dissenting:
The Sixth Amendment guarantees every person accused of a crime — no matter how ugly the crime or how clear the guilt — “the right to a speedy and public trial.” U.S. Const, amend. VI. A public trial “en-sur[es] that judge and prosecutor carry out their duties responsibly ... and discourages perjury.” Waller v. Georgia, 467 U.S. 39, 46, 104 S.Ct. 2210, 81 L.Ed.2d 31 (1984). For these reasons, as the Supreme Court has recognized, the right to a public trial is critical to “[t]he central aim of a criminal proceeding” — affording the accused a fair trial. Id. Accordingly, there is a strong presumption that a trial will be open to the public; a presumption that can only be “overcome” in those “rare” circumstances where “the balance of interests [has been] struck with special care.” Id. at 45, 104 S.Ct. 2210 (emphasis added). Thus, prior to closing a trial, the presiding judge must: (1) determine that the party seeking closure has advanced “an overriding interest that is likely to be prejudiced,” (2) fashion a closure that is “no broader than necessary to protect that interest,” (3) “consider reasonable alterna*176tives to closing the proceeding,” and (4) “make findings adequate to support the closure.” Id. at 48, 104 S.Ct. 2210.
When Ernest Sutton Bell was tried in Pritt County North Carolina Superior Court in January 1994, the presiding judge removed the public and press from the courtroom during the entire testimony of the most critical prosecution witness1 without making findings adequate to support such drastic action. Indeed, nothing in the record even evidences that the judge recognized Waller as controlling, let alone considered alternatives or took “special care” to be sure that the closure “advanced an overriding interest ... likely to be prejudiced” and was “no broader than necessary to protect that interest.”
On direct appeal, Bell’s counsel failed to argue that the closure violated the Sixth Amendment right to a public trial. On state habeas review, North Carolina courts summarily, and without explanation, rejected Bell’s claim that this failure to pursue the Sixth Amendment violation on appeal constituted ineffective assistance of counsel. Because I believe the state court holding was “contrary to, or involved an unreasonable application of clearly established Federal law, as determined by the Supreme Court,” I respectfully dissent from the majority’s refusal to grant Bell’s petition for a writ of habeas corpus.
I.
A federal court may grant habeas relief only when a state’s adjudication of a petitioner’s claim “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.A. § 2254(d)(1). On state habeas review — termed a motion for appropriate relief in North Carolina — the state courts’ adjudication of Bell’s ineffective assistance of counsel claim was perfunctory, at best. The state trial judge’s order stated simply; “The undersigned, having reviewed the Motion for Appropriate Relief, finds that it fails to state a claim upon which relief can be granted. It is, therefore, denied.” This was followed by an Order signed by the Clerk of the North Carolina Court of Appeals that stated, “The petition filed in this cause on the 9th day of December 1996 designated ‘Petition for Writ of Cer-tiorari’ is denied .”
Under circuit precedent, such “perfunctory rejection” of a claim constitutes an “adjudication” for purposes of § 2254(d)(1). See Green v. Catoe, 220 F.3d 220, 223 (4th Cir.2000). See also Cardwell v. Greene, 152 F.3d 331, 839 (4th Cir.), cert. denied, 525 U.S. 1037, 119 S.Ct. 587, 142 L.Ed.2d 491 (1998); Wright v. Angelone, 151 F.3d 151, 156-57 (4th Cir.1998). Nonetheless, as we made clear- in Cardwell and recently reiterated in Green:
[Bjeeause the state court decision fails to articulate any rationale for its adverse determination of Cardwell’s claim, we cannot review that court’s “application of clearly established Federal law,” but must independently ascertain1 whether the record reveals a violation of [a constitutional right].
Green, 220 F.3d at 223; Cardwell, 152 F.3d at 339. See also Bacon v. Lee, 225 F.3d 470, 478 (4th Cir.2000); Baker v. Corcoran, 220 F.3d 276, 291 n. 14 (4th Cir.2000).
Because the North Carolina courts provided no indication of how they applied federal law to the facts of Bell’s constitutional claim, we must review their decision *177under the Cardwell/Green standard. Indeed, without state court analysis, we are left with no choice but to independently evaluate the record and determine if a constitutional violation occurred. Thus, we must (1) ascertain what law the Supreme Court has established as to the constitutional right to a public trial and effective assistance of counsel, (2) “independently ascertain whether the record reveals a violation” of these constitutional rights — i.e., whether the state court erred in denying the writ, and finally (3) determine if the state court decision — if erroneous — is also contrary to, or involves an unreasonable application of, clearly established law as determined by the Supreme Court. The majority omits completely the second part of this analysis.
A.
