dissenting.
A five-justice panel of the New York State Supreme Court, Appellate Division, Second Department, ruled that Guy Zap-pulla’s signed confession to the murder of Jennifer Scarpati was inadmissible because the police failed to re-administer the warnings required by Miranda v. Arizona, 384 U.S. 436, 86 S.Ct. 1602, 16 L.Ed.2d 694 (1966). People v. Zappulla, 282 A.D.2d 696, 697, 724 N.Y.S.2d 433, 434 (2d Dep’t 2001).5 Those same judges, however, unanimously concluded that the error was harmless beyond a reasonable doubt because the remaining evidence of Zappulla’s *476guilt was overwhelming. See id. at 698, 724 N.Y.S.2d at 435. Addressing this issue on Zappulla’s habeas corpus challenge to his state murder conviction, see 28 U.S.C. § 2254, District Judge Jack B. Weinstein also determined that the Miranda error in Zappulla’s case was harmless: “Under the circumstances of this ease and in light of the overwhelming evidence of petitioner’s guilt, there is no reasonable probability that the introduction of his un-Mirandized inculpatory statement had a substantial and injurious effect or influence in determining the jury’s verdict or that it resulted in actual prejudice to petitioner.” Zappulla v. New York, 296 F.Supp.2d 309, 319 (E.D.N.Y.2003) (applying “actual prejudice” standard of review articulated in Brecht v. Abrahamson, 507 U.S. 619, 637, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993)). Because I agree with these able judges’ assessment of the record in this case, I respectfully dissent from the ruling of my colleagues in the majority that it was “objectively unreasonable” for a court to conclude “beyond a reasonable doubt” that the erroneously admitted confession “did not contribute to the verdict obtained.” Supra at [474].
There is no question in this case that the Appellate Division applied the correct standard of harmless error review on direct appeal. See People v. Zappulla, 282 A.D.2d at 698, 724 N.Y.S.2d at 436 (citing People v. Crimmins, 36 N.Y.2d 230, 237, 367 N.Y.S.2d 213, 218, 326 N.E.2d 787 (1975), which expressly adopted the “harmless beyond a reasonable doubt” standard of direct review stated in Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967)). Accordingly, our ability to grant habeas relief is narrowly circumscribed by 28 U.S.C. § 2254(d)(1). Specifically, we may not grant a writ of habeas corpus simply because we conclude that the state court erred in finding admission of the un-Mir-andized confession harmless; “rather, ha-beas relief is appropriate only if the [state court] applied harmless-error review in an ‘objectively unreasonable’ manner.” Mitchell v. Esparza, 540 U.S. 12, 124 S.Ct. 7, 12, 157 L.Ed.2d 263 (2003); see also Lockyer v. Andrade, 538 U.S. 63, 75, 123 S.Ct. 1166, 155 L.Ed.2d 144 (2003) (rejecting the use of a “clear error” standard to assess objective unreasonableness because it “fails to give proper deference to state courts by conflating error (even clear error) with unreasonableness”).
As I.understand the majority opinion, its determination that the state court’s harmlessness conclusion was not simply wrong but “objectively unreasonable” turns on four factors: (1) the weakness of the prosecution’s case against Zappulla without the signed confession, (2) the prosecution’s heavy reliance on the signed confession in making its case to the jury, (3) the fact that the confession was not cumulative but of singular importance to the establishment of Zappulla’s motive for killing Scar-pati, and (4) the particular persuasive influence of a signed confession.6
*477For the reasons stated in this dissent, I am not convinced that the majority’s conclusions as to the first three factors are so compelled by the record that a contrary view by the state court must be deemed “objectively unreasonable.” As for the last factor, I note simply that at the same time that the Supreme Court in Arizona v. Fulminante, 499 U.S. 279, 111 S.Ct. 1246, 113 L.Ed.2d 302 (1991), made the point emphasized by the majority — that a “defendant’s own confession is probably the most probative and damaging evidence that can be admitted against him,” id. at 296, 111 S.Ct. 1246 — it ruled that even a coerced confession was subject to harmless error analysis and may be rendered innocuous if the remaining evidence against the defendant is so overwhelming as to leave no reasonable possibility that the tainted confession contributed to conviction, id. at 310-11, 111 S.Ct. 1246;7 see also Milton v. Wainwright, 407 U.S. 371, 92 S.Ct. 2174, 33 L.Ed.2d 1 (1972) (holding that admission of defendant’s confession was harmless beyond a reasonable doubt where the admissible evidence was so overwhelming as to render the confession cumulative). Indeed, Judge Weinstein fully recognized the prejudicial potential of an erroneously admitted confession but, nevertheless, concluded that the error in this case could reasonably be deemed harmless in light of overwhelming admissible evidence of Zap-pulla’s guilt. See Zappulla v. New York, 296 F.Supp.2d at 319.
1. The Overwhelming Admissible Evidence of Zappulla’s Guilt
Shortly after 6:00 p.m. on March 17, 1998, New York City police officers found Jennifer Scarpati’s dead body stuffed under the mattress and box spring of a bed in Room 234 of the Golden Gate Inn in Brooklyn, New York. She had been strangled to death, causing blood vessels to burst around her eyes and in her neck, and blood to flow from her nose and mouth. A blood-stained towel lay on the floor near her body, and a blood stain was also visible on the wallpaper at the level of the bed-frame.
