Jones v. Murphy

Judge POOLER dissents in part in a separate opinion. GERARD E. LYNCH, Circuit Judge: Petitioner-appellant Jermaine Jones was convicted of murder by a Connecticut jury in 2004. The Connecticut Supreme Court affirmed his conviction by opinion in 2007. In 2010, Jones filed a petition for a writ of habeas corpus under 28 U.S.C. § 2254 in the United States District Court for the District of Connecticut. That court (Janet C. Hall, Judge) denied the petition, and this Court granted a certificate of appealability as to some of his claims. Jones argues that the state trial court’s decision to exclude him from the courtroom violated his federal constitutional right to be present at his trial, and that police interrogation shortly after the murder violated his Miranda rights. We find that any error in initially excluding Jones was harmless because he did not miss any critical stage of the trial as a result of that exclusion. Furthermore, because the subsequent decision to continue Jones’s exclusion was attributable to his own violent conduct, the Connecticut Supreme Court reasonably applied United States Supreme Court precedent in upholding the trial court’s decision. Jones’s two claims of improper police procedure in his interrogation are also meritless. One claim misapprehends Supreme Court precedent; the other was never raised in state court and is thus unavailable for review in federal court. We therefore affirm the denial of the writ. BACKGROUND I. The Crime The facts of the underlying crime are essentially undisputed. A more detailed recitation is available in the Connecticut Supreme Court’s opinion. See State v. Jones, 281 Conn. 613, 916 A.2d 17, 21-22 (2007). In June 2001, Jones lived with his then girlfriend, Erica Minnifield, in Hartford, Connecticut. On June 22, 2001, Jones and Minnifield traveled to Waterbury to visit friends and family. Minnifield, without Jones, then went to a shopping mall with *230 Thomas Williams, the victim, and purchased clothing. Later that evening, Jones saw Minnifield wearing new clothing and suspected that another man had purchased it for her. While being driven back to Hartford together later that night by a friend, Jones confronted Minnifield about the new clothing. During the confrontation, Jones threatened Minnifield with a knife and cut her on the arm. When the couple arrived at home, Minnifield ran into a neighbor’s yard, but Jones found her and dragged her back into the house. He then cut her pants from her body using the same knife. Minnifield fled the next morning. Later that day, Jones saw Minnifield driving with Williams and became enraged. Jones went with a friend to look for Williams. When they found him, Jones approached Williams and cocked a handgun hidden in his jacket. Jones then asked Williams, “[Y]o, you gonna stop fucking around with my girl?” Williams laughed and asked Jones what he meant. Jones told him that he would kill Williams if he saw him again with Minnifield. Jones then removed the handgun from his jacket and shot Williams four times. Williams died shortly thereafter. II. Police Investigation and Jones’s Confession Jones was arrested at his home on June 27, 2001, between 9:00 and 10:00 a.m. Scott Stevenson, a detective with the Waterbury Police Department, executed the arrest warrant and, after searching the home, brought Jones back to the police station,where he was to be interviewed by detectives assigned to the case, between noon and 1:00 p.m. Because those detectives were not available, Stevenson was assigned to guard Jones in the interview room, the door of which did not lock. Stevenson was not instructed to interview Jones or ask him any questions. Because it was lunchtime, Stevenson ordered food for himself and Jones, and the two men had a long conversation on various topics unrelated to the arrest. Eventually, Jones, unprompted, said that he hadn’t killed anyone. Stevenson ignored the statement, but shortly thereafter, Jones said, “I know you guys think it’s about the girl.” According to the Connecticut Supreme Court, “Stevenson then asked the defendant if he had known the victim. The defendant responded by placing his head in his hands and stating, ‘He did not deserve to have happen what I did to him.’ ” State v. Jones, 916 A.2d at 40 (alterations omitted). Following that statement, Stevenson left the room to ask how to proceed; he was instructed to deliver Miranda warnings and then seek a confession. Stevenson advised Jones of his rights, and Jones then provided an oral confession. After signing an acknowledgment that he knew his rights, Jones signed a written confession as well. Jones also helped the police to find the murder weapon by telling police to speak to his brother, who knew the weapon’s whereabouts. When his brother was initially unwilling to aid police, Jones spoke to him on the phone, after which his brother led police to the weapon. III. Pretrial Proceedings and Trial On January 27, 2004, the trial court held a hearing on Jones’s motion to suppress his confession. The court also heard arguments about whether and to what extent the prosecution would be permitted to discuss Jones’s threats to Minnifield and Minnifield’s knife injuries. The parties agreed, and the court ruled, that Minnifield’s “injuries [could] be noted,” since they were relevant to understanding the sequence of events, but that the knife *231 should not be mentioned as the source, as Jones was not charged with assaulting Minnifield. On January 30, the court denied the motion to suppress the confession. Jones attempted personally to dispute the court’s ruling, but the court told Jones that he could appeal the ruling later. On the way back to jail that evening, Jones, angry at the ruling, punched his hand through a Plexiglas window. The trial court learned of this incident during pretrial proceedings on February 2, 2004, the first day of trial. The prosecution called Minnifield as one of its first few witnesses on February 2. During the course of Minnifield’s testimony, the assistant state’s attorney asked Minnifield several questions that went beyond the scope of the court’s order concerning Jones’s possession and use of the knife. Defense counsel did not object. During a recess, the court admonished the assistant state’s attorney. Defense counsel explained that he had not objected at the time because he did not want to draw the jury’s attention to the testimony, but he requested a cautionary instruction. The court agreed that such an instruction was the proper remedy and subsequently delivered it. At the beginning of proceedings on February 3, before the jury entered the courtroom, defense counsel alerted the court that Jones would like to seek a mistrial on the basis of Minnifield’s testimony. Defense counsel then informed the court that although counsel was satisfied that the court had adequately resolved the issue, Jones was not, and “as a result of what [Jones] considered to be [ ] the prejudicial impact of [ ] Minnifield’s testimony,” Jones “would choose personally at this point not to go further with the proceedings.” To the extent there was any ambiguity about the meaning of counsel’s statement, Jones himself made clear that he wanted leave the courtroom. Jones addressed the court himself, stating that in light of the court’s rulings, “I don’t even want to be here then.” A colloquy ensued between the court and Jones. The court warned Jones that “[t]he case is going to proceed, you know, without you” and that Jones was “going to be prejudiced by this.” The court also repeatedly asked if Jones understood what he was doing. The colloquy culminated in the following exchange: THE COURT: Mr. Jones, you understand that you’re going to be prejudicing yourself by leaving today; you understand that? Do you understand that? THE DEFENDANT: Well, I want to— Pm going to put a motion in to dismiss counsel, then. THE COURT: No, it’s— THE DEFENDANT: For the record. THE COURT: No, Mr. Jones, we’re proceeding— THE DEFENDANT: And I’ll take my case myself. THE COURT: — we are proceeding— THE DEFENDANT: I’ll take my case myself, Your Honor. THE COURT: — that’s denied, Mr. Jones. THE DEFENDANT: I have a right, Your Honor. That’s my Constitutional right. THE COURT: All right. THE DEFENDANT: You’re laughing, but I’m serious. THE COURT: Marshals, remove Mr. Jones. A MARSHAL: Let’s go, Mr. Jones. THE DEFENDANT: I’m not going nowhere, man. A MARSHAL: Come on. THE DEFENDANT: No, I’m not going nowhere. *232 A “scuffle ensued” in the courtroom. With order restored, the trial court stated that “Mr. Jones had to be physically restrained by a number of marshals. I guess to say Mr. Jones is a violent man is understating the obvious.” The court ordered a three-hour recess for Jones to calm down and consult with defense counsel. After the recess, and after a further discussion with counsel, the court excused the jury for the remainder of the day. The court also had a discussion with the chief marshal, Anthony Candido, about Jones’s behavior. Candido informed the court that after he was removed, Jones had made threatening comments that he would “continue to fight” and would “bring you people down” and had refused to place his hands behind his back. Candido also noted that one marshal had to be sent to the hospital as a result of injuries sustained in attempting to restrain Jones. Candido recommended that Jones be kept in full restraints if he returned. The court stated: I’m very concerned, having observed ... the defendant over the course of jury selection, especially after ruling on his motions, his reaction thereto, and it just appears every time the defendant gets an adverse ruling, he reacts and reacts violently at times, including breaking his hand after ... what he viewed as losing on the motion to suppress. The court also stated that it was “not at all sure that this trial could proceed in an orderly fashion with Mr. Jones in the courtroom,” and that it was “concerned, as is Marshal Candido, about the safety of people in the courtroom. That includes the prosecutors, the other court personnel, the jury and his own counsel.” The court concluded that Jones had “forfeited his right to ... be in the courtroom by his engaging in this disruptive and volatile, disorderly, disrespectful conduct.” Furthermore, the court was not optimistic that, despite what he says or may say down the road — and I haven’t heard anything yet — but even if he were to promise to try to behave, based on having observed him over the course of the last two weeks or so, I am very concerned for the court personnel, who will be closer to him than I[,] for their safety because he just does not control himself and does not appear to want to control himself, more importantly. The court then engaged counsel in a discussion of how to proceed. Defense counsel stated that Jones no longer wished even to be brought to the courthouse but suggested that Jones be given the option each day of coming to the courthouse or remaining at the jail. The court adopted this suggestion and noted that a holding cell adjacent to the courtroom was equipped with a monitor and speaker system which would permit Jones to monitor the trial. The court also said, however, that it wouldn’t “force him” to attend if he “does not want to come to court.” The court again expressed concern about the “personal safety of court personnel, attorneys, and correctional personnel,” and that “based on everything the Court has observed and stated today and heard today,” it would “not [] bring [Jones] into the courtroom unless he is highly restrained.” The court noted, however, that it would leave the decision to defense counsel and Jones and that it would “listen to any other requests.” On the morning of February 4, defense counsel told the trial court that Jones had come to court and wished to be present in the courtroom, though he had not agreed to wear full restraints. The trial court then asked the advice of the deputy chief marshal, Gino DiMauro, about whether it would be safe to permit Jones back in the *233 courtroom. DiMauro initially indicated that he felt the marshals would be able to “handle” Jones, but he left to speak to Jones and assess the situation. During DiMauro’s absence, the court expressed further concern about the safety of courtroom personnel and Jones’s ability to keep himself under control. The court also suggested that having Jones absent might be less prejudicial than having the jury see him in full restraints. When DiMauro returned, he reported to the court that Jones was being “somewhat confrontational” toward the marshals, and that in his “better judgment” he “[could not] say that the defendant should be present in court during the proceedings.” Di-Mauro also said that he could not guarantee that another outburst would not occur, and that he did not “believe that Mr. Jones should be in these proceedings while they’re going on for the safety of everybody involved.” In discussion with the court, the marshal confirmed that Jones was agitated and was “still talking about yesterday.” In light of the earlier events and DiMauro’s assessment, the court then reconsidered its ruling and found that Jones had “by his disruptive behavior, waived any right to be present during the proceedings.” The court reiterated its concerns that bringing Jones back would be unsafe and disruptive. After consulting Jones about his options, defense counsel informed the court that Jones wanted a new attorney and that Jones wished