Opinion by Judge REINHARDT; Dissent by Judge THOMPSON.
REINHARDT, Circuit Judge:. At a murder trial in which the central question is whether the defendant acted in self-defense, are a defendant’s constitutional rights violated when spectators are permitted to wear buttons depicting the “victim”? We conclude that under clearly established Supreme Court law such a practice interferes with the right to a fair trial by an impartial jury free from outside influences.
Mathew Musladin appeals the district court’s denial of his petition for a writ of habeas corpus. He contends that the buttons worn by the “victim’s” family members at his trial created an unreasonable risk of impermissible factors coming into play, and that the state court was objectively unreasonable in denying this claim both' on direct appeal and in the post-conviction proceedings. ■ In light of clearly-established federal law set forth by the Supreme Court, and persuasive authority from this court concerning the proper application of that law, we hold that the last-reasoned decision of the state court *1073constituted an unreasonable application of Supreme Court law. Accordingly, we reverse the district court’s denial of Musla-din’s petition and remand for issuance of the writ.
I. Factual Background and Procedural History
Musladin was charged in a California state court with first degree murder for the killing of Tom Studer, the fiance of his estranged wife Pamela. On May 13, 1994, Musladin came to the house where Pamela, Studer, and Pamela’s brother Michael Albaugh lived in order to pick up his son for a scheduled weekend visit. Pamela testified that she and Musladin had an argument, and that Musladin pushed her to the ground. According to Pamela, when Studer and Albaugh came out of the house to assist her, Musladin reached into his car to grab a gun and fired two shots at Studer, killing him. Musladin contends, however, that after Pamela fell to the ground, Studer and Albaugh appeared, holding a gun and a machete respectively, and threatened him. Musladin asserted that, after seeing the weapons, he shot in the general direction of Studer out of fear for his own life. Accordingly, at trial Mus-ladin argued perfect and imperfect self-defense. There is no dispute that Musla-din fired the shot that killed Studer, although experts for both sides agree that the fatal shot was the result of a ricochet rather than a direct hit. Under Musla-din’s theory of defense, there was no crime and, thus, no “victim.”
During the 14-day trial, Studer’s family sat in the front row of the gallery. On each of those 14 days, at least three members of the family wore buttons on their shirts with the deceased’s photograph on them. According to declarations submitted by the defendant, the buttons were several inches in diameter and “very noticeable.” Furthermore, the family members were seated in the row directly behind the prosecution and in clear view of the jury. Before opening statements, counsel for Musladin requested that the trial judge instruct the family members to refrain from wearing the buttons in court, out of fear that the button’s expressive content would influence the jury and prejudice Musladin’s defense. The trial judge denied the request. Musladin was convicted of first degree murder and three other related offenses.
Musladin exhausted the available state procedures both on direct review and on post-conviction relief. He then filed a petition for a writ of habeas corpus in the District Court for the Northern District of California. He alleged, among other things, that the state court unreasonably applied clearly-established federal law in determining that his right to a fair trial was not violated by the family members’ wearing of the buttons depicting the deceased. The district court denied the petition and this appeal followed.
II. The AEDPA Standard
Musladin’s petition for habeas corpus is governed by the Anti-Terrorism and Effective Death Penalty Act (AEDPA). Therefore, we may not grant habeas relief to the defendant unless the state court decision was “contrary to, or involved an unreasonable application of,.clearly established Federal law, as determined by the Supreme Court of the United States.” 28 U.S.C. § 2254(d)(1). Because state courts often issue “postcard” denials that offer no rationale for.their dispositions, we determine whether the state court unreasonably applied federal law by looking to the “last reasoned decision of the state court as the basis of the state court’s judgment.” Franklin v. Johnson, 290 F.3d 1223, 1233 n. 3 (9th Cir.2002). In this case, we look to the opinion ofLthg__California Court of Appeal on direct appeal.