Since at least 1986, eight years prior to Bell’s trial, clear and abundant Supreme Court precedent established the necessity of specific findings before removing the public from a criminal trial. The Court initially addressed the right to a public trial in 1980 in the context of the First Amendment. In Richmond Newspapers, Inc. v. Virginia, 448 U.S. 555, 561, 100 S.Ct. 2814, 65 L.Ed.2d 973 (1980) (plurality opinion), the trial court closed the courtroom during a murder trial, refusing to allow the press to witness the proceedings. The question before the Supreme Court was whether the press and the general public had a First Amendment right of access to criminal trials. The Court answered this question in the affirmative, noting that “we are bound to conclude that a presumption of penness inheres in the very nature of a criminal trial under our system of justice.” Id. at 573, 100 S.Ct. 2814.
The Court then held that the trial court violated Richmond Newspapers’s First Amendment right of access to the murder trial by closing the courtroom during the proceeding, because “the trial judge made no findings to support closure.... Absent an overriding interest articulated in findings, the trial of a criminal case must be open to the public.” Id. at 580-81, 100 S.Ct. 2814 (emphasis added).
The Supreme Court again focused on the public’s First Amendment right of access to criminal trials in Globe Newspaper Co. v. Superior Court, 457 U.S. 596, 102 S.Ct. 2613, 73 L.Ed.2d 248 (1982). The Globe Newspaper Co. (“Globe”) unsuccessfully attempted to gain access to a rape trial in which the victims were underage females. Id. at 599, 102 S.Ct. 2613. In denying Globe’s request to gain access to the trial, the trial court cited a state statute requiring the exclusion of the press and the public during the testimony of a minor-victim of a sexual assault. See id. at 599, 602, 102 S.Ct. 2613. Globe objected to the mandatory closure statute, arguing that the statute violated its First Amendment right of access to criminal trials because it did not require a case-by-ease determination of whether the press should be excluded.
The Supreme Court agreed, noting that “the circumstances under which the press and public can be barred from a criminal trial are limited; the State’s justification in denying access must be a weighty one.” Id. at 606, 102 S.Ct. 2613. The Court held that a mandatory closure rule “cannot be viewed as a narrowly tailored means of accommodating the State’s asserted interest.” Id. at 609, 102 S.Ct. 2613. Instead, whether a trial court should close the courtroom must be analyzed on a case-by-case basis in which the trial court weighs “the minor victim’s age, psychological maturity and understanding, the nature of the crime, the desires of the victim, and the interests of parents and relatives.” Id. at 608, 102 S.Ct. 2613 (footnote omitted).
Four years later, in 1986, the Supreme Court specifically addressed a criminal defendant’s Sixth Amendment right to a public trial. See Waller v. Georgia, 467 U.S. 39, 104 S.Ct. 2210, 81 L.Ed.2d 31 (1984). In Waller, the trial court, over the defendant’s objection, closed the court*178room during a hearing on a motion to suppress in a criminal ease. See id. at 42, 104 S.Ct. 2210. The Court considered whether the trial court’s closure of the courtroom violated the defendant’s Sixth Amendment right to a public trial. Acknowledging that its prior precedents concerned the right to a public trial under the First Amendment, the Court nonetheless held “there can be little doubt that the explicit Sixth Amendment right of the accused is no less protective of a public trial than the implicit First Amendment right of the press and public.” Id. at 46, 104 S.Ct. 2210. The Court thus adopted the following standard for closing a courtroom: “the party seeking to close the hearing must advance an overriding interest that is likely to be prejudiced, the closure must be no broader than necessary to protect that interest, the trial court mtíst consider reasonable alternatives to closing the proceeding, and it must make findings adequate to support the closure.” Id. at 48, 104 S.Ct. 2210 (emphasis added). Moreover, the findings supporting closure cannot be “broad and general,” id.; rather, the state’s interest in closing the courtroom must be “articulated along with findings specific enough that a reviewing court can determine whether the closure order was properly entered.” Id. at 45, 104 S.Ct. 2210 (quoting Press-Enterprise Co. v. Superior Court of California, 464 U.S. 501, 510, 104 S.Ct. 819, 78 L.Ed.2d 629 (1984)) (emphasis added).
Thus, of particular relevance here, well prior to Bell’s 1994 trial, Supreme Court precedent — Globe, Waller, Richmond Newspapers, and Press-Enterprise — clearly established that: (1) a trial court cannot close the courtroom to the public during a criminal trial without specific, on-the-ree-ord findings supporting closure; and (2) a trial court may not enforce a per se rule mandating closure of a courtroom simply because the witness is a minor-victim of a sexual assault.
B.
It is clear under Supreme Court precedent that the removal of the public from the courtroom without specific findings supporting the decision violated Bell’s Sixth Amendment right to a public trial. As such, the failure of Bell’s appellate counsel to raise this claim constituted ineffective assistance.
1.