In contrast to the majority, I conclude that, even without Zappulla’s challenged confession, his commission of this brutal crime was established by overwhelming evidence. See, e.g., United States v. Cas-taño, 999 F.2d 615, 618 (2d Cir.1993) (per curiam) (observing that the “strength of the prosecution’s case is probably the single most critical factor in determining whether error was harmless”). Specifically, (a) extensive testimonial and physical evidence established that Zappulla was with Scarpati at and about the time of her death in the room they shared at the Golden Gate Inn; (b) when Zappulla was arrested on unrelated charges in a hallway of that motel at 12:20 a.m. on March 17, 2004, Scarpati’s blood was on his jacket, sweat*478shirt, pants, and boots; and (c) while in custody, Zappulla admitted to a fellow inmate, Feliz Pabon, that he had killed Scar-pati, providing even more detail as to the commission of the crime than he had in his written confession.
(a) Evidence that Zappulla Was with Scarpati At and About, the Time of Her Death
Testimonial proof that Zappulla was with Scarpati at and about the time of her death was supplied by gypsy cab driver Aaron Cohen, who stated that he had brought Scarpati and Zappulla to the Golden Gate Inn after the trio, together with another trial witness, Bruce Brodsky, had spent part of the day on Brodsky’s boat smoking crack cocaine. Indeed, Cohen ád-mitted that he had rented the room for Scarpati and Zappulla to allow the couple to conceal their presence at the hotel. Apparently, Scarpati, who worked as a prostitute, had a particular interest in discretion: motel management had barred her from the premises. Over the course of the next twenty-four hours, Cohen occasionally joined Scarpati and Zappulla in the room to smoke crack. On one occasion, Zappul-la gave Cohen some jewelry to sell in order to purchase crack. At approximately 5:00 a.m. on March 16, 1998, Cohen called Room 234, asking to come over with another prostitute. Zappulla denied the request, saying he wished to sleep. When Cohen called the room at 2:00 p.m. the same day and spoke with Scarpati, she berated him for this request. Cohen again called the room sometime between 5:00 and 5:30 p.m. and asked about Scarpati; Zappulla told him that they had had a fight, and he did not know where she was.
Although the majority cites Cohen’s criminal record, drug abuse, and occasional use of an alias to dismiss him as an “unreliable” witness against Zappulla, supra at [468-69], it appears that this “unreliability” pertains primarily to Cohen’s account of the jewelry exchange, a point discussed more particularly at Part 3 of this dissent, and his report of a fight between Zappulla and Scarpati. His testimony that Zappulla was with the murder victim in Room 234, however, was so heavily corroborated that the majority characterizes this fact as essentially “undisputed.” Supra at [469].
Indeed, the principal piece of corroboration putting Zappulla in Room 234 could not be impeached: a key for that motel room was found in Zappulla’s pocket after his initial arrest on March 17.8 Further, not only was Zappulla arrested at the Golden Gate Inn, the murder site, motel surveillance cameras established his presence on the premises at various times in the interval between 9:00 p.m. on March 16, 1998, and 12:20 a.m. on March 17, 1998, which was consistent with the estimated time of Searpati’s death.
Certainly, these surveillance tapes, like the key, do not conclusively establish that Zappulla murdered Scarpati. What they do demonstrate, however, is his opportunity to commit the crime. The majority is not content to note this distinction. Instead, it suggests that the tapes actually demonstrate the unlikelihood of Zappulla’s being Scarpati’s murderer. Supra at [469]. Preliminarily, let me note that I do not understand this to be a case in which anyone asserts Zappulla’s actual innocence. To the contrary, given that the legal defect with his signed confession is a failure to re-administer Miranda warnings, with no evidence of compulsion, there can be no serious question in this court *479that Zappulla did, in fact, murder Jennifer Scarpati. Thus, the only issue before us is whether the erroneous admission of the un-Mircmdized confession could reasonably have been ruled harmless beyond a reasonable doubt. That being said, I will briefly explain why I do not share my colleagues’ view that the tapes favor Zap-pulla.
The majority states that for Zappulla to have murdered Scarpati, the crime “must have occurred prior to 11:00 p.m.” Supra at [469]. While I agree that Zappulla could have committed the murder before 11:00, it is also possible that he committed the crime sometime between his recorded 11:12 p.m. and 12:18 a.m. departures from the motel, depending on when he returned to the room, presumably through the unmonitored side door. The majority, however, concludes that a time line derived from the tapes makes these scenarios unlikely because
it seems odd that, after brutally murdering Jennifer in the heat of an argument, Zappulla would leave through the front door of the motel, knowing that the side door was an unmonitored exit ... return to the room through the side door, call several escort services while inside the room with Jennifer’s dead body, go back through the lobby and exit through the front door, and briefly converse with the front desk clerk. According to this time line, it is more likely that Zappulla acted this way because he was unaware that a murder had occurred in the motel room and that Jennifer’s dead body had been hidden under one of the beds.
Id. at [469-70]. The fact that murderers act in ways that strike law-abiding persons as “odd” is no new phenomenon. Indeed, in this case we know that Zappulla acted in just the “odd” way described by the majority and, nevertheless, killed Scarpati. Thus, I cannot fault the state court for failing to share the majority’s view that the surveillance evidence is favorable to Zappulla. To the contrary, I consider it entirely reasonable for the Appellate Division to have cited this evidence, together with Cohen’s testimony and Zappulla’s possession of the motel room key, as overwhelming proof, independent of the signed confession, that Zappulla was with Scarpa-ti at and about the time of her death. See People v. Zappulla, 282 A.D.2d at 698, 724 N.Y.S.2d at 435.