*1074AEDPA limits the source of clearly-established federal law to Supreme Court cases. See 28 U.S.C. § 2254(d)(1). Nevertheless, we recognize that precedent from this court, or any other federal circuit court, has persuasive value in our effort to determine “whether a particular state court decision is an ‘unreasonable application’ of Supreme Court law, and ... what law is ‘clearly established.’ ” Duhaime v. Ducharme, 200 F.3d 597, 600 (9th Cir.1999); see also Robinson v. Ignacio, 360 F.3d 1044, 1057 (9th Cir.2004) (“When faced with a novel situation we may turn to our own precedent, as well as the decisions of other federal courts, in order to determine whether the state decision violates the general principles enunciated by the Supreme Court and is thus contrary to clearly established federal law.”); Williams v. Bowersox, 340 F.3d 667, 671 (8th Cir.2003) (“[T]he objective reasonableness of a state court’s application of Supreme Court precedent may be established by showing other circuits having similarly applied the precedent.”); Ouber v. Guarino, 293 F.3d 19, 26 (1st Cir.2002) (“[T]o the extent that inferior federal courts have decided factually similar cases, reference to those decisions is appropriate in assessing the reasonableness vel non of the state court’s treatment of the contested issue.” (internal quotation marks and citation omitted)); Matteo v. Superintendent, SCI Albion, 171 F.3d 877, 890 (3d Cir.1999) (“[W]e do not believe federal habeas courts are precluded from considering the decisions of the inferior federal courts when evaluating whether the state court’s application of the law was reasonable.”) (en banc).
III. Discussion
“Due process requires that the accused receive a fair trial by an impartial jury free from outside influences.” Sheppard v. Maxwell, 384 U.S. 333, 362, 86 S.Ct. 1507, 16 L.Ed.2d 600 (1966). The Supreme Court has held that when the consequence of a courtroom practice is that an “unacceptable risk is presented of impermissible factors coming into play,” there is “inherent prejudice” to a defendant’s constitutional right to a fair trial and reversal is required. Holbrook v. Flynn, 475 U.S. 560, 570, 106 S.Ct. 1340, 89 L.Ed.2d 525 (1986). In order to determine whether Musladin is entitled to federal habeas relief, we must therefore assess whether the buttons depicting Studer worn by spectators at the trial posed a risk of impermissible factors coming into play that is similar to those previously found to exist in other circumstances, such as in compelling a criminal defendant to wear prison garb and shackles before the jury, see Estelle v. Williams, 425 U.S. 501, 96 S.Ct. 1691, 48 L.Ed.2d 126 (1976), and in permitting spectators at a rape trial to wear anti-rape buttons, see Norris v. Risley, 918 F.2d 828 (9th Cir.1990). Because we conclude that no significant difference exists between the circumstances of this case and the “unacceptable risks” found to exist in Williams and Norris, we hold that the state court unreasonably applied established Supreme Court law in denying Mus-ladin relief.
a. Clearly Established Federal Law
The underlying federal law in this case — that certain practices attendant to the conduct of a trial can create such an “unacceptable risk of impermissible factors coming into play,” as to be “inherently prejudicial” to a criminal defendant — was clearly established by the Supreme Court in Estelle v. Williams, 425 U.S. 501, 96 S.Ct. 1691, 48 L.Ed.2d 126 (1976), and Holbrook v. Flynn, 475 U.S. 560, 106 S.Ct. 1340, 89 L.Ed.2d 525 (1986). In Williams, the Court considered whether compelling a criminal defendant to appear at his jury trial dressed in prison clothing violated his right to a fair trial. See Williams, 425 *1075U.S. at 503-06, 96 S.Ct. 1691. The Court found that the compelled wearing of prison clothing constitutes a continuous impermissible reminder to the jury of the defendant’s condition: an accused in custody who is unable to post bail. Id. at 505, 96 S.Ct. 1691. The Court held that the influence of prison clothing, and the message it conveyed to the jurors, impaired a defendant’s presumption of innocence. See id. at 503, 96 S.Ct. 1691. Noting these and other concerns, the Court concluded that because “[t]he defendant’s clothing is so likely to be a continuing influence throughout the trial ... an unacceptable risk is presented of impermissible factors coming into play.” Id. at 505, 96 S.Ct. 1691.
In Flynn, the court reaffirmed its holding in Williams regarding the “inherent prejudice” of courtroom practices that create an “unacceptable risk of impermissible factors coming into play,” but distinguished the case before it on the facts. The defendants in Flynn argued that the presence of four uniformed state troopers sitting in the front row directly behind them at trial led the jury to draw adverse inferences about them. Flynn, 475 U.S. at 563-64, 106 S.Ct. 1340. The Court explained that there are certain “courtroom practices [that it] might find inherently prejudicial,” but that the use of security officers to the extent involved did not fall into that category. Id. at 569, 106 S.Ct. 1340. As the Court explained, the “inferences that a juror might reasonably draw from the officers’ presence” in that case “need not be ... that [the defendant] is particularly dangerous or culpable.” Id. In distinguishing Flynn from Williams, the Court pointed out that the jury may not even have noticed that extra guards were being used in the trial, or most likely, drew no impermissible inference from their presence. The court stated, “[guards] are doubtless taken for granted so long as their numbers or weaponry do not suggest particular official concern or alarm.” Id. at 569. The law concerning the “inherently prejudicial” nature of courtroom practices which convey an impermissible message, however, remained unchanged and clear.