to be returned to the jail rather than sit in the holding cell. The court stated that it saw “no reason to force Mr. Jones to listen to the proceedings” and that Jones had rejected “the most viable option,” which was “to have Mr. Jones be able to listen to the proceedings, participate to a limited extent.” The court also noted that after his removal the previous day, Jones could be heard in the courtroom “kicking or pounding walls,” so that “it might be better” for him not to be present. But the court stated again that “well treat the future proceedings on a day-by-day basis.” The record is unclear as to whether Jones returned to the holding cell or chose to remain at the jail for the remainder of the trial. At any rate, his counsel never again raised the question of his return to the courtroom, nor did the judge, and Jones did not return to the courtroom until his sentencing. See State v. Jones, 916 A.2d at 30-31. He was sentenced to sixty-five years in prison. IV. Subsequent Procedural History Jones appealed his conviction to the Connecticut Supreme Court and argued that the trial court had erred in initially excluding him and then in keeping him out of the courtroom, that he was improperly denied his right to proceed pro se, and that the trial court had erred in rejecting his argument that his confession had been obtained illegally. The Connecticut Supreme Court rejected these arguments and affirmed his conviction, State v. Jones, 281 Conn. 613, 916 A.2d 17 (2007), and the United States Supreme Court denied certiorari, Jones v. Connecticut, 552 U.S. 868, 128 S.Ct. 164, 169 L.Ed.2d 112 (2007). Jones then sought state habeas relief, claiming that he had been denied effective assistance of counsel. Following an evidentiary hearing, the state court denied the writ. Jones v. Comm’r of Corr., No. CV044001435, 2007 WL 1976664 (Conn.Super.Ct. June 15, 2007). The state Appellate Court summarily dismissed his appeal from that denial, Jones v. Comm’r of Corr., 115 Conn.App. 902, 971 A.2d 97 (2009), and the state high court denied a petition for certification to appeal, Jones v. Comm’r of Corr., 293 Conn. 911, 978 A.2d 1109 (2009). Jones, proceeding pro se, then sought federal habeas corpus review before the *234 United States District Court for the District of Connecticut, arguing that his constitutional rights were violated by his absence from the trial, the admission of his confession, and the denial of his request to proceed pro se. The district court denied the petition and denied a certificate of appealability. Jones v. Murphy, No. 3:10— CV-49(JCH), 2010 WL 3829129 (D.Conn. Sept. 21, 2010). A panel of this Court, however, granted a certificate of appealability on July 13, 2011, as to four issues: (1) whether, before removing the Appellant from the courtroom, the trial court should have warned the Appellant that he would be removed if he continued his disruptive behavior; (2) whether the Appellant “reclaimed” his right to be present at trial. See Illinois v. Allen, 397 U.S. 337, 343 [90 S.Ct. 1057, 25 L.Ed.2d 353] (1970); (3) whether the Appellant was interrogated within the meaning of Rhode Island v. Innis, 446 U.S. 291 [100 S.Ct. 1682, 64 L.Ed.2d 297] (1980) and, (4) if so, whether his subsequent, post- Miranda statements are admissible consistent with Oregon v. Elstad, 470 U.S. 298 [105 S.Ct. 1285, 84 L.Ed.2d 222] (1985), and Missouri v. Seibert, 542 U.S. 600 [124 S.Ct. 2601, 159 L.Ed.2d 643] (2004)? DISCUSSION I. Standard of Review Under the Antiterrorism and Effective Death Penalty Act of 1996 (“AEDPA”), Pub.L. No. 104-132, 110 Stat. 1214 (1996), a state prisoner may seek federal collateral review of a state court conviction “only on the ground that he is in custody in violation of the Constitution or laws or treaties of the United States.” 28 U.S.C. § 2254(a). AEDPA “imposes a highly deferential standard for evaluating state-court rulings and demands that state-court decisions be given the benefit of the doubt.” Hardy v. Cross, — U.S.-, 132 S.Ct. 490, 491, 181 L.Ed.2d 468 (2011) (per curiam) (internal quotation marks omitted). When a petitioner “in custody pursuant to the judgment of a State court” advances “any claim that was adjudicated on the merits in State court proceedings” as a basis for federal habeas relief, a federal court may grant relief only if the “adjudication of the claim ... resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States,” 28 U.S.C. § 2254(d), (d)(1), or “resulted in a decision that was based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding,” id. § 2254(d)(2). As we have stated, under § 2254(d)(1), for a state court decision to be “contrary to,” or an “unreasonable application of,” that Supreme Court precedent, the decision must: (1) “arrive[ ] at a conclusion opposite to that reached by [the Supreme Court] on a question of law”; (2) “decide[] a case differently than [the Supreme Court] on a set of materially indistinguishable facts”; or (3) “identif[y] the correct governing legal principle ... but unreasonably appl[y] that principle to the facts of the prisoner’s case.” Portalatin v. Graham, 624 F.3d 69, 79 (2d Cir.2010) (alterations in original), quoting Williams v. Taylor, 529 U.S. 362, 412-13, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000) (opinion for the Court by O’Connor, J.). However, “[w]e apply AEDPA deference only if the state court has disposed of a claim on the merits.” Watson v. Greene, 640 F.3d 501, 508 n. 7 (2d Cir. 2011) (internal quotation marks omitted); see also, e.g., Cone v. Bell, 556 U.S. 449, *235 472, 129 S.Ct. 1769, 173 L.Ed.2d 701 (2009); Rompilla v. Beard, 545 U.S. 374, 390, 125 S.Ct. 2456, 162 L.Ed.2d 360 (2005). II. Exclusion Claims Jones argues that both his initial exclusion from the courtroom and the trial court’s subsequent refusal to allow him to return to the courtroom violated clearly established Supreme Court precedent under Illinois v. Allen, 397 U.S. 337, 90 S.Ct. 1057, 25 L.Ed.2d 353 (1970). The Connecticut Supreme Court adjudicated these questions on the merits. See State v. Jones, 916 A.2d at 32-34 (initial exclusion); id. at 34-37 (subsequent exclusion). Thus, our first task is determining whether either of the state court’s rulings “was contrary to, or involved an unreasonable application of,” Allen or other “clearly established Federal law, as determined by the Supreme Court of the United States,” 28 U.S.C. § 2254(d)(1), or whether either determination “was based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding,” id. § 2254(d)(2). A. Initial Exclusion The Connecticut Supreme Court held that the trial court’s initial order to exclude Jones did not violate Allen because Jones had voluntarily consented to removal. See State v. Jones, 916 A.2d at 32-34. We find that this holding rests on an “unreasonable determination of the facts” within the meaning of § 2254(d)(2), but that any Allen error in excluding Jones was nonetheless harmless. This claim therefore provides no basis for granting the writ. The Connecticut Supreme Court determined that the trial court had ordered Jones removed because Jones had requested to leave, and not because he was disruptive. Id. With respect, we find that this holding cannot reasonably be reconciled with the trial transcript. 1 On the morning of February 3, before the jury arrived, Jones’s counsel indicated that Jones disputed the court’s ruling of the previous day regarding Minnifield’s testimony. After the court declined to reconsider the ruling and Jones had a chance to consult with his counsel, counsel stated: Mr. Jones is indicating to me, Your Hon- or — and I’ve tried to explain to him that as a result of what he considered to be this — the prejudicial impact of Miss Minnifield’s testimony that as a result of that he would choose personally at this point not to go further with the proceedings. Counsel also indicated that he had told Jones that there was “no legal way to interrupt, continue, stop the proceedings.” Jones interrupted, stating, “Your Honor, I — I don’t even want to be here then.” The court then engaged Jones in a discussion about whether he actually wished to remove himself from the courtroom. The court warned Jones that “[t]he case is going to proceed, you know, without you,” that Jones was “going to be prejudiced by this,” and repeatedly asked if Jones understood what he was doing. The colloquy, which spans more than four transcript pages, culminated in the following ex *236 change, which we quote again in full due to its importance: THE COURT: Mr. Jones, you understand that you’re going to be prejudicing yourself by leaving today; you understand that? Do you understand that? THE DEFENDANT: Well, I want to— I’m going to put a motion in to dismiss counsel, then. THE COURT: No, it’s— THE DEFENDANT: For the record. THE COURT: No, Mr. Jones, we’re proceeding— THE DEFENDANT: And I’ll take my case myself. THE COURT: —we are proceeding— THE DEFENDANT: I’ll take my case myself, Your Honor. THE COURT: —that’s denied, Mr. Jones. THE DEFENDANT: I have a right, Your Honor. That’s my Constitutional right. THE COURT: All right. THE DEFENDANT: You’re laughing, but I’m serious. THE COURT: Marshals, remove Mr. Jones. A MARSHAL: Let’s go, Mr. Jones. THE DEFENDANT: I’m not going nowhere, man. A MARSHAL: Come on. THE DEFENDANT: No, I’m not going nowhere. A MARSHAL: Let’s go. THE DEFENDANT: Don’t touch me. The transcript then breaks off as a “scuffle ensued” in the courtroom. After order was restored, the trial court made a record that “Mr. Jones had to be physically restrained by a number of marshals.” Reviewing the transcript, the Connecticut Supreme Court found that “it was not the defendant’s conduct that precipitated his removal,” but rather that “the trial court ordered the defendant’s removal because the defendant had requested that he be permitted to leave the courtroom, which, although inadvisable, was the defendant’s right.” State v. Jones, 916 A.2d at 32. The high court noted that Jones “never explicitly stated that he had changed his mind about absenting himself from the courtroom,” and observed in a footnote that “it does not appear that the trial court perceived the defendant’s request to dismiss counsel and represent himself to be a sincere one.” Id. at 33 & n. 17. Respectfully, we find the Connecticut Supreme Comb’s finding of fact unreasonable. 2 When Jones asserted that he would like to leave, the trial court did not accede to that request. Rather, it engaged him in a discussion of the consequences of a decision to absent himself from the trial. That was entirely proper, and indeed necessary, since a waiver of the right to be present at trial, “as the waiver of any constitutional right in a criminal proceeding, must be knowing and voluntary.” Polizzi v. United States, 926 F.2d 1311, 1319 (2d Cir.1991); see also United States v. Tureseo, 566 F.3d 77, 83 (2d Cir.2009) (at least in federal cases, “[t]o establish waiver, the District Court must conduct a record inquiry to determine whether the defendant’s absence was ‘knowing and voluntary’ ”). A criminal de *237 fendant, who is not an expert in criminal procedure or constitutional law, must generally be advised of the consequences of waiving his rights, and be found by the court to have made a knowing and voluntary waiver, before being permitted to waive such an important right as presence at trial. See, e.g., Johnson v. Zerbst, 304 U.S. 458, 464, 58 S.Ct. 1019, 82 L.Ed. 1461 (1938) (“[C]ourts indulge every reasonable presumption against waiver of fundamental constitutional rights and [ ] we do not presume acquiescence in the loss of fundamental rights. A waiver is ordinarily an intentional relinquishment or abandonment of a known right or privilege.”) (internal quotation marks and citation omitted, emphasis added); cf. Brookhart v. Janis, 384 U.S. 1, 4, 86 S.Ct. 1245, 16 L.Ed.2d 314 (1966) (citing Zerbst and holding that a waiver of constitutional confrontation rights must be knowing and voluntary); see also Schriro v. Landrigan, 550 U.S. 465, 484-87, 127 S.Ct. 1933, 167 L.Ed.2d 836 (2007) (Stevens, J., dissenting) (collecting cases holding that waiver must be knowing and voluntary). 