(b) Scarpati’s Blood on Zappulla’s Clothing on the Day of Her Death
A medical examiner testified that soon after her death, Jennifer Scarpati bled from the nose and mouth. This grim fact allowed her to provide some of the most powerful evidence against Zappulla, identifying him as her murderer with her own blood.
Specifically, Scarpati’s blood was found in ten different locations on the clothing Zappulla was wearing at the time of his March 17 arrest at the Golden Gate Inn: (1) on the right sleeve cuff of his sweatshirt; (2) on the left front of his sweatshirt, beneath the chest area; (3) in the right knee area of his jeans; (4) in the left knee area of his jeans, (5) on the back of the left knee area of his jeans, toward the calf; (6) underneath the left pocket of the jeans; (7) several stains along the instep of his right boot; (8) in a lug on the heel of his right boot; (9) on the back of his jacket, near the left shoulder blade; and (10) on the back of his jacket, further toward the bottom. See Trial Tr. at 426, 432. This evidence, together with the aforementioned proof that Zappulla was with Scarpati in the room where she died at and about the time of her death, would, by itself, permit a reasonable jury to conclude beyond a reasonable doubt that he was the murderer.
*480Without detailing the extent of this blood evidence, the majority dismisses it as inconclusive, noting that the record fails to indicate whether the blood stains were “innocuous, trace amounts” or “substantial.” Supra at [470]. In fact, trial testimony by both a police detective and a forensics examiner reveals that the stains were not innocuous; most were visible to the naked eye, even though Zappulla’s clothing was generally dark-colored. See Trial Tr. at 278, 435-36.
The majority further notes that no prosecution witness could testify as to how or when Scarpati’s blood came to be on Zap-pulla’s clothes. Pointing to evidence that Scarpati had “various cuts on her body in different stages of healing” and “wounds on her wrists from using needles,” supra at [469], it suggests that this offers an alternative innocent explanation for the blood on Zappulla’s clothes. It further hypothesizes that if Scarpati “had already been murdered and hidden under one of the beds when Zappulla returned to the motel” after 11:00 p.m. on March 16, “a certain amount of her blood may have gotten on his clothes and boots without him realizing this.” Supra at [470].
Given that the needle marks were described as “pinpoints,” Trial Tr. at 332, and the cuts as “healing abrasions,” id. at 344, I respectfully suggest that it strains common sense to think that they explain the number of blood stains on Zappulla’s clothing or their myriad locations, spanning Zappulla’s sleeve, shirt front, pants legs, jacket back, boot instep, and even a lug on his boot heel. Even more implausible is the suggestion that Zappulla could have received so many blood stains in so many different locations on his clothing without realizing that he was coming into contact with blood or that Scarpati’s dead body was hidden in their room. Indeed, this hypothesis is undermined by the fact that the blood stains on the back of Zap-pulla’s jacket were consistent with droplets of fresh blood, not smears from blood already fallen onto some other site with which Zappulla came into inadvertent contact. See Trial Tr. at 439. Moreover, the numerous blood stains must be considered in light of the undisputed evidence that Scarpati’s murderer did not simply kill her and leave; he hid the body under the mattress and box spring, which means he had to move it. In this context, the logical inference to be drawn from the number, type, and sometimes curious locations of the blood stains on Zappulla’s clothing, is that Zappulla killed Scarpati, and when he hoisted her dead body to place it underneath the box spring, blood from her nose and mouth dripped directly onto the back of his jacket. Meanwhile, in maneuvering her body under the box spring, her blood also smeared onto the front his sweatshirt, his shirt sleeve, the front and back of his pants, and even the top and bottom of his boots. Because any other conclusion about the blood evidence is too improbable to make sense, it was not “objectively unreasonable” for the Appellate Division to view this compelling forensics evidence of Zappulla’s guilt as further support for a finding of harmless error. See People v. Zappulla, 282 A.D.2d at 698, 724 N.Y.S.2d at 435.9
*481(c) The Detailed Admissions to Felix Pabon
Although the circumstantial evidence of Zappulla’s guilt was remarkably strong in this case, the prosecution also offered direct proof of guilt in the form of admissions made by Zappulla to Felix Pabon, an inmate with whom he was incarcerated in July 1998.
In many cases, the admission of a substantially similar confession, without constitutional infirmity, will suffice to render an erroneously admitted statement cumulative and, therefore, harmless. See, e.g., Parsad v. Greiner, 337 F.3d 175, 185 (2d Cir.2003) (finding harmless error when erroneously admitted statements were cumulative of properly admitted statements); Tankleff v. Senkowski, 135 F.3d 235, 245-46 (2d Cir.1998) (admission of defendant’s un-Mirandized statements was harmless beyond a reasonable doubt because those statements were “substantially the same as some of his later, admissible confession[s]”); Rollins v. Leonardo, 938 F.2d 380, 382 (2d Cir.1991) (affirming district court’s harmless error determination that defendant’s un -Mirandized confession was merely cumulative of subsequent confession and other overwhelming evidence of guilt); Campaneria v. Reid, 891 F.2d 1014, 1022 (2d Cir.1989) (error in admitting defendant’s confession in violation of Miranda was harmless where the confession was “entirely cumulative” and “did not contain anything that [the defendant] had not already stated in his earlier [admissible] statements”).10
This is especially true where the admissible evidence encompasses a more complete and detailed description of the offense than the confession that should have been suppressed. See, e.g., Nova v. Bartlett, 211 F.3d 705, 709 (2d Cir.2000) (error in admitting defendant’s statement in violation of Miranda was harmless where defendant subsequently made more complete, detailed confessions that were admissible); United States v. Daniel, 932 F.2d 517, 521-22 (6th Cir.1991) (where defendant’s second, admissible statement “supplied greater information” than the *482first, unconstitutional confession, the first confession was “cumulative and unnecessary to establishing the case against the defendant” and its inclusion harmless error); Martin v. Wainwright, 770 F.2d 918, 932-33 (11th Cir.1985) (admission of confession obtained in violation of Miranda was harmless error where subsequent confession was not only cumulative of evidence contained in earlier confession, but in fact contained far more detailed description of the murder).
