of any error was thus likely to be magnified. See, e.g., Strickland, 466 U.S. at 696, 104 S.Ct. 2052; see also Bell, 97 Fed.Appx. at 16-17 (citing “overwhelming evidence” and the defendant’s failure to present any reasonable defense as mitigating any possible prejudice from defendant’s clearly justified shackling); Wilkerson v. Whitley, 16 F.3d 64, 69 (5th Cir.1994) (finding that clear eyewitness testimony rendered harmless any prejudice resulting from defendant’s shackling).

In addition, as Ruimveld notes, neither Warden Birkett nor the state appellate court have discussed any significant evidence that would counter Ruimveld’s argument that his shackling weighed against him or that would prove that this substantial injury to Ruimveld’s case was mitigated to the point where any remaining injury would be insubstantial. The only mitigating factor discussed by the state appellate *1017court was the fact that the trial took place in a prison; hardly a factor that would be likely to nullify completely the prejudice resulting from Ruimveld’s shackling. See supra, at 1016; cf. Bell, 97 Fed.Appx. at 17. Further, the state appellate court did not take into account the totality of the evidence or the closeness of the case in determining that the error was harmless, as would be required by Williams.5

Similarly, the State’s brief on appeal does not detail how Ruimveld’s shackling was harmless, beyond simple reliance on the statements of the state appellate court. The brief repeatedly notes that the state appellate court did not find that the error was harmless because of overwhelming evidence or other circumstances of the case, but rather that the state court found that the shackling was harmless only because it “was not prejudicial.” Br. of Appellant at 20-21. Even if one reads this statement in some non-tautological fashion, the conclusion is directly at odds with the Supreme Court precedent discussed supra, logically implying that indicia of guilt do indeed affect a defendant’s presumption of innocence where the shackling is not warranted or outweighed by other factors. A mere statement by the state appellate court claiming the shackling was not prejudicial does not automatically make it so, absent facts that would support this conclusion. And nowhere in the State’s brief does it mention any other factors, beyond Ruimveld’s imprisonment, that in this case would have mitigated the prejudicial' effects of shackling; rather the brief merely repeatedly relies on the mantra that the state appellate court found the shackling not to have prejudiced Ruimveld’s case.

As a result, we conclude that the state court’s cursory harmless error analysis unreasonably discounted the prejudicial effects of shackling noted by the Supreme Court in Allen and Estelle. This is especially true given that there can be little question that a defendant’s shackling in front of the jury always interferes with the presumption of innocence to some extent and can create a substantially injurious inference of guilt in the absence of any circumstances that render shackles necessary. See Allen, 397 U.S. at 344, 90 S.Ct. 1057; Estelle, 425 U.S. at 512-13, 96 S.Ct. 1691; Holbrook, 475 U.S. at 568-69, 106 S.Ct. 1340; see also Ruimveld, 2001 WL 793872, at *3. (Whitbeck, J., dissenting) (“The trial court’s ruling did hot, in my view, just diminish the presumption of innocence, it obliterated it.”). It is true that, on harmless error review, a petitioner bears the burden of “depriv[ing] the appellate court of a ‘fair assurance’ that the error was not outcome determinative.” Beck v. Haik, 377 F.3d 624, 635 (6th Cir.2004); see also Brecht, 507 U.S. at 637, 113 S.Ct. 1710. However, Ruimveld has done so by showing the harm to the presumption of innocence that the Supreme Court has found to be inherent in indicia of guilt such as shackles, by showing that there was no good reason for the shackling, by showing that his was a close case based on *1018purely circumstantial evidence, and by pointing out that the State has not alluded to any evidence that would in any way remove this case from the Supreme Court’s clear conclusion that indicia of guilt can result in significant prejudice when such indicia are not warranted by the security concerns of the trial court. Indeed, it would be difficult to imagine anything else an erroneously shackled defendant could ever show in order to evince a lack of assurance that the error did not affect his case, absent a poll of the jury where jurors stated that the shackling affected their decisionmaking.