3 If the trial court had concluded its colloquy, and Jones had still wished to leave the courtroom after understanding the consequences, the state supreme court’s reading of the situation would be correct. But the transcript indicates that the colloquy instead caused Jones to rethink his decision to leave the trial. Indeed, reconsideration was the intended effect of the colloquy. The trial court was correct that leaving would have prejudiced Jones by, for example, preventing him from participating in his own defense, and the entire point of advising a defendant about the consequences of a waiver is to give him the opportunity to decide whether, in light of those consequences, he persists in his desire to waive. Once the warnings had their intended effect, Jones indicated he wanted to fire his attorney and represent himself, a desire plainly inconsistent with waiving his presence in court. At a minimum, the record does not indicate that Jones had made any final decision to waive his right to be present at trial at the time the court ordered the marshals to “remove Mr. Jones.” Although the Connecticut Supreme Court correctly noted that Jones “never explicitly stated that he had changed his mind about absenting himself from the courtroom,” State v. Jones, 916 A.2d at 33, that observation cannot reasonably support a finding that Jones’s departure was voluntary. Indeed, to require the defendant expressly to revoke his earlier request, at least on the facts presented here, puts the burden in the wrong place. For Jones’s attempted waiver to be valid, the trial court was required to make sure that he understood the consequences of his decision, and that his waiver was knowing and intelligent. Polizzi, 926 F.2d at 1319; Zerbst, 304 U.S. at 464, 58 S.Ct. 1019. The trial court clearly understood this obligation, and proceeded to ask Jones whether he did indeed understand what he was doing. Since there is no indication that Jones’s initial request was knowing and intelligent, a voluntary waiver could not be sustained unless Jones reiterated his request after he was properly advised by the court. He never made such a request. To the contrary, Jones’s words and actions made clear that he did not wish to leave the courtroom. Jones’s comment about proceeding pro se, whether or not it *238 was a “serious” request that could not be summarily denied without further inquiry, see Faretta v. California, 422 U.S. 806, 95 S.Ct. 2525, 45 L.Ed.2d 562 (1975), at a minimum indicated that Jones had rethought his decision to leave the courtroom. And certainly, the circumstances surrounding his exit do not show a man voluntarily departing. The immediate cause of his exit was the trial court’s order to the marshals, the execution of which Jones violently resisted. We therefore hold that the state supreme court’s decision was based on an unreasonable determination of the facts. 4 That Jones did not expressly waive his right to be present, however, does not dispose of the case. In addition to express waiver, a defendant may constructively waive his rights to be present at trial by disruptive behavior. Allen, 397 U.S. at 343, 90 S.Ct. 1057. But the record here presents difficulties in assessing any contention that Jones was properly removed on that basis. The transcript shows a defendant who insisted on speaking personally with the court despite being represented by counsel and who persisted in arguing with the court about its rulings— behavior that, while contentious and improper, would not in itself warrant the extreme response of involuntary exclusion. Nevertheless, caution is appropriate in assessing the trial judge’s response to the interaction. Absent specific record findings by the judge about what occurred in the courtroom, a cold transcript provides no insight into tone of voice, body language, or possible overtly threatening behavior that might cast mere spoken words in a different light. The actual situation facing a judge in the real world is not limited to the words that a court reporter can transcribe. Cf Harris v. Kuhlmann, 346 F.3d 330, 354 (2d Cir.2003). In any event, Jones’s violent resistance to the court’s removal order also supported a decision to continue the exclusion at least until he could comport himself appropriately, which purpose is implicit in the trial judge’s decision to declare a three-hour recess. It is not necessary, however, for us to decide whether AEDPA deference applies to our review of the trial court’s decision to exclude Jones 5 or *239 whether the exclusion was proper, because any error in the initial decision to remove Jones from the courtroom was harmless. 6 See Lucky, 569 F.3d at 108 (stating that harmless error applies to erroneous exclusion from the courtroom); cf. Spears v. Greiner, 459 F.3d 200, 204 (2d Cir.2006) (declining to decide whether to apply AEDPA deference because claim failed even without deference). Under the Sixth Amendment’s Confrontation Clause, a defendant has the right to be present at trial to confront the witnesses against him. See, e.g., Faretta, 422 U.S. at 816, 95 S.Ct. 2525; Allen, 397 U.S. at 338, 90 S.Ct. 1057. This right extends as a matter of due process to “critical stages” of the trial beyond the presentation of evidence when the defendant’s “presence has a relation, reasonably substantial, to the fulness of his opportunity to defend against the charge.” Kentucky v. Stincer, 482 U.S. 730, 745, 107 S.Ct. 2658, 96 L.Ed.2d 631 (1987) (internal quotation marks omitted); accord Tureseo, 566 F.3d 77 at 83. No further trial proceedings occurred on February 3 after Jones was excluded. Immediately after the exclusion, the trial court ordered a three-hour recess to permit Jones to cool off and discuss his situation with defense counsel. After a colloquy with both the prosecution and the defense about how to proceed in light of Jones’s behavior, the trial court decided to dismiss the jury for the rest of the day and resume proceedings on the following day. The minimal proceedings conducted in Jones’s absence on February 3 did not bear on Jones’s “opportunity to defend against the charge,” Stincer, 482 U.S. at 745, 107 S.Ct. 2658, but only on the entirely collateral issue of how the court should deal with Jones’s own dangerous and disruptive behavior. This was not a critical stage of the trial: the jury did not hear evidence, no motions were argued, jurors were not selected. Any argument that Jones had a right to be present during the discussion of the consequences of his own violent and disruptive behavior would be circular, and would imply that a court could never exclude a defendant under Allen. In sum, Jones’s exclusion for the remainder of February 3 was not an exclusion from a critical stage and did not prejudice Jones. Jones’s initial exclusion thus *240 provides no basis for federal habeas corpus relief. B. Subsequent Exclusion Jones was also absent from the courtroom on February 4. We must therefore decide whether this continued exclusion was proper, particularly in light of any efforts by Jones to reclaim his right to be present. 7 The Connecticut Supreme Court held that even if Jones did attempt to return to the courtroom, the trial court’s subsequent exclusion of Jones on February 4 was not an abuse of its discretion under Allen. We hold that the high court’s conclusion did not represent an unreasonable application of Allen. Even if the trial court erroneously removed Jones on February 3, Jones’s own violent and disruptive actions thereafter prevented him from returning on February 4. Cf. Norde v. Keane, 294 F.3d 401, 413 (2d Cir.2002) (“The fact that [the § 2254 petitioner’s] conduct may have been based on what he believed to be a compelling reason ... does not excuse his misconduct. Allen makes clear that a defendant does not have the right to disrupt the trial proceedings.”). Although Allen permits a court to find that a defendant has constructively waived his right to be present at his own trial, the Supreme Court also expressly made clear that “[o]nce lost, the right to be present can, of course, be reclaimed as soon as the defendant is willing to conduct himself consistently with the decorum and respect inherent in the concept of courts and judicial proceedings.” 397 U.S. at 343, 90 S.Ct. 1057. Thus, Jones is correct that once the trial court excluded Jones on February 3, it was required to permit him to return — but only if Jones satisfactorily demonstrated that he would not be violent or disruptive. The record shows not only that Jones failed to do so, but also that he engaged in further violent behavior that independently supported his exclusion from the courtroom, without regard to whether the initial removal was proper. On the morning of February 4, the court specifically cited Allen and noted that it believed it was permissible to exclude Jones based on his intervening misconduct. But as the Connecticut Supreme Court noted, the trial court was initially ready to permit Jones’s return on the morning of February 4. See State v. Jones, 916 A.2d at 35. However, after assessing Jones in person, a marshal reported to the court that Jones was being “somewhat confrontational toward” the marshals, and that in his “better judgment” the marshal “[could not] say that the defendant should be present in court during the proceedings.” The marshal also said that he could not “guarantee that a possible outburst won’t happen again” and that he did not “believe that Mr. Jones should be in these proceedings while they’re going on for the safety of everybody involved.” In discussion with the court, the marshal confirmed that Jones was agitated and was “still talking about yesterday.” In light of the earlier events and the marshal’s assessment, the court then “reconsider[ed]” its ruling and found that Jones, “by his disruptive behavior, [had] waived any right to be present during the proceedings.” The court reiterated its concerns that bringing Jones back would be unsafe and disruptive. The Connecticut Supreme Court found that the trial court did not abuse its discretion in excluding Jones, and we cannot find that conclusion unreasonable. *241 8 Initially, we are mindful of the Supreme Court’s guidance that “evaluating whether a rule application was unreasonable [under AEDPA] requires considering the rule’s specificity,” and that “[t]he more general the rule, the more leeway courts have in reaching outcomes in case-by-case determinations.” Yarborough v. Alvarado, 541 U.S. 652, 664, 124 S.Ct. 2140, 158 L.Ed.2d 938 (2004). Although the right to be present at one’s own trial is a right of paramount importance, appellate courts— let alone courts considering a case on collateral review many years later — lack the direct perception of the situation that informs the trial court’s judgment. This distance from courtroom realities explains why we review decisions to exclude a defendant for abuse of discretion. When that fact-specific standard of review is viewed through the additionally “deferential lens of § 2254(d),” Knowles v. Mirzayance, 556 U.S. 111, 121 n. 2, 129 S.Ct. 1411, 173 L.Ed.2d 251 (2009), the bar to relief is a high one. See Renico v. Lett, — U.S. -, 130 S.Ct. 1855, 1865, 176 L.Ed.2d 678 (2010) (noting the “dual layers of deference required by AEDPA” and an underlying standard asking “whether the [trial] judge exercised sound discretion”). Applying that deferential standard to this case, we conclude that the state high court reasonably determined that the trial court had a valid basis to exclude Jones on February 4, even if he wished to return. The Connecticut Supreme Court offered three reasonable considerations to support its conclusion. See State v. Jones, 916 A.2d at 35-36. First, and most importantly, it noted that Jones had reacted with serious physical violence to the court’s previous exclusion order on February 3. Id. at 35. In addition to its personal observations of Jones’s violence on the previous day, the trial court was also aware that after a previous adverse pretrial ruling, Jones had punched his hand through a Plexiglas window. Though Jones had been placed in restraints on February 4, we do not think the trial court’s concern for the safety of the lawyers, jurors, and witnesses was misguided. Second, the Connecticut Supreme Court noted that Jones had demonstrated no ability to control his temper and seemed unwilling to comply with court rulings or orders from the marshals. Id. at 35-36. This conclusion was supported by the marshal’s observation that Jones remained agitated and “confrontational,” and by the marshal’s concern that he could not guarantee the safety of courtroom personnel. Third, the Connecticut Supreme Court noted that the alternative — permitting Jones in the courtroom only in full-body restraints — presented its own difficulties. Id. at 36. As the Allen Court held, the “sight of shackles and gags might have a significant effect on the jury’s feelings about the defendant.” 397 U.S. at 344, 90 S.Ct. 1057. Moreover, contrary to Jones’s argument in our Court, it was not clear that Jones had consented to being shackled; his attorney had reported to the court that Jones ignored the “issue of restraints” in his discussions with counsel that morning. *242 We conclude that the Connecticut high court’s decision was reasonable under § 2254(d). The trial court was faced with a violent and unpredictable defendant. Putting aside the evidence that he had violently assaulted Minnifield and shot Williams to death, Jones had reacted violently (albeit outside the courtroom) to one adverse ruling, and had in the court’s presence violently resisted the removal order, requiring numerous marshals to restrain him, at least one of whom was injured in the fray. He then continued his obstreperous behavior once removed, threatening the marshals and “kicking or pounding [the] walls.” Jones manifestly waived his right to be present based on the extraordinary violence he had displayed during the preceding removal. Even if the court had erred in ordering that Jones be removed from the courtroom, Jones’s proper recourse was to comply with the order and seek his return through legal argument, not to offer violent resistance. Cf. Norde, 294 F.3d at 413 (“The fact that [the § 2254 petitioner’s] conduct may have been based on what he believed to be a compelling reason ... does not excuse his misconduct.”). All of the behavior related above was either known to, or had been personally observed by, the trial judge. Moreover, the court was entitled to rely on the marshal’s assessment that, whatever Jones might say, he remained volatile and confrontational, and continued to present a danger to persons in the courtroom. On this record, we cannot say that the Connecticut Supreme Court unreasonably applied Allen in affirming the trial court. 