This court’s decision in Norris v. Risley, 918 F.2d 828 (9th Cir.1990), has persuasive value in an assessment of the meaning of the federal law that was clearly-established by Williams and Flynn and whether the state court’s application of that law in the case before us is objectively unreasonable. Like the present case, Norris involved the application of the Supreme Court’s “inherent prejudice” rule in assessing whether buttons worn by audience members during a trial created an “unacceptable risk of impermissible factors coming into play.” See Norris, 918 F.2d at 831-32. In Norris, the defendant was facing a criminal charge of rape. During the trial, several women sat in the spectator’s gallery wearing buttons that read “Women Against Rape.” Id. at 829. We noted that at any given time in Norris’s trial, approximately three women in the audience would be wearing the anti-rape buttons. Id. at 831. Faced with these facts, we applied Williams and concluded that “[j]ust as the compelled wearing of prison garb during trial can create an impermissible influence on the jury, throughout trial the buttons’ message ... constituted a continuing reminder that various spectators believed Norris’s guilt before it was proven, eroding the presumption of innocence.” Id. at 831. As we explained, because of the button’s clear communicative purpose, its impermissible message was far more clear and direct than that deemed unlawful in Williams:
Thus, though far more subtle than a direct accusation, the buttons’ message was all the more dangerous precisely because it was not a formal accusation. Unlike the state’s direct evidence, which could have been refuted by any manner *1076of contrary testimony to be judged ultimately on the basis of each declarant’s credibility, the buttons’ informal accusation was not susceptible to traditional methods of refutation. Instead, the accusation stood unchallenged, lending credibility and weight to the state’s case without being subject to the constitutional protections to which such evidence is ordinarily subjected.
Id. at 833.
Our reliance on Norris is appropriate for another reason: the last reasoned state court opinion identified Norris as setting forth the operative law as announced by the Supreme Court, and the state court sought to apply Norris when reaching its determination. Indeed, the state court’s unreasonable application of federal law lies in its misapplication of the Williams test, as it was explained in Norris, to the facts of this case.
The state court’s decision to apply Norris, and ours to afford it persuasive weight when determining the federal law as established by Williams, are particularly significant in light of the striking factual similarities between Norris and the present case. See Richardson v. Bowersox, 188 F.3d 973, 978 (8th Cir.1999) (“In determining whether a state court’s decision involved an unreasonable application of clearly established federal law, it is appropriate to refer to decisions of the inferior federal courts in factually similar cases.”).
b. Unreasonable Application of The Law
Although the state court identified the correct federal law to apply in adjudicating Musladin’s claim, citing Williams for the controlling principle, and properly looking to our decision in Norris as a persuasive application of that federal law in a factually similar case, the state court was objectively unreasonable both in its ultimate conclusion and in the rationale it employed in denying Musladin’s appeal. The California Court of Appeal justified its disposition of Musladin’s claim as follows:
[i]n contrast to the buttons in Norris, the message to be conveyed by the Stu-der family wearing buttons is less than clear. The simple photograph of Tom Studer was unlikely to have been taken as a sign of anything other than the normal grief occasioned by the loss of a family member. While we consider the wearing of photographs of victims in a courtroom to be an “impermissible factor coming into play,” the practice of which should be discouraged, we do not believe the buttons in this case branded defendant “with an unmistakable mark of guilt” in the eyes of the jurors.
People v. Musladin, No. H015159 at 21-22 (Cal.Ct.App. Dec. 9, 1997) (unpublished decision) (citing Flynn, 475 U.S. at 570-71, 106 S.Ct. 1340).
By disposing of Musladin’s claim in the above manner, the state court unreasonably applied federal law in two ways: (1) The state court’s attempt to distinguish this case from Norris on the basis of the facts misconstrued the Supreme Court’s rule regarding the “message” that certain courtroom practices send to the jury; and (2) the state court unreasonably imposed an additional and unduly burdensome requirement — demanding that the challenged practice cause the “brand[ing]” of the defendant with an “unmistakable mark of guilty” — even though the Williams test for finding “inherent prejudice” had already been met.