Further, the inoculating effects of an admissible confession are particularly strong where the incriminating statements therein are corroborated by independent evidence. See United States v. Newton, 369 F.3d 659, 679 (2d Cir.2004) (finding erroneous admission of defendant’s incul-patory statement of gun possession harmless beyond a reasonable doubt because the defendant had essentially acknowledged ownership of the gun in another properly admitted statement, which was corroborated by additional testimony); United States ex rel. Moore v. Follette, 425 F.2d 925, 928 (2d Cir.1970) (finding admission of an improperly obtained confession harmless because a second, admissible confession covered every element of the crime and was supported .by corroborating evidence).
In this case, Zappulla’s admission to Pa-bon was considerably more detailed about events related to Scarpati’s murder than his signed confession to the police. Further, the details were extensively corroborated by other forensic and testimonial evidence.
For example, in his admissions to Pabon, Zappulla specifically identified his victim as a “prostitute,” Trial Tr. at 569, a detail not mentioned in his confession to the police but corroborated by the testimony of Aaron Cohen, see id. at 151.11 Similarly, Zappulla told the police little about the circumstances leading to his arrival at the Golden Gate Inn, but in his admission to Pabon, he stated that the motel stay was preceded by an incident where he had “dis’d” his “wife.” Id. at 574. Later, he saw police outside his house, prompting him to go to a motel “with a couple of *483people, some guys and girls,” where they all got “high on crack, marijuana, alcohol and pills.” Id. The majority cites an error in this account, noting that only one “girl,” Scarpati, was with Zappulla at the motel. Supra at [470-71], But the error seems trivial when compared to the accuracy of the numerous other details attributed to Zappulla by Pabon. Notably, Zappulla’s disclosure to Pabon that he had gone to the motel with more than one person was correct, as Aaron Cohen confirmed. See Trial Tr. at 159-61. Moreover, Zappulla told Pabon, but not his arresting officers, about a domestic dispute that had resulted in police coming to his home, a detail confirmed by his girlfriend’s March 15, 1998 police report that Zappulla had threatened her and stolen her fur coat and jewelry. See Hearing Tr. at 35, 55, 140. Further, Pabon’s account that Zappulla admitted using crack, marijuana, and pills in the motel room was corroborated by the discovery of these substances by crime scene detectives. See Trial Tr. at 360-63.
According to Pabon, Zappulla also stated that he killed Scarpati when he discovered that “he was missing some of his jewelry that he had in his pocket.” Id. at 575. That Zappulla kept his girlfriend’s jewelry in a pocket was not revealed in his signed confession; nevertheless, this fact was confirmed by police recovery of jewelry from one of his pockets in the course of a post-arrest search. See id. at 20-21, 27.
With respect to the circumstances of Scarpati’s death, Zappulla’s signed confession and his admission to Pabon were substantively identical, though each statement contained details omitted from the other. In his signed confession, Zappulla stated that his argument with Scarpati was preceded by an argument with Aaron Cohen when the taxi driver brought drugs to the motel room that he had acquired by selling jewelry. Zappulla reported that “he got mad because Aaron and Jennifer had taken advantage of him while he was high”; he insisted that “[h]e never gave them permission to sell [his girlfriend’s] jewelry.” Id. at 497. After Cohen left the motel room, Zappulla started to argue with Scarpati about the missing jewelry, threatening to report her to the police. Zappulla claimed that Scarpati then attacked him, and he responded by choking her. He stated that when he left the room, Scarpati was unconscious, but Zappulla did not know whether she was dead. See id.
In his admission to Pabon, Zappulla made no mention of Aaron Cohen’s involvement with his jewelry. He reported waking up in the motel room, discovering some jewelry missing, blaming Scarpati, and starting to “beat” her before he “choked her to death.” Id. at 575-76. Pabon recalled Zappulla telling him that when he killed Scarpati, “some blood ... came out” of the victim. Id. at 575. Although Zappulla was not specific as to the site of the bleeding, he did tell Pabon that some of the blood stained the wall. See id. Zappulla further told Pabon that “he broke [Scarpati’s] back and put her under the mattress.” Id. These details, omitted from the signed confession, are significant in several respects. That Zappulla beat as well as strangled Scarpati finds corroboration in the medical examiner’s discovery of hemorrhages on the dead woman’s head, which were consistent with either blows or a fall. See id. at 336. That Scarpati bled visibly as a result of the strangling — from her nose and mouth — was similarly corroborated by the medical examiner. See id. at 331, 338-39. Crime scene evidence also confirmed that some of Scarpati’s blood was found on the wall adjacent to the bedframe where her body was hidden. See id. at 216-18, 339. Finally, although the medical examiner reported no fractures to Scarpati’s back, her body was certainly found where Zappulla told Pabon *484he had hidden it: under the mattress. Moreover, crime scene photographs apparently depicted the body shoved into the bed frame at such a distorted angle that Zappulla might well have thought that he had broken his victim’s back. See People’s Exh. 15; Hearing Tr. Feb. 9,1999 at 57 (describing finding Scarpati’s body “in a U-shape[ ] — stuffed under the bed”).