The following colloquy represents the only analysis conducted by the state trial court before closing the courtroom:
PROSECUTOR: And Judge, we have an outstanding motion I forgot to bring up earlier that the State had about closing the courtroom when at least Wendy testifies or all of the children testify at the appropriate time.
THE COURT: Is there any objection to that motion?
BELL’S COUNSEL: Yes, sir, we would object to closing the courtroom. We believe that would impact on our client’s constitutional right to a public trial. We would oppose it.
THE COURT: Well—
PROSECUTOR: I would argue that that is contrary to case law in this state.
THE COURT: The Court is going to allow that motion and we’ll do it in the most discreet way possible so that the jury doesn’t even notice it unless someone else calls it to their attention. We can take a short recess, and I can excuse the jury and I can tell the others — other people in the courtroom that this is testimony of an apparent delicate nature. I don’t see anything wrong with that. I am going to allow that motion.
J.A. 280-81 (emphasis added).
The trial court made no findings to support closure of the courtroom.2 The *179court’s conclusory statement that the testimony was of an “apparent delicate nature” obviously does not constitute “findings” that are “specific enough that a reviewing court can determine whether the closure order was properly entered.” Waller, 467 U.S. at 45, 104 S.Ct. 2210 (quoting Press-Enterprise, 464 U.S. at 510, 104 S.Ct. 819). The trial judge failed even to acknowledge defense counsel’s assertion that closing the courtroom would impact Bell’s constitutional right to a public trial. Indeed, the judge stated that he “didn’t see anything wrong with” clearing the courtroom without ever asking the witness if an open courtroom would inhibit her testimony, thus failing the first Waller requirement. He did not ask whether she feared for her safety, see, e.g., Woods v. Kuhlmann, 977 F.2d 74, 77 (2d. Cir.1992) (upholding closure because the judge questioned the witness to determine if fears of specific threats were credible), or whether she felt so intimidated as to warrant the drastic action of banning the public, see, e.g., Guzman v. Scully, 80 F.3d 772, 775-76 (finding a public trial violation because the judge never asked the witness whether she felt intimidated, or whether any intimidation was sufficient to overcome the presumption of an open trial).
Without knowing if a witness is threatened or intimidated, and if so, why, it is impossible to determine whether the threat or intimidation is so drastic as to warrant closure or whether removing the public would even cure the problem. For instance, if Wendy had indicated that Bell’s presence intimidated her, removing the public would not have alleviated that intimidation. The judge’s lack of inquiry falls far below the “special care” required to overcome the presumption that court proceedings should be open to the public.
In fact, it is hard to imagine that a court could grant a motion to close a courtroom in a more perfunctory manner. Indeed:
The trial court’s order excluding the public from the witness’ testimony, without any inquiry or findings concerning the specific condition of the witness in this case, is essentially equivalent to the blanket legislative closure rejected in Globe Newspaper.
Davis v. Reynolds, 890 F.2d 1105, 1111 (10th Cir.1989).
The trial judge also failed to consider reasonable alternatives to excluding the public and press from the courtroom. Waller, 467 U.S. at 48, 104 S.Ct. 2210. This is unsurprising, however, because it is difficult to consider alternatives to solve a yet-to-be determined problem. Moreover, the trial judge’s “solution” was certainly “broader than necessary” to protect any possible interest that could have been articulated by Wendy, id., because the judge not only cleared the courtroom during Wendy’s testimony, but also closed the courtroom prior to opening statements.3
As a result, contrary to established Supreme Court precedent, the state trial court violated Bell’s Sixth Amendment right to a public trial by closing the courtroom without making the necessary findings.
2.
Because Bell’s right to a public trial was so clearly violated, it is equally clear that appellate counsel’s failure to raise the public trial issue on appeal rendered his assistance constitutionally ineffective. The right to the effective assistance of counsel extends to a criminal defendant’s first appeal as of right. See Evitts v. Lucey, 469 U.S. 387, 396, 105 S.Ct. 830, 83 L.Ed.2d 821 (1985).
An ineffective assistance of counsel claim usually has two elements: counsel’s performance must have fallen “below an objective standard of reasonableness,” see Strickland v. Washington, 466 U.S. 668, 688, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), *180and there must be a “reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” Id. at 694, 104 S.Ct. 2052. However, as the majority properly recognizes, when, as here, the deficient performance constitutes structural error “the prejudice component of the Strickland analysis may be presumed.” Ante at 165, (citing McGurk v. Stenberg, 163 F.3d 470, 473-74 (8th Cir.1998)). See also Neder v. United States, 527 U.S. 1, 8, 119 S.Ct. 1827, 144 L.Ed.2d 35 (1999). Thus, if appellate counsel’s failure to pursue Bell’s right to a public trial fell below an objective standard of reasonableness, prejudice is presumed and the representation must be held constitutionally ineffective.