First, the state court erred in attempting to distinguish Norris. Norris simply cannot reasonably be distinguished. The message conveyed in the present case is even stronger and more prejudicial than the one conveyed in Norris. The state court unreasonably justified its conclusion by stating that, when compared to the buttons worn by spectators in Norris, the
*1077“message conveyed by the Studer family wearing buttons is less than clear.” This is simply not the case. Just as we held that the message sent by the anti-rape buttons was substantially more direct and clear than the message conveyed by the prison clothing in Williams, see Norris, 918 F.2d at 831, the message conveyed by the buttons depicting Studer in the case before us is substantially more direct and clear than that of the anti-rape buttons in Norris. In Norris, the buttons expressed the wearer’s position against rape but did not specify the defendant or the victim. In this case, the buttons actually depicted the individual that the defendant was charged with murdering and represented him as the innocent party, or the “victim.” Here, the direct link between the buttons, the spectators wearing the buttons, the defendant, and the crime that the defendant allegedly committed was clear and unmistakable. The primary issue at Mus-ladin’s trial was whether it was the defendant or the deceased who was the aggressor. The buttons essentially “argue” that Studer was the innocent party and that the defendant was necessarily guilty; that the defendant, not Studer, was the initiator of the attack, and, thus, the perpetrator of a criminal act.
The California court’s belief that buttons depicting the victim were “unlikely to have been taken as a sign of anything other than the normal grief occasioned by the loss of a family member” is even more incorrect as a matter of law than the view that interpreting the “Woman against Rape” buttons in Norris served no purpose other than women announcing a general statement against rape or expressing solidarity with, or support for, the rape victim in Norris’s case.1 We did not excuse the wearing of the buttons on that ground in Norris, and it was objectively unreasonable in light of Norris for the state court to do so here. See Norris, 918 F.2d at 831. In both Norris and the case before us, the law requires the courts to look beyond the general sentiment a button reflects and to determine the specific message that the button conveys in light of the particular facts and issues before the jury. Doing so here, a reasonable jurist would be compelled to conclude that the buttons worn by Studer’s family members conveyed the message that the defendant was guilty, just as the buttons worn by spectators in Norris did in that case.
Second, the state court unreasonably applied federal law by imposing an additional requirement when assessing Musladin’s “inherent prejudice” claim. Here, the court specifically found “the wearing of photographs of victims in a courtroom to be an ‘impermissible factor coming into play’ ” (emphasis added). Under Williams and Flynn, that finding, in itself establishes “inherent prejudice” and requires reversal. 'Indeed, the present finding goes beyond the finding that was held to require reversal in Williams and Norris and that was described as requiring rever*1078sal in Flynn. In the case before us, the state court found not only that an “unreasonable risk” existed that an impermissible factor would come into play, but that an impermissible factor actually had come into play. Nevertheless, after setting forth this finding, the state court added that, although the practice of wearing such buttons “should be discouraged,” Musladin was not entitled to relief because “the buttons in this case [did not] brand[ ] defendant ‘with an unmistakable mark of guilt’ in the eyes of the jurors.” The state court was unreasonable in imposing this additional requirement after it had concluded that the “inherent prejudice” elements had already been fully established.
The Supreme Court announced in Williams and Flynn that following a finding of an unacceptable risk of impermissible factors coming into play, no further showing is necessary because the practice is then deemed “inherently prejudicial.” Here, the state court flouted that rule: it required that the challenged practice not only constitute an unacceptable risk of an impermissible factor coming into play but also that it “brand” the defendant with an “unmistakable mark of guilt.” This additional test imposes too high and too unreasonable a burden on defendants and is contrary to established Supreme Court law. See Benn v. Lambert, 283 F.3d 1040, 1051 n. 5 (9th Cir.2002).2