Zappulla also shared with Pabon numerous details about events after the murder that he had not mentioned in his police confession. Zappulla explained that after leaving the motel, he realized that he had forgotten to take the fur coat out of the room and went back to retrieve it. At that time, police officers, who were at the motel to investigate an unrelated incident, stopped Zappulla, leading to his arrest. This sequence of events is corroborated by the motel surveillance tape, which shows Zappulla’s 12:18 a.m. departure from the motel followed by a 12:19 re-entry. It is also corroborated by police testimony that an unrelated investigation about an argument at the motel had brought officers to the scene. At 12:20 a.m., they encountered Zappulla in a hallway and arrested him when one of the officers recognized him from a photograph disseminated at the station earlier that day. See Trial Tr. at 379-82, 463. Further corroborating this part of Pabon’s account is the fact that a fur coat was indeed still in the motel room at the time of the arrest.
Pabon also testified that in a subsequent conversation, Zappulla told him that he expected to “beat” his case because paperwork provided by his lawyer indicated that police had conducted a warrantless search of his motel room on the morning of March 17, 1998. Id. at 578. The statement was corroborated by the fact that at approximately 8:45 a.m. on March 17, there was a warrantless entry into Room 234 by a hotel employee accompanied by two police officers who remained outside. See id. at 247-48. Later in the day, pursuant to a search warrant, police entered the room and found Scarpati’s dead body.
As the majority notes, in considering whether Zappulla’s admission to Pabon renders harmless Zappulla’s xm-Miran-dized confession, careful scrutiny of Pa-bon’s credibility is required. He is, after all, a convicted felon with an extensive and serious criminal record and a history of drug abuse, both in and out of prison. Further, he testified pursuant to a cooperation agreement with the prosecution, which he hoped would result in a reduced sentence on pending charges against him. These circumstances do not, however, make it “objectively unreasonable” for a state court to have relied on Pabon’s testimony in making its b^milessness determination.
The extraordinary detail in the statements attributed to Zappulla by Pabon together with extensive corroboration of these details preclude a conclusion that Pabon simply fabricated the admissions to which he testified out of whole cloth. The majority nevertheless suggests that “there was another way” for him “to have learned the details of Zappulla’s case other than from a confession from Zappulla.” Supra at [12]. Specifically, he might have heard these details from an “inmate law clerk who was preparing Zappulla’s legal paperwork,” 12 or he might have read them in *485legal papers shown to him by Zappulla himself. Id.
The problem with this theory is that it derives largely from Pabon’s acknowledgment (1) that an inmate law clerk first told him that the victim in Zappulla’s murder case was a woman — until that time, Pabon had simply assumed that Zappulla had killed a man, see Trial Tr. at 568 — and (2) that Zappulla himself had shown him certain documents, specifically, a list of the charges against him, a list of grievances he was maintaining against his wife/girlfriend, and a photocopied picture of the victim’s body, which Pabon stated was so dark that little was discernable, id. at 607-08. This conversation and the identified documents might have given Pabon some information about Zappulla’s case, but hardly the extensive details about which he testified.
Pabon acknowledged that Zappulla had other documents pertaining to his case in his prison cell, but he denied seeing these papers or having any further conversations with the inmate law clerk about Zap-pulla’s case. Of course, a jury is not required to believe a witness’s denials, but that does not mean that a reviewing court, which has not observed the witness’s demeanor, may speculate that the fact denied is the truth. See generally Dyer v. MacDougall, 201 F.2d 265, 269 (2d Cir.1952) (L.Hand, J.) (holding that witness demeanor may persuade a jury to “assume the truth of what he denied,” but a court cannot allow a case to go to the jury on such evidence). In this case, although the record indicates that Zappulla had copies of unspecified legal documents in prison, it nowhere demonstrates that the particular factual details attributed to Zappulla by Pabon were contained in these documents,13 much less that Pabon had seen these materials. In any event, Pabon’s powers of memory and synthesis would have had to have been extraordinary to cull from diverse court papers, police reports, and forensics analyses, the detailed admissions he attributed to Zappulla.
In sum, I do not agree with the majority’s conclusion that the admissible evidence of Zappulla’s guilt was weak.14
*486Where, as in this case, an informant (even one with a record and a deal), recounts an admission by the defendant that is remarkably detailed and largely corroborated, it is not “objectively unreasonable” for a state court to conclude that such evidence, together with proof that the defendant was with the murder victim at and about the time of her death and that the victim’s blood was on the defendant, so overwhelmingly establishes guilt that “there [was] no reasonable possibility that the error of admitting defendant’s [un-Mirandized confession] into evidence might have contributed to his conviction.” People v. Zappulla, 282 A.D.2d at 698, 724 N.Y.S.2d at 435.