Of course, appellate counsel does not have a duty to raise every nonfrivolous argument on appeal, see Jones v. Barnes, 463 U.S. 745, 754, 103 S.Ct. 3308, 77 L.Ed.2d 987 (1983), but “a petitioner may establish constitutionally inadequate performance if he shows that counsel omitted significant and obvious issues while pursuing issues that were clearly and significantly weaker.” Mayo v. Henderson, 13 F.3d 528, 533 (2d Cir.1994). When appellate counsel fails to raise issues that are “clearly stronger than those presented,” the presumption that counsel rendered effective assistance will be overcome. Gray v. Greer, 800 F.2d 644, 646 (7th Cir.1985).
As demonstrated within, well prior to Bell’s 1994 trial, Supreme Court precedent required that trial courts express on-the-record findings before closing a courtroom. Moreover, this is not a case where trial counsel failed to object and appellate counsel was required to plumb the record to find the unconstitutional closure and contend that the closure constituted plain error. Here, Bell’s trial counsel did properly object to this clear constitutional violation, and yet appellate counsel still did not find this issue important enough to brief. Instead, appellate counsel chose to assert evidentiary challenges that had no likelihood of success. Appellate counsel’s failure to raise the one issue that would have been a “dead-bang winner” renders his assistance constitutionally ineffective. See Banks v. Reynolds, 54 F.3d 1508, 1515 (10th Cir.1995) (“[A]n appellate advocate may deliver deficient performance and prejudice a defendant by omitting a ‘dead-bang winner,’ even though counsel may have presented strong but unsuccessful claims on appeal”) (quoting United States v. Cook, 45 F.3d 388, 394-95 (10th Cir.1995)); see also Fagan v. Washington, 942 F.2d 1155, 1157 (7th Cir.1991) (“His lawyer failed to raise either [meritorious] claim, instead raising weaker claims.... No tactical reason — no reason other than oversight or incompetence — has been or can be assigned for the lawyer’s failure to raise the only substantial claims that [defendant] had.”).
In sum, controlling Supreme Court precedent establishes that Bell’s right to a public trial was violated and that appellate counsel’s failure to raise this claim constituted ineffective assistance of counsel.
C.
The only remaining question is whether the North Carolina courts’ clearly incorrect rejection of Bell’s ineffective assistance of counsel claim was “contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court.” 28 U.S.C.A. § 2254(d)(1).
The Supreme Court has instructed that a state court decision is “contrary to ... clearly established Federal law, as determined by the Supreme Court of the United States,” if the state court “arrives at a conclusion opposite to that reached by th[e Supreme] Court on a question of law,” Williams v. Taylor, 529 U.S. 362, 120 S.Ct. 1495, 1523, 146 L.Ed.2d 389 (2000). A state court decision involves an unreasonable application of clearly established federal law if the state court “identifies the correct governing legal principle from th[e Supreme] Court’s decisions but unreason*181ably applies that principle to the facts of the prisoner’s case.” Id.
In this case, of course, we simply do not know what governing legal principles (if any) formed the basis of the state courts’ rejection of Bell’s ineffective assistance claim. We do not know why they concluded that closure of Bell’s trial during the testimony of the most crucial prosecution witness, without any indication that the required Waller factors had even been considered, let alone correctly analyzed, did not constitute obvious reversible error that should have been raised by any competent appellate counsel. The North Carolina courts did not provide any reasoning or cite a single legal principle or precedent, let alone Supreme Court precedent, in support of their decision.
The record suggests that the North Carolina'courts most likely followed legal principles directly contrary to those mandated by the Supreme Court. In opposing Bell’s objection to closure at trial, the State asserted that refusal to grant the closure would be “contrary to case law'in this state.” Similarly, when opposing Bell’s petition for state habeas relief, the State relied heavily on State v. Burney, 302 N.C. 529, 537, 276 S.E.2d 693, 698 (1981), a pre-Waller case in which the state court concluded that the defendant could not “fashion support for a sixth amendment claim from a case which has manifest first amendment underpinnings” and in any event, any error in closing the proceedings was “harmless.” If the state courts based their decision on either or both of these rationales, then certainly their decisions are contrary to “clearly established Federal law, as determined by the Supreme Court.” 28 U.S.C.A. § 2254; see Williams, 120 S.Ct. at 1523. This is so because both of these rationales are diametrically opposite to the conclusions reached by the Supreme Court in Waller.4
Of course, although it seems unlikely, the state habeas courts may have “identified the correct governing legal principles” from controlling Supreme Court precedent (Waller, Globe, Richmond Newspapers, and Press-Enterprise) and concluded that the state trial court had complied with those principles. This seems improbable in view of the fact that the State never argued on state habeas review that the trial court’s closure accorded with Waller, Globe, Richmond Newspapers, Press-Enterprise, or other Supreme Court precedent. If the state habeas courts had chosen to rely on an argument never made by the state, it seems likely that they would have thought the issue worth more than a perfunctory denial. But assuming — even though there is no evidence this occurred — that the state habeas courts did identify the correct governing legal principles, I believe that their denial of Bell’s habeas petition was an “unreasonable application of, clearly established Federal law, as determined by the Supreme Court.”