We note that the “branding” with an “unmistakable mark of guilt” language employed in Flynn constituted only a descriptive comment. See Flynn, 475 U.S. at 571, 106 S.Ct. 1340 (quoting Williams, 425 U.S. at 518, 96 S.Ct. 1691 (Brennan, J., dissenting)).3 Both Williams and Flynn are clear as to the legal standard, and neither suggested that “branding” was necessary. Indeed, under the state court’s interpretation, the holding in Williams would not survive its own test. The Williams Court never found, or even implied, that the compelled donning of prison clothing would “brand [the] defendant ‘with an unmistakable mark of guilt’ in the eyes of the jurors.” Rather, the court’s concern was directed purely at the clothes’ role as a “constant reminder of the accused’s condition” — a “continuing influence throughout the trial,” principally because the requirement that defendants wear prison clothes “operates usually against only those who cannot post bail prior to trial.” Williams, *1079425 U.S. at 504-06, 96 S.Ct. 1691. At most, the Williams Court found that the shackling and prison clothes were “unmistakable indications of the need to separate a defendant from the community at large,” not that they would “brand” the defendant with an “unmistakable mark of guilt.” Flynn, 475 U.S. at 569, 571, 106 S.Ct. 1340. Although a practice that brands a defendant as guilty would surely be sufficient to demonstrate “inherent prejudice” and require reversal, branding is not a necessary element of establishing such prejudice. The state court’s imposition of the additional “branding” requirement was contrary to clearly established federal law and constituted an unreasonable application of that law.4
IV. Conclusion
In finding the wearing of buttons depicting the deceased to be an “impermissible factor coming into play,” the state court reached the point at which the Supreme Court “went no further and concluded that the practice [at issue wa]s unconstitutional.” Flynn, 475 U.S. at 568, 106 S.Ct. 1340. Instead of granting relief, however, the state court, disregarding the consideration that the central question was one of self-defense, unreasonably stated that the message conveyed through the wearing of the buttons in this case was not as clear as that conveyed by the anti-rape buttons in Norris. The state court then also unreasonably held that “branding” the defendant with “an unmistakable mark of guilt” is necessary to grant relief even though it had already found that “impermissible factors” had come into play before the jury. The state court did not simply engage in an incorrect application of Supreme Court law. Rather, its application of that law was contrary to the Court’s established rule of law and was objectively unreasonable. Accordingly, we reverse the district court’s denial of Musladin’s petition for habeas corpus and remand for issuance of the writ.
Reversed and Remanded.
. This is not to say that a showing of actual prejudice has no role in a Flynn challenge. As we held in Nonis, reversal is required if a defendant can prove either actual or inherent prejudice. See Norris, 918 F.2d at 828; accord Woods, 923 F.2d at 1457; see also United States v. Halliburton, 870 F.2d 557, 560 (9th Cir.1989) ("If we find that the jurors' brief view of Halliburton in handcuffs was not inherently prejudicial, we must then determine whether appellant has carried his burden of affirmatively showing actual prejudice.”).
. Justice Brennan’s dissent in Williams was the first time a member of the Court employed the "branding” language. Justice Brennan explained that regardless of whether prison clothing was compelled or not, the wearing of those clothes adversely affects a defendant's presumption of innocence, because, among other things, the clothing "surely tends to brand him in the eyes of the jury with an unmistakable mark of guilt.” Williams, 425 U.S. at 518, 96 S.Ct. 1691 (Brennan, J., dissenting) (emphasis added). When it was repeated in Flynn, the majority had already dismissed the defendant's challenge, and was simply noting in the alternative that "[e]ven had the jurors been aware that the deployment of troopers was not common practice in Rhode Island, we cannot believe that the use of the four troopers tended to brand respondent in their eyes 'with an unmistakable mark of guilt.’ ” Flynn, 475 U.S. at 571, 106 S.Ct. 1340 (quoting Williams, 425 U.S. at 518, 96 S.Ct. 1691 (Brennan, J., dissenting)). Flynn never stated that in order to prevail the defendant would have had to meet that standard. Rather, it may have simply been refuting the defendant's contention that the challenged practice had the described effect.
. We have consistently granted relief in this circuit on 'Williams-Flynn grounds without requiring the defendant to show that the challenged practice "branded” him with "an unmistakable mark of guilt.” See, e.g., Norris, 918 F.2d at 831; United States v. Olvera, 30 F.3d 1195, 1198 (9th Cir.1994) ("We conclude that [the challenged practice] posed an unacceptably high risk of influencing the jury’s judgment in a manner that undermined the presumption of innocence.”). Although we have found that proving a mark of unmistakable guilt is sufficient to establish prejudice, we have never held that it was necessary. To the contrary, we have noted that it may serve an alternative to the basic test, which remains the traditional standard by which we determine the existence of inherent prejudice. See Williams v. Woodford, 384 F.3d 567, 588 (9th Cir.2004) ("The Supreme Court has said that security measures at trial are inherently prejudicial when they 'tend[ ] to brand [the defendant] in [the jurors’] eyes with an unmistakable mark of guilt,' or when they create 'an unacceptable risk ... of impermissible factors coming into play.' " (emphasis added) (internal citations omitted)). Thus, branding may be an alternative means of establishing a violation, but it is not an additional requirement to be imposed after the violation has already been established under the traditional Williams-Flynn standard.