2. The Prosecution’s Reliance on the Signed Confession Does Not Render “Objectively Unreasonable” the State Court Determination of Harmless Error
As the majority notes, in the prosecution’s opening statement at Zappulla’s second trial, it identified the signed confession as the first piece of evidence and discussed in more detail the interaction between Zappulla and the police that led to this statement. On the other hand, no greater emphasis was placed on the confession during the evidentiary phase of the second trial, and the prosecution did not even refer to the confession in summation until twenty-five pages into the argument. The summation discussion of the confession takes up a scant few of the more than forty-five total pages, with at least equal time devoted to a discussion of Zappulla’s admissions to Pabon.
This conduct does not compare to that in Campaneria v. Reid, where the prosecutor’s summation emphasized an inadmissible confession almost to the exclusion of the properly received evidence. 891 F.2d at 1023-24 (Kearse, J., dissenting) (noting that in a fourteen-page summation, the prosecutor directed the jury’s attention to the inadmissible confession no less than a dozen times). At issue in Compañería was a tape-recorded interrogation conducted after defendant had invoked his right to remain silent. Despite the summation concerns raised by Judge Kearse, the majority concluded that receipt of the inadmissible evidence was harmless in light of a trio of untainted admissions by defendant. Id. at 1022. Given the overwhelming admissible evidence of Zappulla’s guilt already discussed, I cannot conclude that the prosecutor’s use of the un-Mirandized confession at the second trial was so prejudicial as to render the Appellate Division’s finding of harmlessness “objectively unreasonable.”
3. The Signed Confession Was Cumulative of the Admission to Pabon and Did Not Prejudice Zappulla’s Defense that Cohen Was the Real Murderer
The majority concludes that the confession in this case cannot be deemed cumulative because it “filled in a missing link to the prosecution’s case: motive.” Supra at [472], I respectfully disagree. Zappulla’s admission to Pabon put before the jury the same motive for Scarpati’s death as did the confession: a dispute with Scarpati over missing jewelry. To this extent the statement was cumulative.15
*487The majority states that receipt of the confession into evidence prejudiced the defense by resolving in Pabon’s favor an inconsistency between Zappulla’s admission that he killed Scarpati in a dispute about missing jewelry and Aaron Cohen’s testimony that Zappulla had voluntarily given him jewelry to sell for drugs. The conflict was important to the defense argument that “Cohen was actually the murderer.” Supra at [473]. Once again, I must disagree.
First, I am not persuaded that the identified conflict exists within the admissible evidence, i.e., between Zappulla’s admission as reported by Pabon and Cohen’s testimony. According to Pabon, Zappulla was angry with Scarpati, not Cohen, about the missing jewelry. Thus, a jury could reasonably have believed both Pabon and Cohen, concluding that Zappulla had voluntarily given Cohen some jewelry to sell for drugs, but had then become angry with Scarpati when he discovered other jewelry missing.16
Second, it is Zappulla’s un-Mirandized confession, not his admission to Pabon, that establishes the conflict identified by the majority. Zappulla told the police that he had argued with Cohen as well as Scar-pati about missing jewelry, implying either that he had never given any jewelry to Cohen (contrary to Cohen’s testimony) or, if he did, that he never intended the jewelry to be sold.17 The problem for Zappulla is that, in the same confession, he admitted killing Scarpati, thereby rendering any conflict about the sale of the jewelry irrelevant to identification of the murderer. A defendant can hardly claim prejudice from the fact that the same piece of evidence that affords him a possible line of impeachment renders the subject of impeachment irrelevant to the critical issue in the case.
I confess to a certain puzzlement as to how resolution of the jewelry conflict would, in any event, support a defense that Cohen, not Zappulla, was Scarpati’s murderer. If the jury decided that Cohen was lying about receiving jewelry voluntarily from Zappulla, that would leave intact Zappulla’s motive for killing Scarpati. If, on the other hand, the jury decided Cohen was telling the truth, it is hard to see how veracity would make it more likely that he was a murderer. At best, such a finding might raise questions about Zappulla’s credibility in telling the police that he argued with Cohen about jewelry, or about the credibility of the police in reporting Zappulla’s confession. But since Pabon did not testify to any statements by Zap-pulla regarding a jewelry dispute with Cohen, Cohen’s testimony would not impeach him.
In considering how, if at all, the un-Mirandized confession prejudiced Zappul-la’s purported defense that Cohen killed Scarpati, it is worth noting that this possibility was raised only obliquely in defense counsel’s summation at Zappulla’s second trial. In cross-examining Cohen, defense counsel never directly confronted the cab *488driver with the possibility that he was Scarpati’s murderer. Indeed, the cross-examination, conducted on a first-name basis, was almost cordial, see Trial Tr. at 181 (“Q: Good afternoon, Aaron. A: Good afternoon, Sam.”), a result, perhaps, of Cohen’s willingness to speak with the defense as well as the prosecution in advance of trial. Significantly, counsel carefully avoided probing Cohen’s whereabouts during the hours when the murder likely took place, although Cohen testified that law enforcement authorities had required him to provide such information. Instead, relying on the fact that Cohen did not attempt to contact Scarpati any time after 5:30 p.m. on March 16, counsel made a single attempt to insinuate that Cohen might be the real murderer without ever actually saying so: “Does it make sense to anyone of you here that this person with all that contact, all of a sudden stopped at 5:30 after hearing his friend, who he is worried about, had a fight at 5:30? Didn’t look for her, didn’t find her, didn’t go to the police; they came to him a week later.” Trial Tr. at 692.