I recognize that not every incorrect decision is unreasonable, see Williams, 120 S.Ct. at 1522, and that a federal habeas court cannot simply substitute its own judgment for that of the state court. Rather, in a “close” case, we must uphold the state court decision even if we disagree with the outcome. See Tucker v.. Catoe, 221 F.3d 600, 614 (4th Cir.2000) (denying habeas on a “close issue” even though the state court was incorrect). See also Francis v. Stone, 221 F.3d 100, 113 (2d Cir.2000) (upholding state court habeas denial because it was a “close question,” even though state court might well have been wrong).
*182But just as surely, when a state court fails to provide any rationale for its decision, and when our mandated independent review of the record reveals that the constitutional violation was not a close question, it is our duty to correct an unreasonable application of Supreme Court precedent. For example, in Delgado v. Lewis, the Ninth Circuit’s examination of a state court denial of habeas was “impeded ... because no rationale for [the state court’s] conclusion was supplied.” 223 F.3d 976, 981 (9th Cir.2000). The federal court reversed the state court decision because it had “a definite and firm, conviction that an error ha[d] been committed.” Id. (quoting Van Tran v. Lindsey, 212 F.3d 1143, 1153 (9th Cir.2000)) (emphasis added).
The constitutional ineffectiveness of Bell’s appellate counsel is not a close issue. There was “no semblance of a tactical reason” for not pursuing Bell’s public trial claim. Washington v. Smith, 219 F.3d 620, 630 (7th Cir.2000). Bell’s counsel raised four arguments on appeal, none of which possessed the obvious strength of the public trial violation. Indeed, disputing a judge’s evidentiary discretion rather than appealing his complete disregard for the mandates of the Supreme Court on a matter of constitutional law is hardly the action of reasonable counsel. Because the constitutional violation was clear, the state courts’ summary dismissal of Bell’s habeas petition was an “objectively unreasonable” application of federal law.
II.
I find puzzling and unconvincing the majority’s holding that North Carolina’s rejection of Bell’s claim was not objectively unreasonable.
A.
Puzzling, because the majority never attempts an essential part of the analysis, i.e., an independent determination of whether Bell’s appellate counsel was in fact constitutionally ineffective. See Green, 220 F.3d at 223; Cardwell, 152 F.3d at 339. Instead, the majority simply cites Strickland and concludes that the state courts’ denial of Bell’s ineffective assistance claim was “neither contrary to nor an unreasonable application of clearly established federal law.” Ante at 175. The majority engages in no independent determination of whether the state courts erred; only a conclusion that if there was error, it was not unreasonable. This directly contradicts the Cardwell/Green mandate that we “independently ascertain whether the record reveals a violation of [a constitutional right].” Green, 220 F.3d at 223; Cardwell, 152 F.3d at 339.
The majority does not contend to the contrary. Rather, in response to my suggestion that it has failed to follow recent, directly controlling circuit precedent, the majority simply overrules that precedent.5 Although I recognize, that the en banc court can create and destroy circuit precedent as it chooses, I believe that the majority’s action today is unwise. There is no reason, except expediency, to reject the well-established practice of this court — followed by all of our active judges, including every member of today’s majority-of making an independent determination of whether a constitutional violation has occurred when evaluating whether a summary state court decision “involved an unreasonable application” of federal law under § 2254(d). See Goins *183v. Angelone, 226 F.3d 312, 324 (4th Cir.2000) (finding that the state court “correctly rejected th[e] claim and, by definition, its conclusion was not unreasonable”); Bacon v. Lee, 225 F.3d 470, 479 (4th Cir.2000) (concluding that counsel was not ineffective and thus the state court habeas denial was not “unreasonable” under § 2254(d)); Tucker v. Catoe, 221 F.3d 600, 614 (4th Cir.2000) (finding error, but nonetheless concluding that the state court was not unreasonable); Barnabei, 214 F.3d at 469 (discussed infra at n. 5); Baker v. Corcoran, 220 F.3d 276, 293 (4th Cir.2000) (concluding that a jury instruction was correct and an ineffective assistance claim was without merit, and thus the state habeas court did not unreasonably apply federal law in denying the writ); Cardwell, 152 F.3d at 339.