Accordingly, the state court’s failure to recognize the likelihood of substantially injurious effects resulting from Ruimveld’s unnecessary shackling was, as the district court noted, an unreasonable application of the harmless error standard clearly stated by the Supreme Court in Brecht. Therefore, the district court correctly found that Ruimveld was entitled to a conditional writ of habeas corpus.

Because we affirm the district court’s decision, we need not address the other grounds on which Ruimveld has petitioned for the writ.

III.

For these reasons, we AFFIRM the district court’s grant to Ruimveld of a conditional writ of habeas corpus.

Notes

. "Shackles,” throughout the rest of this opinion, includes all of the bindings worn by Ru-imveld, including leg shackles, belly chains, and handcuffs.

. Of course, some other cases discuss the potential for prejudice resulting from the at*1012mosphere in which a trial is conducted. See, e.g., Murphy v. Florida, 421 U.S. 794, 802, 95 S.Ct. 2031, 44 L.Ed.2d 589 (1975); cf. Coy v. Iowa, 487 U.S. 1012, 1020-21, 108 S.Ct. 2798, 101 L.Ed.2d 857 (1988) (considering potential for situations when prejudicial effects of violations of Confrontation Clause rights might be vitiated by other important public policy concerns). However, none of these cases are directly relevant to the case at bar.

. Indeed, even under a stricter interpretative standard, we would still reach the same conclusion, since the holdings in these cases clearly logically result from a concern that shackling harms the presumption of innocence. See, e.g., Taylor, 288 F.3d at 855 (Boggs, J., concurring in the judgment) (arguing that "clearly established Federal law” should only include results that logically follow from a Supreme Court decision).

. Ruimveld would have us find that Smith v. Stegall, 385 F.3d 993 (6th Cir.2004), which held that Sixth Circuit precedent as to what is "clearly established Federal law” under AEDPA is binding upon later panels of this Court, binds us to find that shackling is always inherently prejudicial when not warranted by the defendant’s conduct or record, because of the precedential holding of Kennedy. However, Kennedy, a pre-AEDPA case, did not solely rest its holding upon the holdings of the Supreme Court, and, regardless, did not consider whether or not a shackling-related error could ever be harmless if not found to be warranted due to the defendant's circumstances. Indeed, the Kennedy Court found that Kennedy’s circumstances did warrant shackling, and thus found no constitutional error without ever reaching the harmless error issue. 487 F.2d at 110-12. Accordingly, and without deter*1015mining whether a pre-AEDPA case could ever create Stegall-style precedent for our interpretation of "clearly established Federal law,” we disagree with Ruimveld’s contention that Kennedy and Stegall together require a finding that unwarranted shackling is always so prejudicial that reversal is required. As the State argues, to the extent that the district court relied solely on statements from Kennedy for its grant of the writ, it also was in error.

. The dissent argues that the state court “did not need’’ to make certain findings of fact, because there was "ample evidence” to support the conviction. We note that the dissent's claim that there is "ample evidence” to support the conviction is irrelevant since the question before us is not whether there was sufficient evidence to support the jury's conclusions, but rather whether the evidence is so strong that a reviewing court can be assured that the shackling did not affect the jury’s conclusions. See, e.g., Beck, 377 F.3d at 634-35. More importantly, however, and regardless of how "ample” the evidence was, the harmless error analysis as required by Brecht requires an examination of the totality of the evidence, among other factors. See, e.g., id.; Williams, 529 U.S. at 397-99, 120 S.Ct. 1495; Strickland, 466 U.S. at 696, 104 S.Ct. 2052. The state appellate court’s harmless error analysis evinces no such examination.



Dissenting Opinion

SILER, Circuit Judge,

dissenting.

I respectfully dissent, for I believe that the Michigan Court of Appeals did not unreasonably apply clearly established federal law, as determined by the Supreme Court of the United States, when it affirmed Ruimveld’s conviction. See 28 U.S.C. § 2254(d).