The majority nevertheless concludes that such a defense “was not far-fetched” because Cohen abused crack, had access to Room 234, and had argued with Scarpati on the afternoon of her death. Supra, at [473]. These points, raised by Zappulla’s counsel on appeal, were not in fact ever argued to the jury as a basis for thinking Cohen was the real murderer. In any event, it was Cohen who disclosed the fact that Scarpati was angry with him — not necessarily the other way around- — for wanting to bring another prostitute to Room 234 early in the morning on March 16. Presumably, the majority thinks this dispute provides a motive for murder. Indeed, the majority concludes that precisely because Cohen and Scarpati were “very, very close friends” that it was “more likely” he would “kill her than a random stranger, like Zappulla.” Supra at [473]. The majority’s assumption that Zappulla and Scarpati were “random stranger[s]” is belied by the evidence. Bruce Brodsky, with whom Scarpati lived, testified that Zappulla and the murdered woman were neighborhood friends. See Trial Tr. at 471. Indeed, at the pre-trial suppression hearing, it was revealed that Zappulla’s girlfriend, in making her March 15 complaint to the police, had stated that Zap-pulla might be found in the company of “Jennifer Imparato,” another name used by Scarpati. See Hearing Tr. at 36, 79. Even if the couple had been strangers before March 15, it is not obvious to me why a prostitute’s murderer was “more likely” to have been a “very, very close friend” than the “random stranger” with whom she was sharing the room in which she died, particularly when that “stranger” is found to have the woman’s blood at ten different locations on his clothing and provides a detailed admission of guilt to a fellow inmate.
In sum, like the justices of the Appellate Division and Judge Weinstein, I am convinced that the admissible evidence of guilt — Zappulla’s presence in Room 234 with Scarpati at and about the time of her murder; the dead woman’s blood on his clothing on the day of her death; and Zappulla’s detailed admission of guilt to Felix Pabon, corroborated in almost every respect by other admissible evidence — is overwhelming. Further, I am satisfied that Zappulla’s mi-Mirandized confession of guilt was cumulative of the more-detailed admission to Pabon on the issue of motive, did not prejudice his defense, and was not unduly highlighted by the prosecution to the exclusion of the admissible evidence. Accordingly, I conclude that the Appellate Division’s finding of harmless error in this case was not “objectively un*489reasonable” and I would affirm the denial of habeas corpus relief on that ground.
4. Ambiguities in the Record Relating to Whether There Was a Miranda Error in this Case
Because I am convinced that the Appellate Division finding of harmless error was objectively reasonable, I would not reach respondent’s argument that there was, in fact, no Miranda error in this case. Because my colleagues in the majority do address this point and summarily conclude that the totality of the circumstances support a finding of error, see supra at [474-75],18 I briefly explain why I do not find this question so easy to answer on the record before us.19
Preliminarily, it is important to recall that the issue presented by Zappulla’s confession is not one of Fifth Amendment compulsion but one of due process. Specifically, after the police gave Zappulla Miranda warnings when he was first arrested. in connection with theft allegations by his girlfriend, did due process demand that those warnings be re-administered approximately twenty-four hours later before police questioned Zappulla about Jennifer Scarpati’s murder?
If, 'as the majority states, upon first being advised of his rights, Zappulla “refused to speak” with the police about the crime then under investigation, • the theft from his girlfriend, supra at [474], then I would agree that the law permits interrogation about a different crime only upon re-advice of rights. See Campaneria v. Reid, 891 F.2d at 1021 (noting that even after a defendant has asserted his right to remain silent, “[questioning can be resumed after fresh Miranda warnings are given and the right to remain silent is otherwise- scrupulously honored, for example, by renewing the questioning only after the passage of a significant period of time and by limiting the renewed questioning to a different subject matter than the original *490interrogation”) (citing Michigan v. Mosley, 423 U.S. 96, 104-07, 96 S.Ct. 321, 46 L.Ed.2d 313 (1975)). But the record is somewhat ambiguous as to whether Zap-pulla actually refused to answer any questions about the theft or simply responded to all police inquiries by diverting the conversation to other matters.
Further, to the extent interrupted custody and the passage of time are relevant factors in determining when Miranda warnings must be re-administered, the unusual circumstances in this case must be noted. Zappulla’s police custody was in-tei'rupted only because he attempted to escape and, in the process, was hit by a car, requiring a hospital visit that delayed the second interrogation. For purposes of considering the police obligation to re-administer Miranda warnings, Zappulla’s situation is hardly analogous to that of a person released from custody with freedom to go where he wished for twenty-four hours, or to a person whose own conduct played no role in delaying his interrogation.
Even assuming, however, that police were required to re-advise Zappulla of his Miranda rights, the law would demand suppression only of custodial statements made in response to interrogation. See Rhode Island v. Innis, 446 U.S. 291, 301, 100 S.Ct. 1682, 64 L.Ed.2d 297 (1980). The record suggests that after being told that the police had found Searpati’s body and wished to question him on the matter, Zappulla promptly stated that he wanted a lawyer. See Trial Tr. at 274, 486. However, when the police responded that “if he wanted an attorney [they] would provide one to him, but that [they] weren’t going to speak to him any longer,” Zappulla clarified that “no, you don’t understand. I don’t want an attorney for me.. I want a[n] Assistant District Attorney here so that we can get some stuff down in writing.” Id. at 488; see also id. at 274 (“[N]o, not a lawyer for me, I want the D.A. down here.”).20 Zappulla explained that he wanted the district attorney present “to get this over with” so that he could be “guaranteed the death penalty.” Id. at 275; see also id. at 486. (“I feel that I want to have a lawyer here so that we can all sit down together and get this thing straight, because I want the death penalty.”). Apparently, no issue was raised in the state courts and, therefore, no findings were ever made, as to whether Zappulla’s statements regarding an attorney and, more importantly, the clarification that followed, were responses to interrogation. I note this point because if these statements were admissible, particularly Zappulla’s professed desire for the death penalty, it would certainly reinforce the Appellate Division’s conclusion that receipt of the written confession was harmless error.