Indeed, the Supreme Court has twice recently indicated that our well-established practice is proper. In Williams, while not addressing this exact question, the Court’s language suggests that ours is the appropriate methodology. The Court instructed:
Under 2254(d)(l)’s “unreasonable application” clause, then, a federal habeas court may not issue the writ simply because that court concludes in its independent judgment that the relevant state-court decision applied clearly established federal law erroneously or incorrectly. Rather, the application must also be unreasonable.
Williams, 120 S.Ct. at 1522 (emphasis added). The Williams Court thus recognized the appropriateness of federal habeas courts independently analyzing asserted claims as long as they “also” engage in the AEDPA reasonableness determination. Moreover, just last term the Supreme Court itself addressed an AEDPA claim by first evaluating whether the state court decision was erroneous, and only after concluding that it was not, holding that the state court had not unreasonably applied federal law. Weeks v. Angelone, 528 U.S. 225, 120 S.Ct. 727, 734, 145 L.Ed.2d 727 (2000). See also Van Tran, 212 F.3d at 1155 (requiring federal courts to first determine whether the state court’s decision was erroneous because it “promotes clarity in our own constitutional jurisprudence and also provides guidance for state courts, which can look to our decisions for their persuasive value.”)6
Under the AEDPA, federal courts have no business reversing even perfunctory state habeas decisions that are “close” to the mark. See Tucker, 221 F.3d at 614. A federal court must, however, find the mark when the state court fails to do so by issuing a summary decision. At some point in the judicial process, even a person convicted of heinous crimes deserves a rigorous and complete analysis of his constitutional claims. If no state court provides such analysis, this task falls to the federal courts. A summary conclusion that a case is “close,” without more, does not afford habeas petitioners a full and fair opportunity to have their habeas claims diligently considered. In many cases, issues that appear “close” at first may, in fact, show themselves as clearly wrong after an exacting review.
That is not to say that federal courts may substitute their analysis for a state *184court decision. But, in the process of evaluating constitutional violations, federal ha-beas courts should satisfy themselves, the reviewing court, and the petitioner that a “close” issue was in fact “close,” and a state court decision, no matter how cursory, was reasonable.
It seems to me that the majority’s careful avoidance of the question of whether the state courts’ decision was contrary to establish federal constitutional law can only be explained as at least implicit recognition that counsel’s failure to appeal the blatant public trial violation clearly violated federal law. This obviously undermines the majority’s conclusion that the decision was nonetheless objectively reasonable. After all, the North Carolina courts enunciated no reason for their decision and from what can be gleaned from the record this most likely was because they were unfaithful to well-established Supreme Court precedent. State courts should not be allowed to insulate their decisions by failing to express their reasoning.
B.
The majority attempts to obscure the clear and unreasonable ineffectiveness of Bell’s appellate counsel by portraying the trial judge’s courtroom closure as a “proper” decision. The majority offers two grounds for this portrayal. Neither is convincing.
First and principally, the majority constructs a new record to take the place of the missing findings. Thus, the majority recounts in some detail what the state judge was “aware of’ or “made aware of’ or “information” that he “possessed.” Ante at 170-72. This may or may not be so; we simply do not know what the trial judge was aware of and what information he possessed. More importantly, we do not know how, if at all, matters of which he was “aware” or “information he possessed” figured in his decision to close the courtroom. We do not know because, contrary to clear Supreme Court precedent, the trial court did not provide findings justifying the closure.
The sole source the majority provides for its suggestions as to the trial judge’s “awareness” is a pretrial hearing. But, no witnesses testified at this hearing. Whatever information the trial judge gleaned from it was based entirely on the prosecutor’s argument; surely a lawyer’s argument does not provide a satisfactory basis for assessing guilt, or denying a public trial. Nor does an affidavit filed by the state trial judge, four years after the trial, provide a basis for the closure. Tellingly, in this affidavit, the judge merely states that he observed that Wendy was under eighteen and gives assurances that he took the embarrassment and emotional well-being of the child into account in deciding to close the courtroom. Not even in this after-the-fact statement does the judge assert that “an overriding interest likely to be prejudiced” compelled the closure he ordered or that he made “implicit” findings supporting the closure.
Even more significantly, in Waller, the Supreme Court expressly rejected the view that the failure to make express findings can be corrected by post hoc analysis of the record. See Waller, 467 U.S. at 49 n. 8, 104 S.Ct. 2210 (“The post hoc assertion by the Georgia Supreme Court that the trial court balanced petitioners’ right to a public hearing against the privacy rights of others cannot satisfy the deficiencies in the trial court’s record.”). Indeed, Waller made it clear that trial courts must make findings “specific enough that a reviewing court can determine whether the closure order was properly entered.” Id. at 45, 104 S.Ct. 2210. The majority’s analysis turns Waller on its head. Under the majority’s reasoning, a trial court no longer has to make specific findings so that an appellate court can properly review those findings; instead, an appellate court can now correct the errors of the trial court by relying on the trial judge’s post hoc justifications or by creating its own “findings,” *185albeit without considering any evidence or viewing a single witness.