I am in basic agreement with the majority in finding that if there was a constitutional error in this case, it is subject to a harmless error test. However, I differ on the question of whether there is clearly established federal law determined by the Supreme Court that trying a defendant in shackles is unconstitutional. Some dicta suggest such a result, as the majority relates, but I do not think it has been “clearly established.” See Williams v. Taylor, 529 U.S. 362, 412, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000) (clearly established law “refers to the holdings, as opposed to the dicta, of this [Supreme] Court’s decisions as of the time of the relevant state-court decision.”). However, I would leave to another day the determination as to whether the law has been clearly established, for this case can be determined on the issue of whether, assuming there was a constitutional error, the resolution by the Michigan Court of Appeals in this matter was an unreasonable application of that law.

The majority correctly decides that the harmless error test by the Michigan Court of Appeals was the appropriate procedure after finding error by the trial court.1 But then, the majority finds fault with how the Michigan Court of Appeals determined the error to be harmless. The majority says the state court did not find there to be “overwhelming evidence” against Ruim-veld, but it found that there was sufficient evidence for a conviction, and the district court found “there was ample evidence” to convict Ruimveld beyond a reasonable doubt. The majority also criticizes the Michigan Court of Appeals because it did not examine whether other prisoner witnesses, and specifically Billy Sirrene, ap*1019peared before the jury in shackles. However, in the same language the trial court used in administering the oath for Ruim-veld to testify, as stated by the majority herein (“Sir, to the extent you can, would you raise your right hand for me?”), it also swore in other inmate witnesses, including Sirrene, James Ingram, Mark Anthony Collingwood, Larry Darnell Schuler and Gerald Hoskins. Therefore, the record was clear that other inmates who testified were also in shackles. The Michigan Court did not need to make a finding about that fact.

The majority also criticizes the Michigan Court for not looking at the totality of the evidence in determining harmlessness. It says that the evidence was “merely circumstantial.” Yet we know that “[c]ir-cumstantial evidence ... is intrinsically no different from testimonial evidence.” Holland v. United States, 348 U.S. 121, 140, 75 S.Ct. 127, 99 L.Ed. 150 (1954). Accord United States v. Frost, 914 F.2d 756, 762 (6th Cir.1990). Only two inmates, Ruim-veld and Sirrene, had possession of the cleaning solution that apparently was put into the coffee that Corrections Officer Snyder drank. Two witnesses saw Ruim-veld enter the break room within minutes of the time that Snyder drank the coffee. No one testified that Sirrene or any other inmate entered the break room around that time. Circumstantial? Yes, but it was “ample” to sustain the conviction.

The majority also criticizes the “cursory harmless error analysis” by the Michigan court, but Ruimveld’s shackling was discussed in two full paragraphs in the opinion. The quality of the ruling was never meant to depend upon the verbiage in the state court’s decision, or else Congress would have included such criteria in AED-PA. The majority also suggests that because the jury deliberated for three hours and asked questions of the court, this somehow shows that the shackling was prejudicial, but the Supreme Court has not yet held that such jury conduct illustrates prejudice. It means nothing. If anything, a short deliberation along with no inquiries to the court might suggest that the jury had made up its mind quickly because of the shacMes.

It was never meant for this court to be a super appellate court to the Michigan Court of Appeals. See Payne v. Janasz, 711 F.2d 1305, 1310 (6th Cir.1983). That is the job of the Michigan Supreme Court, which declined to hear the case. I would deny the writ, finding the Michigan Court of Appeals did not unreasonably apply clearly established law as determined by the United States Supreme Court.

. The majority relates that the Michigan Court of Appeals “clearly found” constitutional error, but I failed to discover where that occurs in the opinion. Instead, the court found error, cited only Michigan state cases, and concluded that the trial court abused its discretion by ordering Ruimveld to remain shackled during the trial.