9 The case on which Jones principally relies, United States v. Ward, 598 F.3d 1054 (8th Cir.2010), is not to the contrary. During pretrial proceedings in that case, the defendant expressed a wish to speak for himself and speak with his lawyer aloud, rather than communicate with the court through counsel and with counsel through written notes, as the court had asked him to do. Id. at 1057. The court ordered him involuntarily removed. Id. He was then excluded for the rest of his trial because defense counsel could not guarantee to the court that the defendant would remain quiet. Id. The Eighth Circuit vacated the conviction and remanded for a new trial, finding problematic both the defendant’s initial exclusion and the trial court’s reliance on the representations of counsel, rather than the defendant himself, about the defendant’s willingness to comport *243 himself properly and reclaim his right to be present. Id. at 1058-60. The facts of Ward differ significantly from this case. Critically, the Ward court noted that [u]nlike many reported exclusion cases, including Allen ..., Ward’s presence at trial did not pose an apparent risk of physical injury to anyone. For example, he did not threaten the judge or anyone else involved in the trial, he had not assaulted anyone outside the courtroom, and he was not charged with a crime of violence. A trial judge with a legitimate concern for safety in the courtroom faces a very different situation and clearly has discretion to take firm action. Id. at 1059. By contrast, Jones, who was charged with murder, was violent both within the courtroom and without, had injured a marshal in resisting compliance with a court order, and had expressly threatened further violence. Thus, the trial judge here had precisely the “legitimate concern for safety in the courtroom” that was absent in Ward. Furthermore, although the Ward court was disturbed by the trial court’s reliance on conversations with defense counsel rather than the defendant himself, the trial court in this case relied on the marshal’s judgment about courtroom safety. Given the roles of judges and marshals, we have held that a trial judge may rely on a marshal’s judgment about safety. See United States v. Zuber, 118 F.3d 101, 103 (2d Cir.1997) (rejecting, “as a matter of law, the contention that the district court erred in deferring to the recommendation of the Marshals Service on the need to restrain the defendant at his sentencing hearing,” and distinguishing Allen). Finally, Ward was decided on direct review, not under the deferential AEDPA standard. Nothing in the Supreme Court’s decision in Allen clearly establishes a rule that a trial judge must converse a minimum number of times in person with a removed criminal defendant regarding the right to return before continuing the exclusion. Allen dictates only that a defendant be allowed to return once “willing to conduct himself consistently with the decorum and respect inherent in the concept of courts and judicial proceedings.” 397 U.S. at 343, 90 S.Ct. 1057. It was not unreasonable for Connecticut’s Supreme Court to conclude that the trial court acted within its discretion in finding that, through his words and conduct, Jones exhibited no such willingness. 10 *244 The trial court may (or may not) have acted precipitously in ordering Jones removed from the courtroom on February 3. In any event, it would have been preferable for the trial court to have recalled Jones to the courtroom, under restraint if necessary, outside the presence of the jury, on February 4 to instruct Jones about the standard of behavior that would be expected of him, question him about his commitment to comply with that standard, and make its own assessment, with due regard to the marshal’s opinion, about the reliability of whatever assurances Jones may have offered. We cannot conclude, however, that Jones suffered any prejudice from his removal on February 3, or that the Connecticut Supreme Court unreasonably determined that his continued exclusion on February 4, in light of his extreme and violent actions the day before, was consistent with United States Supreme Court precedent. 11 The district court thus did not err in denying the writ. III. Miranda Claims Jones also argues that the state courts failed to recognize that his interrogation violated Miranda v. Arizona, 384 U.S. 436, 86 S.Ct. 1602, 16 L.Ed.2d 694 (1966), and its progeny. He makes two distinct claims. First, he argues that because he was subjected to conduct that amounted to interrogation without receiving a Miranda warning, his subsequent statements were rendered involuntary and should not have been admitted under Rhode Island v. Innis, 446 U.S. 291, 100 S.Ct. 1682, 64 L.Ed.2d 297 (1980), and Oregon v. Elstad, 470 U.S. 298, 105 S.Ct. 1285, 84 L.Ed.2d 222 (1985). Second, he argues that the police employed a deliberate two-step procedure to circumvent Miranda, in violation of Missouri v. Seibert, 542 U.S. 600, 124 S.Ct. 2601, 159 L.Ed.2d 643 (2004). Neither claim provides a basis for a grant of the writ: The first misunderstands the relevant Supreme Court law, and the second was never presented to any state court. A brief summary of the pertinent facts is required. See State v. Jones, 916 A.2d at 39-41. In a pretrial suppression hearing, the government presented evidence that after his arrest, Jones had been *245 placed in an interview room with a Detective Stevenson, who was assigned to prevent Jones from leaving because the door did not lock. Stevenson was not directed to take any statement or ask any questions of Jones. Because it was lunchtime, Stevenson ordered food for both men, and they had an informal and wide-ranging conversation. As stated by the Connecticut Supreme Court: Stevenson testified that, shortly after lunch, however, the defendant spontaneously stated something “along the lines of, I didn’t kill anybody.” According to Stevenson, he ignored the comment, but the defendant stated shortly thereafter, “I know you guys think it’s about the girl.” Stevenson then asked the defendant if he had known the victim. The defendant responded by placing his head in his hands and stating, “[H]e did not deserve to have happen what I did to him.” Stevenson left the room to inform [his superior officer] O’Leary about the defendant’s statement. O’Leary asked Stevenson whether the defendant had been advised of his rights and, upon learning that he had not, instructed Stevenson to do so immediately and then to ask the defendant if he would be willing to speak to Stevenson about the victim’s murder. Id. at 40. Thereafter, Stevenson returned and read Jones his rights. After Stevenson asked Jones if he understood these rights, Jones replied that he did and signed a card indicating that he had been advised of his rights. Jones then confessed to the murder. After verifying again that he understood his lights, Jones also dictated a written confession to Stevenson, which Jones signed. A. Innis and Elstad Arguments Jones argues first that the apparently informal discussion he had with Detective Stevenson over lunch was actually an interrogation within the meaning of Miranda. He relies on Innis, in which the Supreme Court held that the Miranda safeguards come into play whenever a person in custody is subjected to either express questioning or its functional equivalent. That is to say, the term “interrogation” under Miranda refers not only to express questioning, but also to any words or actions on the part of the police (other than those normally attendant to arrest and custody) that the police should know are reasonably likely to elicit an incriminating response from the suspect. 446 U.S. at 300-01, 100 S.Ct. 1682 (footnote omitted). Whether or not any of Stevenson’s conversation with Jones preceding Jones’s apparently spontaneous statements that he didn’t kill anyone and that the police thought “it’s about the girl,” we can assume for purposes of the argument that Stevenson’s direct question to Jones whether he knew the victim Williams, which elicited the more incriminating statement that Williams “did not deserve ... what I did to him,” constituted interrogation. None of these statements, however, were offered against Jones at trial. Rather, the prosecutor offered into evidence only the full oral and written confessions that were made after Stevenson advised Jones of his rights and began a formal interview. In Elstad, the Supreme Court held that even if a defendant is questioned without Miranda warnings, rendering any statement made during such questioning coerced and inadmissible, that violation does not automatically taint subsequent statements made after he is advised of his rights. 470 U.S. at 318, 105 S.Ct. 1285. “The relevant inquiry is whether, in fact, the second statement was also voluntarily *246 made.” Id. Thus, even assuming arguendo that the pre-warning police conduct was “interrogation,” and that Jones’s pre-warning statements were therefore not admissible, Jones’s post-warning statements — -the only ones received in evidence — were admissible unless they were involuntary. Here, the facts as reasonably found by the Connecticut Supreme Court — and which Jones does not contest in this proceeding — demonstrate that Jones’s confession was voluntary. See State v. Jones, 916 A.2d at 39-41. Indeed, the Connecticut Supreme Court noted that Jones “does not dispute that the evidence adduced by the state, if credited, was sufficient to establish that he confessed to the murder and revealed the whereabouts of the murder weapon only after a knowing and voluntary waiver of his rights.” Id. at 42. According to the facts as found by the state courts, Detective Stevenson advised Jones of his rights by reading aloud from a card and asked Jones if he understood; Jones replied that he did. Id. at 40. After making an oral confession, but before making a written one, Jones signed a card advising him of his rights, and two police witnesses verified that Jones had “read aloud from a voluntary statement rights form and initialled] each line.” Id. at JO-41. Later, Jones told the police that his brother would lead them to the murder weapon; after his brother initially refused to cooperate, Jones spoke to him on the phone, and his brother led officers to the weapon. Id. at 41. It is true that Jones disputed this version of events at his suppression hearing, but the court found the officers more credible, and the Connecticut Supreme Court found no clear error in that finding. Id. at 42. Even if Jones attempted to challenge this credibility determination, under the substantially more deferential standard of § 2254(d)(2) and (e), 12 we cannot conclude that the findings of the state courts were unreasonable. B. Seibert Claim Jones attempts to raise a second claim under Missouri v. Seibert. That case created an exception to the Elstad rule, discussed above, by holding that police may not use a deliberate two-step procedure to circumvent Miranda. See Seibert, 542 U.S. at 621, 124 S.Ct. 2601 (Kennedy, J., concurring in the judgment) (“When an interrogator uses [a] deliberate, two-step strategy, predicated upon violating Miranda during an extended interview, postwarning statements that are related to the substance of prewarning statements must be excluded absent specific, curative steps.”). 13 Jones contends that the police employed such a deliberate two-step procedure here, rendering his post-warning confessions inadmissible. Because Jones did not present this argument to the state courts, however, this Court may not consider it. 14 Under AEDPA, a prisoner in custody pursuant to a state court judgment must generally exhaust state court *247 remedies before seeking federal habeas corpus review. See 28 U.S.C. § 2254(b)(1); see also Carvajal v. Artus, 633 F.3d 95, 104-05 (2d Cir.2011). Moreover, “[i]f a habeas applicant fails to exhaust state remedies by failing to adequately present his federal claim to the state courts so that the state courts would deem the claim procedurally barred, we must deem the claim procedurally defaulted.” Carvajal, 633 F.3d at 104 (internal quotation marks and brackets omitted); see also Aparicio v. Artuz, 269 F.3d 78, 90 (2d Cir.2001) (“[W]hen the petitioner failed to exhaust state remedies and the court to which the petitioner would be required to present his claims in order to meet the exhaustion requirement would now find the claims procedurally barred, federal habeas courts also must deem the claims procedurally defaulted.” (internal quotation marks omitted)). “An applicant seeking habeas relief may escape dismissal on the merits of a procedurally defaulted claim only by demonstrating ‘cause for the default and prejudice’ or by showing that he is ‘actually innocent’ of the crime for which he was convicted.” Carvajal, 633 F.3d at 104, quoting Aparicio, 269 F.3d at 90. Jones makes several arguments that his Seibert claim is not barred. The first and most substantial is that he presented essentially the same claim in his Connecticut Supreme Court appeal. Jones notes that “citing chapter and verse of the Constitution” is not necessary to exhaust a claim before the state courts. Daye v. Att’y Gen. of State of N.Y., 696 F.2d 186, 194 (2d Cir.1982) (en banc). Jones argues that his Seibert claim was “fairly presented,” Picard v. Connor, 404 U.S. 270, 275, 278, 92 S.Ct. 509, 30 L.Ed.2d 438 (1971), to the state courts when Jones made other, similar arguments about his Miranda rights. We are not persuaded. In the Connecticut Supreme