The majority’s post hoc creation of a record to support the closure of a courtroom also has the effect of violating the Supreme Court’s holding in Globe, which prohibited a per se closure of a courtroom during the testimony of a minor-victim of a sexual assault. As the trial court never personally evaluated Wendy’s state of mind or inquired into why she could not testify in open court, it could only have based its decision upon the general nature of the crime alleged and the age of the witness. To base a closure decision on these general characteristics that are present in virtually all child abuse crimes is to apply a per se rule. A case-by-case inquiry requires the trial court to actually inquire into why a particular witness would encounter difficulty testifying in a public courtroom.
Requiring explicit on-the-record findings is not a meaningless formality. This requirement forces trial courts to weigh the evidence supporting and opposing closure in a systematic fashion, thereby ensuring that courtrooms are not closed after perfunctory dismissal of a defendant’s Sixth Amendment rights. Under the majority’s analysis, however, a trial court need not make an individualized determination that closure is necessary. Rather, the majority would allow a trial court to close the courtroom during the testimony of every minor-victim of a sexual assault without any express findings supporting closure.
The majority’s other reason for finding that the state courts did not “unreasonably reject” Bell’s claim is that the instant case assertedly involved only a “partial closure.” Ante at 172-75. The majority contends that it was not unreasonable for North Carolina to fail to apply the principles so clearly articulated in Supreme Court precedent because the Supreme Court cases involved complete closures of a courtroom, while the present case involves only a “partial, closure.” But in fact, the Supreme Court dealt with a “partial closure” of a courtroom in Press-Enterprise, 464 U.S. at 503, 104 S.Ct. 819. There, three days of a six-week voir dire were open to the public, and the remainder of the proceeding was closed. See id. The proceeding thus was a “partial closure” under the majority’s analysis. The Supreme Court nonetheless required “findings specific enough that a reviewing court can determine whether the closure order was properly entered.” Id. at 510, 104 S.Ct. 819. Accordingly, the Court held that petitioner’s First Amendment right of access to a criminal trial was violated, because there had been “a failure to articulate findings with the requisite specificity.” Id. at 513, 104 S.Ct. 819. The majority’s “partial closure” distinction thus finds no support in Supreme Court precedent.7
The majority also attempts to buttress its “partial closure” distinction by relying on decisions from other circuits to support the notion that express findings are not always necessary when dealing with some “partial closures.” See ante at 172-75. The majority apparently believes that North Carolina’s decision cannot be “unreasonable” under the AEDPA if the decision finds some support in case law from other circuits. Such a rationale echoes reasoning specifically rejected by the Supreme Court in Williams. There the Court made it clear that “[t]he federal habeas court should not transform the inquiry into a subjective one by resting its determination ... on the simple fact that at least one of the Nation’s jurists has applied the relevant federal law in the same manner the state court did in the habeas petitioner’s case.” Id. at 1521-22. Even if some circuits had held that every closure of less than an entire proceeding *186constituted a partial closure, and that such partial closures do not need to be justified by express findings, such a holding would still be unreasonable because it directly conflicts with Supreme Court precedent, i.e. Press-Enterprise.8
Of course, no circuit has held anything of the kind. Rather, even in those cases on which the majority so heavily relies, courts have either ordered removal of only members of the defendant’s family, see United States v. Osborne, 68 F.3d 94, 99 (5th Cir.1995) (ordering removal only of defendant’s sister allowing all other spectators including other relatives of the defendant to remain in court); Woods v. Kuhlmann, 977 F.2d 74, 76 (2d Cir.1992) (permitting exclusion only of members of defendant’s family during one witness’s testimony), or required findings of, or at least some knowledge of, a specific threat to the testifying witness, in order to bar the press and public. See Ayala v. Speckard, 131 F.3d 62, 64-65 (2d Cir.1997) (en banc) (closure of courtroom permitted to protect security of undercover police officers after a hearing on the issue); United States v. Farmer, 32 F.3d 369, 372 (8th Cir.1994) (upholding closure because there was evidence in the record that the defendant had threatened the victim/witness who feared retaliation by the defendant and his family).
In contrast, Bell’s trial was totally closed during the testimony of the most crucial prosecution witness; neither the public nor the press was allowed to remain to safeguard the fairness of the proceeding.9 Moreover, the trial judge held no hearing, made no findings, questioned no witnesses, knew of no specific threat to any witness, and based his decision on nothing more than the “apparent delicate nature” of the testimony. See also United States v. Galloway, 937 F.2d 542, 547 (10th Cir.1991) (in a case very similar to Bell’s, reversing and remanding to district court for further findings regarding a partial closure). Consequently, it is clear that Bell’s right to a public trial was violated and counsel’s failure to pursue that argument on appeal was undeniably ineffective assistance.