Court, Jones did argue that his confession was involuntary. But his brief did not so much as cite Seibert, nor did he in any way articulate the core of a Seibert claim — that the police had employed a deliberate two-step procedure to circumvent Miranda. Instead, Jones contended that the trial court had erred in its factual determinations, and, citing State v. Finder, 250 Conn. 385, 736 A.2d 857, 878 (1999), requested that the high court conduct its own “scrupulous examination of the record.” Furthermore, Jones’s failure to alert the government and the state courts to this claim is not a mere technicality. This is not a case in which the habeas petitioner has simply applied a slightly different label to what is essentially the same claim. Adjudicating a Seibert claim would require factual findings that were never made by the trial court, based on evidence that was never adduced in the state courts — for example, evidence of the subjective intent of the officers. See United States v. Williams, 681 F.3d 35, 43 (2d Cir.2012) (stating that in assessing a Seibert claim, “a court should review the totality of the objective and subjective evidence surrounding the interrogations” (internal quotation marks omitted)). Thus, Jones’s failure to raise the claim prevented the development of a proper record for assessing it. Connecticut law is clear that “rights of constitutional magnitude may be waived,” State v. Paige, 304 Conn. 426, 40 A.3d 279, 284 (2012), and the Connecticut Supreme Court has declined to consider similar arguments in similar procedural postures. Cf. State v. Mullins, 288 Conn. 345, 952 A.2d 784, 795-97 (2008) (declining to consider claim that confession was coerced because defendant had not objected below and record was inadequate to review the claim). We are confident that *248 “the court to which the petitioner would be required to present his claims in order to meet the exhaustion requirement would now find the claims procedurally barred.” Aparicio, 269 F.3d at 90 (internal quotation marks omitted). Because “the state courts would deem the claim procedurally barred, we must deem the claim procedurally defaulted.” Carvajal, 633 F.3d at 104 (internal quotation marks and brackets omitted). 15 Jones makes two additional arguments that his Seibert claim is preserved, but neither is persuasive. First, he notes that although he did not raise the claim in his initial § 2254 petition below, he attempted to make the claim in a reply brief that the district court rejected as untimely. This argument is irrelevant — it addresses a separate default, namely Jones’s failure to raise the claim below, in the federal district court, to preserve it for this federal appeal. Second, Jones notes that this Court’s certificate of appealability mentioned Seibert. But the purpose of a certificate of appealability is only to provide jurisdiction for appeal, see 28 U.S.C. § 2253(c), not to determine the merits of claims or investigate whether those claims were properly presented in state proceedings. Finally, Jones does not attempt to show cause and prejudice or actual innocence, which are necessary to raise a defaulted claim. See Carvajal, 633 F.3d at 104. Thus, we cannot reach Jones’s argument that the police interrogation violated Seibert. CONCLUSION For the foregoing reasons, the judgment of the district court denying the petition for a writ of habeas corpus is AFFIRMED. POOLER, Circuit Judge, concurring in part and dissenting in part: I join the majority opinion fully as to Parts II.A and Part III. I respectfully dissent as to Part II.B, denying Jones habeas relief for his unwarned and continued exclusion from trial. The majority concludes that the state court’s decision that this exclusion did not deprive Jones of his right to presence was a reasonable application of clearly established Supreme Court law. I disagree. Background “ ‘A leading principle that pervades the entire law of criminal procedure is that, after indictment found, nothing shall be done in the absence of the prisoner.’ ” United States v. Canady, 126 F.3d 352, 360 (2d Cir.1997) (quoting Lewis v. United States, 146 U.S. 370, 372, 13 S.Ct. 136, 36 L.Ed. 1011 (1892)). The right to presence is “scarcely less important to the accused than the right of trial itself.” Diaz v. United States, 223 U.S. 442, 455, 32 S.Ct. 250, 56 L.Ed. 500 (1912). It is “rooted to a large extent in the Confrontation Clause of the Sixth Amendment,” United States v. Gagnon, 470 U.S. 522, 526, 105 S.Ct. 1482, 84 L.Ed.2d 486 (1985) (per curiam), but *249 due process also protects the right of a defendant to “be present in his own person whenever his presence has a relation, reasonably substantial, to the fullness of his opportunity to defend against the charge,” Snyder v. Massachusetts, 291 U.S. 97, 105-06, 54 S.Ct. 330, 78 L.Ed. 674 (1934), overruled in part on other grounds by Malloy v. Hogan, 378 U.S. 1, 2, 84 S.Ct. 1489, 12 L.Ed.2d 653 (1964). The right to presence, crucial though it is, can be constructively waived by misbehavior in the courtroom. In Illinois v. Allen, the Supreme Court case which guides all analysis in this area, the Court considered the appropriateness of removing a defendant from the courtroom who had “argue[d] with the judge in a most abusive and disrespectful manner”; threatened the judge by telling him “ ‘When I go out for lunchtime, you’re (the judge) going to be a corpse here’ ”; and t[ore] up “the file which his attorney had and threw the papers on the floor.” 397 U.S. 337, 339-340, 90 S.Ct. 1057, 25 L.Ed.2d 353 (1970). The defendant continually interrupted voir dire, badgered prospective jurors and was generally obstreperous. Id. The judge warned him that after another outbreak, he would be removed. The defendant continued to make “abusive remarks” and was, indeed, removed from the courtroom. The next day the defendant was returned to the courtroom. The judge admitted him, but told him he would be only “would be permitted to remain in the courtroom if he ‘behaved (himself) and (did) not interfere with the introduction of the case.’” Id. In spite of this second warning, the defendant again acted disruptively. Id. at 340-341, 90 S.Ct. 1057. He was again removed; during the intermittent time he was present, he “responded to one of the judge’s questions with vile and abusive language.” Id. at 341, 90 S.Ct. 1057. The judge still told him that “he could return to the court room whenever he agreed to conduct himself properly.” Id. When the defendant eventually gave such assurances, and ultimately complied with them, he “was permitted to be present through the remainder of the trial, principally his defense.” Id. In considering whether such “deplorable” removals, id. at 347, 90 S.Ct. 1057, were constitutionally sound, the Court explicitly h[e]ld that a defendant can lose his right to be present at trial if, after he has been warned by the judge that he will be removed if he continues his disruptive behavior, he nevertheless insists on conducting himself in a manner so disorderly, disruptive, and disrespectful of the court that his trial cannot be carried on with him in the courtroom. Once lost, the right to be present can, of course, be reclaimed as soon as the defendant is willing to conduct himself consistently with the decorum and respect inherent in the concept of courts and judicial proceedings. Id. at 343, 90 S.Ct. 1057. Allen thus sets forth two requirements for a constitutional removal from his trial. First, it requires that a defendant be warned before he is removed. Id. Second, it requires that a defendant who has previously lost the right to presence be able to reclaim it. These holdings bound the Connecticut Supreme Court in analyzing Jones’s case, and now, in habeas, they bind us. Analysis I. Allen Warnings This troubling case implicates both of Allen’s major holdings: the necessity of warning and reclamation of right. Taking first the question of warning, the state court found that no warning was required before Jones’s removal because he was taken from the courtroom only upon his “own request that he be permitted to leave.” Jones, 281 Conn. at 638-39, 916 *250 A.2d 17. I am in full agreement with the majority opinion that such a conclusion based on the record was unreasonable, as the record unmistakably reflects that Jones was not originally removed pursuant to his request. Maj. Op. 236-37. The majority concludes that because the exclusion itself on February 3rd was-harmless, further warning analysis is obviated. Maj. Op. 239 fn. 6. I agree that Jones’s removal that day was harmless, but even assuming the error in failure to warn on that day was subsumed by the harmlessness of the exclusion itself, the majority inexplicably fails to account for the fact that Jones was excluded again, without warning, on February 4 (and indeed, every other day of the trial). If warning is required before a defendant loses his right to be present, then a warning was required before Jones was deemed to have waived his right to be present in the courtroom the day after his initial wrongful exclusion. This, of course, raises the question of whether warning is required — or to be more precise, whether it has been clearly established by the Supreme Court that a defendant must be warned that he may lose his right to be present if he fails to comport himself in an appropriate manner. There is no doubt to my mind that it has. In Allen the Court “explicitly” held that a defendant can lose his right to be present at trial if, after he has been warned by the judge that he will be removed if he continues his disruptive behavior, he nevertheless insists on conducting himself in a manner so disorderly, disruptive, and disrespectful of the court that his trial cannot be carried on with him in the courtroom. 397 U.S. at 343, 90 S.Ct. 1057. This passage is completely unambiguous about what Allen requires in terms of warning. There is no doubt about its force (“we explicitly hold”) or what is required from it (“can lose his right to be present at trial, if, after he has been warned, ” (emphasis added)). A defendant can lose his right to presence if he has been warned about the consequences of inappropriate behavior. Jones never was. I am compelled, of course, to admit that our court has not always read Allen this way. In two cases, Gilchrist v. O’Keefe, 260 F.3d 87, 97 (2d Cir.2001) and Norde v. Keane, 294 F.3d 401, 413 (2d Cir.2002), we indicated that Allen does not require a warning before removal. In Gilchrist, we examined a state court’s determination that a defendant could “forfeit[ ] his right to counsel based on a single, apparently spontaneous violent incident,” without warning that such behavior might light to a loss of counsel. 260 F.3d at 97. We considered Allen in relationship to the warning question, and found that “while [it] stated that a defendant could be removed from the courtroom ‘after he has been warned by the judge,’ it did not indicate whether such a warning was a requirement in every situation.” Id. at 96. We concluded that Allen “stand[s] for the proposition” that “even absent a warning, a defendant may be found to have forfeited certain trial-related constitutional rights based on certain types of misconduct,” id., and so denied the petitioner relief. In Norde v. Keane, we cited to the “even absent warning” language in Gilchrist and upheld a removal where the court did not “explain to [the defendant] the potential ramifications of his removal” for misbehavior. 294 F.3d at 413. The judge in Norde did, however, warn the defendant that “he would be removed if he did not remain quiet.” Id. As an initial matter, these cases do not bind us in habeas. “ ‘Clearly established federal law’ refers only to the holdings of the Supreme Court.” Rodriguez v. Miller, 537 F.3d 102, 106 (2d Cir.2008) (citing *251 Williams v. Taylor, 529 U.S. 362, 412, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000)). Moreover, to the extremely limited extent they are relevant, those precedents are themselves an unreasonable application of Supreme Court law. 1 “[T]he warning requirement from Allen cannot be interpreted in any non-mandatory way, lest we substitute our own judgment of what the rule should be for that of the Court.” Gray v. Moore, 520 F.3d 616, 624 (6th Cir.), cert. denied by Gray v. Moore, 555 U.S. 894, 129 S.Ct. 216, 172 L.Ed.2d 162 (2008). There is no way to read the Supreme Court’s holding in Allen as anything other than “requiring] a trial court to give the accused one last chance to comply with courtroom civility before committing the ‘deplorable’ act — in the Allen Court’s words — of removing that person from his own trial.” Id. Moreover, neither Norde nor Gilchrist’s view of Allen constitutes their holdings. The language in Gilchrist is dicta, since the question before the court was not whether warnings were required before a defendant was removed for disruptive behavior, but whether the right to counsel could be forfeited absent warning. Norde is also distinguishable by the crucial fact that the defendant there actually received the only warning — that misbehavior would lead to removal — that Allen requires. The majority opinion itself recognizes to some extent that warning is required prior to removal, Maj. Op. 237 fn. 3, though it fails to apply that warning to second day of Jones’s exclusion. 2 Under the applicable Supreme Court precedent, a warning was required before Jones was not allowed to return to court on the second day. The failure to warn is particularly disturbing in the circumstances of this case. The record reflects that Jones’s behavior on February 4th sprung in part from his anger at being excluded from court the previous day, as well as his distress at the court’s (appropriate) refusal to let him proceed pro se without further discussion. The marshal told the court that Jones was “still talking about yesterday” and that he was “reliving the past.” The marshal further indicated that Jones did not think that he had done anything wrong the day before and believed that he had been treated unjustly. I do not disagree with the majority that “[t]he fact that [Jones’s] conduct may have been based on what he believed to be a compelling reason ... does not excuse his misconduct,” Norde 294 F.3d at 413, Maj. Op. 240, though I note that we are not confronting Jones’s mere belief that he had a compelling claim — this court apparently agrees with him that his initial removal was wrongful. But what concerns me is not whether Jones’s behavior could be excused in light of the previous day’s violation, but whether he had a right to be told that if he did not behave he would continue *252 to be excluded. Under Allen, he clearly did. 3 Allen warnings were not only required, but especially important in this case. Jones, who wanted to attend court and whose misbehavior was occasioned in part by the court’s previous initially wrongful removal, might have heeded the court’s warning that he had to behave appropriately or risk further exclusion, had one been given. Moreover, given that Jones’s initial removal was precipitated only by his “insist[ing] on speaking personally with the court despite being represented by counsel and ... [his] persistence] in arguing with the court about its rulings,” Maj. Op. 238, Jones may not have made the connection between his violent behavior and his exclusion. From his perspective, the court was liable to remove him simply for insisting — correctly—that he had a right to represent himself. Maj. Op. 235-36. If he had been warned, however, about what sort of behavior actually warranted removal, he might have understood what was required of him to retain his right to presence, and the “deplorable” act of “removing] a man from his own trial,” Allen, 397 U.S. at 347, 90 S.Ct. 1057, avoided. Warnings are not ancillary under Allen’s framework; they are crucial. First, as just noted, warnings help prevent the loss of the right to presence where it can be avoided. Allen itself tells us that “courts must indulge every reasonable presumption against the loss of constitutional rights,” Id. at 343, 90 S.Ct. 1057 (citing Johnson v. Zerbst, 304 U.S. 458, 464, 58 S.Ct. 1019, 82 L.Ed. 1461 (1938)). Second, and most importantly, warnings ensure a defendant’s implied waiver by misconduct is both knowing and voluntary. As the majority notes a waiver of the right to be present at trial, “as the waiver of any constitutional right in a criminal proceeding, must be knowing and voluntary.” Polizzi v. United States, 926 F.2d 1311, 1319 (2d Cir.1991); see also United States v. Lucky, 569 F.3d 101, 107-08 (2d Cir. 2009). A criminal defendant, who is not an expert in criminal procedure or constitutional law, must generally be advised of the consequences of waiving his rights, and be found by the court to have made a knowing and voluntary waiver, before being permitted to waive such an important right as presence at trial. Maj. Op. 235-36. Allen’s warning requirement and its reliance on Zerbst, which established the knowing and voluntary waiver standard, tell us that even an implied waiver of the right to presence based on conduct must be knowing and voluntary — a contention with which the majority agrees. Maj. Op. 237 fn. 3. See also Brookhart v. Janis, 384 U.S. 1, 4, 86 S.Ct. 1245, 16 L.Ed.2d 314 (1966) (“There is a presumption against the waiver of constitutional rights, and for a waiver to be *253 effective it must be clearly established that there was ‘an intentional relinquishment or abandonment of a known right or privilege.’ ” (internal citation omitted)); United States v. Benabe, 654 F.3d 753, 768-69 (7th Cir.2011) (noting that while a defendant can implicitly waive his right to presence based on his conduct, under Allen, such a waiver must still be knowing and voluntary) (quoting Allen, 397 U.S. at 343, 90 S.Ct. 1057); Proffitt v. Wainwright, 685 F.2d 1227, 1258 fn. 45 (11th Cir.1982) (noting Allen “supports retention of the knowing-and-voluntary waiver standard in right-to-presence cases”). If a defendant is warned that misconduct will lead to exclusion, and engages it in anyway, he has waived his right to presence knowingly and voluntarily. Not so with a defendant who has not been warned — especially one, like Jones, who was initially excluded wrongfully, and who cannot have been expected to connect his behavior to the loss of the right. Under Allen, as well as Zerbst and the Supreme Court’s continued jurisprudence on knowing and voluntary waiver of rights, a defendant must be warned before he is excluded from his trial. 4 He must be given an opportunity to comport himself appropriately, and more importantly, given the opportunity to do so with the knowledge that a failure to behave will lead to the loss of his right to be present. Without warning, we cannot say Jones’s misbehavior either on February 3rd in the courtroom or the next day with the marshal constituted a knowing waiver of his right to be present. The Supreme Court has clearly established that the right to presence cannot be implicitly waived without warning. A defendant edging towards or having already crossing the line of excludable behavior must be given a chance to step back — or if he fails to do so, knowingly and voluntarily face the consequences of his actions. Jones faced the consequences of his misdeeds without ever having been warned about what they would be. Continued trial in his absence based on the lack of warning alone was a violation of his right to be present and a violation of clearly established Supreme Court law. II. Reclamation Jones’s continued exclusion from trial also implicates Allen’s dictate that “[o]nee lost, the right to be present can ... be reclaimed as soon as the defendant is willing to conduct himself consistently with the decorum and respect inherent in the concept of courts and judicial proceedings.” 397 U.S. at 343, 90 S.Ct. 1057. 5 *254 The Connecticut Supreme Court found no error in the trial court’s refusal to let Jones come back to court on February 4, the day after this initial removal. The court wrote that it acknowledged that Jones originally had refused to return to court in full body restraints but thereafter agreed to do so. In other words, as between banishment from the courtroom and returning to the courtroom fully restrained, the defendant expressed a preference for the latter. Although it may be true that a court normally should defer to the preference of the accused in this regard, we are not persuaded that the court was required to do so in the present case. First, the court understandably was concerned that, due to the defendant’s demonstrated propensity for violence, he posed a particular danger to the personnel responsible for transporting him to and from the courtroom. For those persons, even attempting to outfit the defendant with body restraints gave rise to a safety risk. Second, the court reasonably concluded that, under all of the circumstances, including the defendant’s volatile temper, his habitual unwillingness to accept adverse court rulings, his confrontational attitude toward the marshals and his lack of contrition for his prior violent behavior, the defendant was very likely to engage in disruptive and obstreperous conduct if permitted to return to the courtroom. Finally, as the trial court observed, the need for and use of visible restraints undoubtedly would have prejudiced the defendant in the eyes of the jurors. In light of the relevant considerations, in particular, the nature and severity of the defendant’s misconduct, ■ his refusal to acknowledge the impropriety of that misconduct and his stubbornly antagonistic attitude, we cannot say that the trial court acted unreasonably in concluding that the defendant should not be allowed to return to the courtroom. Jones, 281 Conn, at 642-44, 916 A.2d 17. The majority finds each of these conclusions reasonable applications of Allen. I disagree entirely. Taking first the conclusion that the “nature and severity of’ Jones’s misbehavior allowed not just for his removal, but his permanent exclusion from the courtroom, I turn back to Allen, which presented similar facts. The Allen defendant was extremely disruptive, even before the jury, and continued to be despite multiple warnings and removals. He also threatened the judge and made “abusive” and “vile” comments throughout the proceedings. And yet in considering his case the Supreme Court still held that a defendant — previously engaged in vile, abusive and threatening behavior — who promises to comport himself appropriately can reclaim his right to be present. Such a conclusion is unsurprising. A defendant who is attempting to reclaim his right to presence is, by definition, a person who has' previously been so disruptive so as to have been removed. It would not only be circular to say that past disruption itself prevents reclamation, but also a perversion of Allen, the holding of which would be made meaningless in light of this circular logic. That Jones’s disruptions involved violent behavior does not change this conclusion. The defendant in Allen was, if not violent, *255 threatening, and yet he still was able to reclaim his right. Even if that were not so, Allen does not limit its holding to any particular class of misbehavior. The Court did not hold there that a defendant can reclaim his right unless he commits a violent act or unless the court has “concernís] for [] safety.” Maj. Op. 241. Rather, it tells us unequivocally that “[o]nce lost, the right to be present can ... be reclaimed as soon as the defendant is willing to conduct himself consistently with the decorum and respect inherent in the concept of courts and judicial proceedings.” 397 U.S. at 343, 90 S.Ct. 1057. Neither the Connecticut Supreme Court, nor we, have the right or the power to read in a limitation where none exists. To the extent Jones’s violent behavior was relevant, I am troubled by some of the considerations made by the majority in its analysis. First, the majority notes that the trial court was aware of the fact that “after a previous adverse pretrial ruling, Jones had punched his hand through a Plexiglas window.” Maj. Op. 241. But that incident took place outside of the courtroom, and thus is completely irrelevant to whether Jones could be excluded from his trial. Allen, and the power it gives to courts to conduct proceedings in a defendant’s absence, is about decorum in the courtroom, not about a defendant’s behavior outside of it. The Allen court framed the “[t]he question presented [in that ease] ... [a]s whether an accused can claim the benefit of this constitutional right to remain in the courtroom while at the same time he engages in speech and conduct which is so noisy, disorderly, and disruptive that it is exceedingly difficult or wholly impossible to carry on the trial.” 397 U.S. at 338, 90 S.Ct. 1057 (emphasis added). Jones’s violent out-of-court conduct cannot be excused nor condoned, and there are many punishments which could have been constitutionally imposed in response. But exclusion from trial is simply not one of them. Exclusion under Allen is simply not a punishment at all; it is a merely a means of preventing defendants from disrupting trial proceedings, not a way of punishing them if they do. Likewise, I am concerned by the majority’s suggestion that Jones’s violent nature and the crimes, which he was at the time accused and is now convicted of committing, are relevant to the court’s decision to exclude him. The majority writes that [t]he trial court was faced with a violent defendant. Putting aside the evidence that he had violently assaulted Minnifield and shot Williams to death, Jones had reacted violently (albeit outside the courtroom) to one adverse ruling, and had in the court’s presence violently resisted the removal order, requiring numerous marshals to restrain him, at least one of whom was injured in the fray ... Maj. Op. 241-42. See also Maj. Op. 242-43 (“Jones, who was charged with murder, was violent both within the courtroom and without, had injured a marshal in resisting compliance with a court order, and had expressly threatened further violence.”). The majority is almost certainly right that Jones was a violent person, and probably even a bad one. But the right to presence belongs to the peaceful and the pugnacious, as well as the murderous and the mad. This conclusion should warrant no remark: Since anyone exercising his right to presence is, by default, a person accused of a crime, the right would mean little if being accused, or even guilty of a crime, could undermine it. Nothing in Allen even remotely suggests that a person’s character or their crimes are appropriate to consider when determining whether he can reclaim his right to be present. Contrary to the majority’s suggestion, Allen does not limit, qualify or otherwise constrain *256 the right to reclaim to those people who are not inherently violent or who have not been arrested for violent crimes. No matter what Jones had done, no matter what his character, he had a right to be present during his trial. Even if we appropriately limit our analysis to Jones’s violent behavior inside the courtroom, the majority still misreads Allen by suggesting that Jones could and forever waive his right to return to court based on the incident of February 3. Maj. Op. 242 (“Jones manifestly waived his right to be present based on the extraordinary violence he had displayed during the preceding removal.”). But this is simply without any support from Allen. While Allen certainly supports the notion that a defendant can waive his right to be present based on violent behavior, in no sense does it support the proposition that there are some acts which might forever disqualify a person from attending their trial. The whole of Allen’s framework of warning and reclamation cuts against any such argument, and nothing in its holding on reclamation is in any way qualified or limited. Perhaps there are some acts so heinous that once committed by a defendant in court they have forever lost their right to presence, but if so, the Supreme Court has yet to tell us what they are. That is not to say that safety in the courtroom is an irrelevant consideration; of course, it is not. But that is why the Supreme Court has dictated alternative, constitutional means of ensuring that safety — namely shackles and gags. The majority finds the state court did not err in crediting the trial court’s refusal to allow Jones into the courtroom, even bound and gagged, but I fail to see how this constitutes a reasonable application of Allen. It is right, as the majority notes, to say that Allen indicates “that the sight of shackles and gags might have a significant effect on the jury’s feelings about the defendant.” 