III.
If Ernest Sutton Bell committed the ugly crimes charged against him, he merits stern punishment. But no matter how dreadful his crimes, they do not entitle a state to deny his constitutional right to a public trial, as North Carolina did. When Bell maintained that his appellate counsel provided ineffective assistance in failing to so argue, North Carolina courts rejected his claim without explanation. That rejection was either contrary to, or involved an unreasonable application, of clearly established law as determined by the Supreme Court. Accordingly, Bell’s petition for a writ of habeas corpus should be granted.
Judge MICHAEL joins this dissent.
BUTZNER, Senior Circuit Judge,dissenting:
I dissent for the reasons stated in Bell v. Jarvis, 198 F.3d 432 (4th Cir.1999).
. The majority notes that of the ''700-plus-page trial transcript, spanning four days, approximately forty-four pages comprised the testimony of Wendy,” ante at 156; however, two trial days and more than 400 transcript pages were devoted to jury selection, argument of counsel, and other non-evidentiaiy matters. Undoubtedly, Wendy’s testimony, which was more than twice as long as that of any other prosecution witness, constituted the most critical evidence against Bell. Yet, over Bell’s objection, the trial judge closed the courtroom during this crucial testimony, without citing any precedent or making adequate findings.
. As noted above, because the presiding judge failed to make adequate findings, nothing in the record demonstrates that he considered, *179let alone correctly applied, the other Waller factors.
. There is no indication that the public or press ever re-entered the courtroom.
. In Waller the Court expressly held "there can be little doubt that the explicit Sixth Amendment right of the accused is no less protective of a public trial than the implicit First Amendment right of the press and public,” 467 U.S. at 46, 104 S.Ct. 2210, and that "the defendant should not be required to prove specific prejudice in order to obtain relief for a violation of the public-trial guarantee,” thus, "the harmless error rule is no way to gauge the great, though intangible, societal loss that flows from closing' courthouse doors.” Id. at 49, 50 & n. 9, 104 S.Ct. 2210 (internal quotation marks omitted).
. I agree with the majority that the language in Cardwell seemingly requiring de novo review cannot be reconciled with the AEDPA. Of course, this does not in any way require disavowal of our well-established practice of initially making an independent determination as to whether a state court decision was erroneous. Indeed, in Barnabei v. Angelone, 214 F.3d 463 (4th Cir.2000), we recognized that although "de novo review by a federal habeas court remains inappropriate under § 2254(d),” a summary state court adjudication forces us to "independently ascertain whether the record reveals a violation.” Id. at 477 (internal quotations omitted).
. As the Van Tran Court noted, this approach is the only one consistent with the Supreme Court’s recent clarification, in the analogous area of qualified immunity, that before a court may decide whether a stale employee violated a constitutional right that was “clearly established,” it "must” decide whether a constitutional right was violated in the first place. Wilson v. Layne, 526 U.S. 603, 609, 119 S.Ct. 1692, 143 L.Ed.2d 818 (1999) (holding that it is improper to dismiss a claim by simply concluding that a constitutional right was not “clearly established” without reaching the antecedent question); County of Sacramento v. Lewis, 523 U.S. 833, 841 n. 5, 118 S.Ct. 1708, 140 L.Ed.2d 1043 (1998) (“[T]he better approach to resolving cases ... is to determine first whether the plaintiff has alleged a deprivation of a constitutional right at all. Normally, it is only then that a court should ask whether the right allegedly implicated was clearly established at the time of the events in question.”)
. While the closure in Press-Enterprise was more extensive than the closure in this case, it only involved the voir dire of prospective jurors while the closure here involved a far more critical portion of the trial — the testimony of the key prosecution witness, whose testimony was necessary to secure a conviction for the offenses charged.
. To the extent that we should look beyond Supreme Court precedent under the AEDPA (and we should not), the application of Supreme Court precedent by the North Carolina Court of Appeals has substantially more probative value than decisions by other federal circuits. And the North Carolina Court of Appeals has required express, on-the-record findings before allowing a "partial closure” of a courtroom during the testimony of a victim of a sexual assault. See State v. Jenkins, 115 N.C.App. 520, 525, 445 S.E.2d 622, 625 (1994). Jenkins was governing precedent in North Carolina at the time of Bell’s appeal, making it clear beyond any doubt that appellate counsel’s future to raise the public trial argument was indeed the failure to raise a "dead-bang winner” and, therefore, ineffective assistance. See Banks, 54 F.3d at 1515.
. The trial judge allowed the complaining witness’s family to remain, but that certainly did nothing to protect Bell’s right to a fair trial.