397 U.S. at 344, 90 S.Ct. 1057. But Allen also held that “bind[ing] and gag[ging] [the defendant], thereby keeping him present” was a constitutional means for a court “handle an obstreperous defendant.” Id. at 343-44, 90 S.Ct. 1057. The Court additionally noted that “binding and gagging might po[s]sibly be the fairest and most reasonable way” to handle such a defendant. Id. at 344, 90 S.Ct. 1057. Thus, while binding and gagging are certainly problematic, they are constitutional. What is not constitutional is removal of a defendant without warning and without allowing him to reclaim once he has made the required promise. It cannot fairly be said that it was more in accord with Allen and the constitutional right to presence to have wrongfully excluded Jones, failed to warn him that continued misbehavior would lead to further exclusion, and finally refused to allow him to return after he promised to behave appropriately, than to have allowed him to attend his trial in shackles. On this point, I note that the majority’s conclusion that “it was not clear that Jones had consented to being shackled,” Maj. Op. 241, is unwarranted by the trial record. While Jones’s counsel remarked that Jones “didn’t address the issue of restraints” in his discussion with counsel, the Marshal indicated afterwards that Jones was “in agreement that he [Jones] will still in full restraints.” Even after the marshal changed his assessment, he indicated that the conversations with Jones took place “in the process of putting the restraints on him.” Though the marshal ultimately concluded that Jones should not be brought to court, he never indicated that Jones had refused in any way to be shackled, or that it was his assessment that it was too dangerous for his deputies to attempt to do so. I find the majority’s analysis to be thus factually and legally wrong. *257 This leads to the state court’s error in confusing disruption with likelihood of disruption and the majority’s adoption of that error. The Connecticut Supreme Court wrote that the [trial] court reasonably concluded that, under all of the circumstances, including the defendant’s volatile temper, his habitual unwillingness to accept adverse court rulings, his confrontational attitude toward the marshals and his lack of contrition for his prior violent behavior, the defendant was very likely to engage in disruptive and obstreperous conduct if permitted to return to the courtroom.’ Jones, 281 Conn, at 643, 916 A.2d 17. See also Maj. Op. 241. This statement, and the trial record it relies upon, constitutes an unreasonable reading of Allen, which gives absolutely no succor to the proposition that the defendant’s likelihood of reoffending is sufficient to deny him the chance to try to behave appropriately. The majority does not argue that Jones’s behavior with the marshal was itself sufficient to exclude him; indeed, it would be hard for such a conclusion to be warranted, given that the strongest statement the marshal made about Jones’s behavior in the moment was that he was being “somewhat confrontational.” The marshal’s view was clearly not that Jones was behaving too improperly to be brought into court, which would present a different, and perhaps easier, question, but rather that he was likely to misbehave if brought back. The marshal told the court that his “feeling [was] [Jones] just wants to get in the courtroom to cause further discomfort,” and that he couldn’t “guarantee that a possible outburst won’t happen again. I think it’s just a matter of time before one does occur again ...” The trial court relied fully on this future likelihood: “I’ll also find that, based on the [marshal’s] assessment ... were [the defendant] allowed into the courtroom, it is very likely that he will cause additional disruptions before the jury ... possibly resulting] in further injuries to either court personnel, the jury, witnesses or spectators for that matter.” (emphasis added). I do not disagree that Jones was likely to continue misbehaving. He may even have been especially likely to do so in light of the fact that he was never warned that such behavior would lead to his exclusion. But the defendant in Allen, who was continually disruptive, even after two removals from the courtroom and repeated warnings, could also be said to be likely to reoffend. He, if anyone, was likely to “cause additional disruptions before the jury,” but the Supreme Court still told us he could reclaim his right on his promise to comport himself appropriately. The mere fact that we are considering a question of reclamation means we are considering a defendant who has already been so disruptive that he has been removed from court. Such a person is almost by definition likely to reoffend, or, in the words of the majority, be a person with “no ability to control his temper and [an] unwilling[ness] to comply with court rulings or orders from the marshals.” Maj. Op. 241-42. Yet Allen counsels us that even such a person is entitled upon his promise to behave to return to court, despite the likelihood that will almost certainly exist in any such case that he will not do so. There is simply no support in Allen for the proposition that a defendant can either be excluded in the first instance or not allowed to reclaim his right to presence based on what he is likely to do. I further note that the state court’s conclusion that exclusion was warranted because Jones “refus[ed] to acknowledge the impropriety of that misconduct and his *258 stubbornly antagonistic attitude” is completely erroneous. In the first instance, Jones was never given the chance to “acknowledge the impropriety of his misconduct,” at least before the court, since he was never returned to court after the initial wrongful removal. And Allen in no sense supports the notion that “a stubbornly antagonist attitude” undercuts the right to reclaim; it firmly stands against it, given the defendant in that case’s behavior. Apologies or regret are not necessary under Alien', only a promise to behave is required. That is perhaps the most crucial point the majority opinion elides. Allen refers to the “constitutionally permissible” means for a trial judge to deal with a disorderly defendant as “tak[ing] him out of the courtroom until he promises to conduct himself properly.” 397 U.S. at 344, 90 5.Ct. 1057 (emphasis added). 6 This language clearly posits the promise as the trigger point for the defendant’s ability to return. Once the promise is made, the defendant can return, unless and until he demonstrates that he will not conduct himself properly. That promise was made in this case, because although Jones was never given the opportunity to make the required promise to the trial judge himself, or to do so formally, he gave the marshal “his word” that there would be “no further outbursts.” Once this was done, the state court was obligated to give Jones himself the chance to show that he was “willing to conduct himself consistently with the decorum and respect inherent in the concept of courts and judicial proceedings,” id. at 343, 90 S.Ct. 1057. Only if and when he failed to do so could he be constitutionally removed from court. Conclusion The clearly established law of the Supreme Court is that before a defendant can be excluded from his trial, he must be warned “that he will be removed if he continues his disruptive behavior.” Allen, 397 U.S. at 343, 90 S.Ct. 1057. Even after he has been warned and removed, he can reclaim his right to presence upon his promise to behave, without regard to his character or to his crimes. Id. The Allen defendant was volatile, confrontational, destructive, and threatening, and he continued to be despite two removals and repeated warnings. The Supreme Court held that he was entitled to a second chance at exercising his fundamental rights. So too was Jones, and failure to comprehend that constitutes an unreasonable application of clearly established law. The sad history of this case represents failure upon failure to comport with Allen. Jones was wrongfully removed from the courtroom in the first instance. He was given no warning to make clear what was required of him and what the consequences would be if he failed to meet the court’s expectations. And even after promising to comport himself appropriately and complying with the application of full restraints, the court still excluded him. Jones’s trial was conducted almost completely in his absence. This history represents the sort of “extreme malfunction[ ],” Harrington v. Richter, — U.S.-, 131 S.Ct. 770, 786, 178 L.Ed.2d 624 (2011), that habeas is meant to guard against. The profound failures of the trial court and the *259 Connecticut Supreme Court to understand and properly apply Allen compel my respectful dissent. Footnotes [1] The trial transcript is the same evidence upon which the state supreme court relied. See 28 U.S.C. § 2254(d)(2) (providing that a writ "shall not be granted with respect to any claim that was adjudicated on the merits in State court proceedings unless the adjudication of the claim ... resulted in a decision that was based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding " (emphasis added)). [2] The Connecticut Supreme Court characterized its holding on this point as a finding of fact based on the transcript. See State v. Jones, 916 A.2d at 32 ("We reject this claim because it is predicated on a faulty factual premise. The record reflects that the trial court did not warn the defendant that he would be removed from the courtroom if he persisted in misbehaving for the simple reason that it was not the defendant’s conduct that precipitated his removal.”). [3] Indeed, even a constructive waiver of the right to be present through disruptive behavior must in effect be knowing; in Allen, the Supreme Court held that a defendant must ordinarily be "warned by the judge that he will be removed if he continues his disruptive behavior” before he may be removed. 397 U.S. at 343, 90 S.Ct. 1057. [4] The Supreme Court has not yet clarified the precise relationship between § 2254(d)(2), which provides that relief is only available if the state court decision "was based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding," and § 2254(e)(1), which provides that "a determination of a factual issue made by a State court shall be presumed to be correct.” See Wood v. Allen, 558 U.S. 290, 130 S.Ct. 841, 848-49, 175 L.Ed.2d 738 (2010) (declining to reach the question). Our Circuit has also not yet decided the question. See Green v. Travis, 414 F.3d 288, 298 n. 6 (2d Cir.2005), citing Channer v. Brooks, 320 F.3d 188, 194 (2d Cir.2003). We assume, without deciding, that § 2254(d)(2)'s unreasonableness standard is the proper one for review of this claim. We may so assume because even if "presumed to be correct” consistent with § 2254(e)(1), the state court’s conclusion cannot be supported by the transcript, which we find to be "clear and convincing evidence,” 28 U.S.C. § 2254(e)(1), contrary to the state court's conclusion. Moreover, as we also hold, any error is harmless. Thus, the standard under which the state court's findings are reviewed does not decide this case. [5] It is not clear that the trial court’s decision would receive heightened AEDPA deference, since it was not the final state adjudication of the exclusion claim on the merits. See Stenhouse v. Hobbs, 631 F.3d 888, 894-95 (8th Cir.2011) ("Our court has not directly addressed whether, if the reasoning of the state appellate court cannot pass muster under AEDPA, the rationale of the state trial court also merits deference under § 2254(d).”), cert. denied, - U.S. -, 132 S.Ct. 308, 181 L.Ed.2d 189 (2011). The usual rule, as stated by the Ninth Circuit, is that "[wjhen more *239 than one state court has adjudicated a claim, we analyze the last reasoned decision,” Barker v. Fleming, 423 F.3d 1085, 1091 (9th Cir. 2005), and also defer only to that last reasoned decision, see id. at 1092-93 (declining to consider an intermediate state appellate court’s decision). See also Brian R. Means, Postconviction Remedies § 29:4 (2011); cf. Greene v. Fisher, — U.S. --, 132 S.Ct. 38, 44-45, 181 L.Ed.2d 336 (2011) (for retroactivity purposes, relevant state decision is the ultimate merits determination by a state court). We have previously discussed and deferred to both a state trial court’s initial decision and an appellate court’s affirmance thereof, see McKinney v. Artuz, 326 F.3d 87, 89, 100-01 (2d Cir.2003), but that was in a case where, under the Ninth Circuit’s taxonomy, "the last reasoned decision adopted or substantially incorporated the reasoning from a previous decision and, as a result, it was reasonable for the reviewing court to look at both decisions to fully ascertain the reasoning of the last decision,” Barker, 423 F.3d at 1093. Nonetheless, because of our harmless error determination, we do not decide this question today. We note, however, that even on direct review, “a district court's determination of whether the defendant waived his right to be present at a particular stage of trial is reviewed for abuse of discretion.” Tureseo, 566 F.3d at 83. Thus, even without AEDPA deference, we would review the trial court’s decision here under a deferential standard. [6] Our finding of harmless error obviates the need to decide the first question in the certificate of appealability, i.e., whether Jones should have been warned before he was excluded. [7] This is the second question referred to in the certificate of appealability. [8] The Connecticut Supreme Court’s decision is the “last reasoned decision,” Barker, 423 F.3d at 1091, on the question of Jones’s subsequent exclusion, and we therefore review whether the high court’s affirmance of the trial court was unreasonable. Of course, the Connecticut Supreme Court’s ruling necessarily depended in large part on facts found and judgments made in the first instance by the trial court. Thus, the instant case presents a scenario in which "the last reasoned decision adopted or substantially incorporated the reasoning from a previous decision and, as a result, it [is] reasonable for the reviewing court to look at both decisions to fully ascertain the reasoning of the last decision,” id. at 1093, even while we formally review only the final determination. See note 5, supra. [9] In her thoughtful dissent, Judge Pooler argues that the Supreme Court’s statement in Allen that a disruptive defendant can be removed from the courtroom if he continues to misbehave after being warned creates an absolute requirement of a warning. As she acknowledges, however, this Court has already held that it does not. Norde, 294 F.3d at 413; Gilchrist v. O'Keefe, 260 F.3d 87, 97 (2d Cir. 2001). We respectfully disagree with her contention that these cases do not bind us in habeas. First, these cases were themselves habeas cases. Second, the habeas statute provides that the writ may only be granted when the state court's 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). It most certainly does not say that Circuit precedent is irrelevant in deciding that the state court’s decision was not such an "unreasonable application.” Indeed, it is difficult to see how we could hold that the Connecticut courts unreasonably applied Supreme Court precedent in not finding a warning an absolute prerequisite to continued exclusion from the courtroom, when we ourselves have read Allen exactly the same way. Finally, we agree with the Gilchrist panel's conclusion that Allen need not be read to create an absolute warning requirement. As that Court correctly stated, “while Allen stated that a defendant could be removed from the courtroom ‘after he has been warned by the judge,’ it did not indicate whether such a warning was a requirement in every situation.” 260 F.3d at 96. [10] Judge Pooler takes a different view of the situation, contending that in the circumstances of this case, Jones’s behavior did not warrant the conclusion that he was unwilling to conduct himself properly in the courtroom. We do not suggest that her analysis is unreasonable, and, at least based on the cold record available to us, we agree that a trial judge could well have taken the view advanced by our dissenting colleague. But see supra at 238 (recognizing the limits of a cold record in depicting the situation faced by the trial judge). In any event, on habeas review, it is not for us to decide how we might have assessed the situation. We must instead decide whether the state courts’ assessment was an unreasonable one in light of Supreme Court precedent. Moreover, like any reviewing court, we must give appropriate deference to the discretion of the trial judge, who was faced not with a cold record, but with firsthand experience of a volatile defendant. We do disagree, however, with several of Judge Pooler’s conclusions. For example, in assessing whether a defendant who has already engaged in violent behavior in the courtroom will be able to conduct himself properly in the future, we do not believe it is "completely irrelevant” that the defendant had very recently engaged in explosive, property-damaging and self-injuring conduct in response to a judicial ruling, even if that particular outburst took place outside the courtroom. Cf. Dissent at 254-55. And while we of course agree that the right to presence extends to all defendants, whatever crimes they are charged with, the issue here is the *244 reasonableness of the trial court's assessment of whether Jones was willing and able to "conduct himself consistently with the decorum and respect” necessary in judicial proceedings, Allen, 397 U.S. at 343, 90 S.Ct. 1057, and indeed whether the safety of courtroom personnel could be assured. The court was fully aware from the suppression hearing of substantial evidence that Jones had not merely committed a serious violent crime, but had engaged in a course of impulsive and volatile violent behavior. Such evidence was hardly irrelevant to the assessment the court had to make. Finally, we cannot agree that the passing comment in Allen, italicized in the dissent, that a disruptive defendant may be removed from the courtroom "until he promises to conduct himself properly,” Dissent at 258, quoting Allen, 397 U.S. at 344, 90 S.Ct. 1057, establishes the extraordinary principle that a defendant’s mere verbal pledge to behave — a promise that Jones's attorney never even suggested that Jones had proffered — triggers an absolute right to return to the courtroom, regardless of the evidence that his word cannot be relied upon, or of the record of prior misconduct he has compiled. Certainly Allen, which affirmed a defendant’s exclusion from the courtroom, cannot be read to have held any such thing. See Rodriguez v. Miller, 537 F.3d 102, 106 (2d Cir.2008) (only the Supreme Court's holdings, and not its dicta, constitutes "clearly established federal law” under § 2254(d)(1)), citing Taylor, 529 U.S. at 412, 120 S.Ct. 1495 (opinion for the Court by O'Connor, J.) and Carey v. Musladin, 549 U.S. 70, 74-77, 127 S.Ct. 649, 166 L.Ed.2d 482 (2006). [11] Although the record is unclear about the remainder of trial proceedings after February 4, Jones makes no further argument that he was improperly excluded after that point. [12] The precise relationship between § 2254(d)(2) and (e)(1) is an unresolved question. See note 4, supra. On this point, Jones would not succeed under either standard, so we need not decide the question. [13] Seibert did not elicit a majority opinion. However, our Court has clarified that Justice Kennedy’s opinion, which provided the fifth vote for the result in the case, is controlling. See United States v. Carter, 489 F.3d 528, 535-36 (2d Cir.2007) (joining several Circuits in adopting this view); see also United States v. Capers, 627 F.3d 470, 475-77 (2d Cir.2010) (recounting evolution of Seibert claims within this Circuit). [14] Indeed, Jones did not raise this claim in his habeas petition to the district court, alluding to it for the first time in an untimely reply brief below. [15] We note that Seibert had not been decided at the time of Jones’s pretrial suppression hearing or his trial. However, because Seibert was decided before Jones’s appeal to the Connecticut Supreme Court, the Seibert rule was retroactively applicable to his case under Griffith v. Kentucky, 479 U.S. 314, 107 S.Ct. 708, 93 L.Ed.2d 649 (1987). The Connecticut Supreme Court has established a test for review of constitutional errors not preserved at trial. See, e.g., State v. Fabricatore, 281 Conn. 469, 915 A.2d 872, 877 (2007) (discussing test); Mullins, 952 A.2d at 795-97. Thus, Jones could have raised his Seibert claim in the Connecticut Supreme Court. (Indeed, the government’s opposition brief in that proceeding cited Seibert in passing.) [1] The majority suggests "it is difficult to see how we could hold that the Connecticut courts unreasonably applied Supreme Court precedent in not finding a warning an absolute prerequisite to continued exclusion from the courtroom, when we ourselves have read Allen exactly the same way.” Maj. Op. 242 fn. 9. But such a position suggests that while state courts might unreasonably apply Supreme Court law, this court never could — a statement with which, deeply respectful of my colleagues though I am, I cannot agree. We, no less than state courts, can and do make mistakes, including unreasonable ones. [2] It also misstates the Court’s decision in that case as holding "that a defendant must ordinarily be 'warned by the judge that he will be removed if he continues his disruptive behavior' before he may be removed.” Maj. Op. 237 fn. 3 (emphasis added). The Court in Allen, it bears repeating, inserted no such qualifier, and instead "h[e]ld that a defendant can lose his right to be present at trial if, after he has been warned by the judge that he will be removed if he continues his disruptive behavior.” 397 U.S. at 343, 90 S.Ct. 1057. [3] Even if the majority is correct that there are situations in which a warning preceding removal is not required by Allen, nothing in this case after February 3 suggests it would be one. Nor does the majority even attempt to explain why this particular case so qualified. I concede that Jones’s behavior after the court’s initial wrongful removal of him may have necessitated an immediate emergency removal, and I do not read Allen to say that the judge was required to shout out a warning before safety could be restored to the courtroom. But once Jones was secured or the courtroom otherwise made safe, the court was required to warn the him before continuing to exclude him from the trial. Nothing in this case even remotely suggests that the court’s failure to warn Jones on the second day was necessitated by exigent circumstances. Accordingly, even if Allen does not, as the majority tells us, "create[ ] an absolute requirement of warning,” Maj. Op. 242 fn. 9, a warning was required here. [4] I note that, as a technical matter, Jones was not ‘removed” from the courtroom on February 4, but rather never allowed to enter. This distinction is completely irrelevant for Allen purposes, however, as the only question is whether a defendant must be warned before his right to be presence can be deemed waived and the trial continued in his absence. Whether the defendant is technically inside or outside of the courtroom at this point is of no moment, as the majority implicitly recognizes in finding that the marshal’s out-of-court assessment was sufficient for the court’s determination that Jones could not return. Moreover, any distinction between "removal” and "refusal to let enter” in this context would eviscerate Allen's warning requirement, for all a judge would need to do to avoid it is, as here, effectuate a wrongful, but harmless removal and then refuse to let a defendant reenter the courtroom. I also note that the events of February 4th could to some extent be characterized as a second exclusion, rather than a refusal to let reclaim, since the judge initially determined that Jones could reenter, and then changed his mind based on Jones’s apparent change in behavior. See infra fn. 3. [5] As noted supra fn. 4, the events of February 4 can be characterized to some extent as a second exclusion, not just a continuation of the first, which implicates whether the second *254 removal was itself warranted rather than whether the refusal to let Jones reclaim was. As noted infra, it would be hard to say that a second exclusion would be warranted based on Jones’s interaction with the marshal alone, but given, that Jones was never actually returned to court, and thus never regained die right to presence, the better analysis is probably the reclamation analysis undertaken by the majority. [6] The majority argues that this language is dicta because Allen “affirmed a defendant’s exclusion from the courtroom.” Maj. Op. 243-44 fh. 10. But Allen affirmed the defendant's removal in part because when he promised that he would behave, he was allowed to return — despite the "evidence his word [could not] be relied upon ... [and] the record of prior misconduct he ha[d] compiled,” id. This language about promise cannot then be passed off as dicta so easily.