While I do not urge remand to clarify these ambiguities because I would affirm the Appellate Division’s ruling on harmless error, because the majority is not of that view and, nevertheless, addresses none of these concerns, I cannot join its holding that Miranda necessarily mandated the suppression of all Zappulla’s statements regarding the Scarpati murder.
The majority observes that it is not “too much” in a free society to require the state to retry a defendant who has been deprived of due process in a way that is not harmless. See supra at [475]. I agree. But the cost of such a retrial should not be underestimated; it is borne not only by the state, but by victims, witnesses, the *491judicial system, and society itself. Nor should blithe assumptions be made that if I (and six other reviewing judges) am right that the admissible evidence of Zappulla’s guilt was overwhelming, “the State should not find it difficult to properly obtain a conviction” for murder. Supra at [475]. The passage of time can complicate retrial of even the strongest cases: witnesses die or become difficult to locate; memories fade; evidence degrades. For all these reasons, habeas relief is warranted only if a petitioner demonstrates that his conviction was, in fact, infected by constitutional error' and that error was not harmless.
This is not such a case. The Appellate Division expressly found that any Miranda error was harmless beyond a reasonable doubt, and the overwhelming admissible evidence of guilt simply does not permit me to label this ruling objectively unreasonable. Accordingly, I respectfully dissent from the court’s grant of a writ of habeas corpus on this appeal.
. As the majority notes, Zappulla had received Miranda warnings approximately twenty-four hours earlier when he was initially arrested in connection with the theft of jewelry and a fur coat from his girlfriend, Margaret Tarulli. See supra at [465]. He was no stranger to these warnings; Zappulla had ten prior convictions, four for felony crimes, including one for racketeering with a predicate of solicitation to commit murder. See Sentencing Tr., Mar. 30-31, 1999, at 4-5, 13. On the occasion here at issue, Zappulla agreed to speak with authorities but only discussed matters unrelated to the theft. The Appellate Division concluded that the lapse of time between the two interrogations and the difference in the crimes at issue required Miranda warnings to be re-administered. See People v. Zappulla, 282 A.D.2d at 697, 724 N.Y.S.2d at 433.
. The majority's opinion might be read to hold that a determination of objective unreasonableness by itself warrants granting a writ of habeas corpus. Some of our sister circuits, however, have concluded that a second step of analysis is required.
Most recently, in Medina v. Homung, the Ninth Circuit ruled that a federal court engaged in habeas review of a harmless error challenge should (1) "determine whether the state court’s harmless error analysis was objectively unreasonable,” and, if it was, (2) "engage in a Brecht analysis” of the error. 372 F.3d 1120, 1126. (9th Cir.2004). The Seventh and Tenth Circuits have also approved a two-step review. See Aleman v. Stemes, 320 F.3d 687, 690-91 (7th Cir.), cert. denied, 539. U.S. 960, 123 S.Ct. 2653, 156 L.Ed.2d 659 (2003) (noting that a federal court may grant habeas relief only if it finds *477that the state court unreasonably applied Chapman and the error is cognizable under Brecht); Saiz v. Burnett, 296 F.3d 1008, 1012 (10th Cir.2002) ('‘[T]he proper question on federal habeas review is whether the Colorado appellate court’s application of the Chapman standard was objectively unreasonable .... [If a court concludes that it was], the court should engage in an independent harmless error analysis applying the standard articulated in Brecht v. Abrahamson.”).
I expect the majority would conclude that the result it reaches today would pertain even under the two-step analysis employed by the Seventh, Ninth, and Tenth Circuits. Nevertheless, it would be useful to clarify whether in this circuit (1) a second step of analysis is necessary, and, if so, (2) that step requires Brecht or Chapman review.
. In this case, no evidence indicates that Zap-pulla's confession was the product of any coercion. The inadmissibility of the evidence is based solely on a failure to re-administer Miranda warnings.
. Zappulla was first arrested on his girlfriend's March 15, 1998 complaint that he had broken into her apartment, threatened her, and taken jewelry and a fur coat.
. The majority faults the prosecution for "shoddy forensics” work in this case, asserting that "any doubt regarding the identity of the killer could have been relieved had the prosecutor ordered forensic tests on the two articles that the killer clearly touched — the underwear found around Jennifer's neck and the blood-stained towel found under the bed.” Supra at [473].
It is unclear what forensics tests the majority has in mind. Trial testimony indicated that fingerprints, which investigators did, in fact, lift from other items in the motel room, can rarely be retrieved from cloth exhibits, a con-*481elusion that would seem obvious with respect to terry cloth. In any event, there is no contrary evidence in the record before us. As for DNA analysis, nothing in the record even suggests that there was any genetic material on the towel other than Scarpati’s blood. Even if there were, or if Zappulla’s semen had been identified on Scarpati's underpants, that would hardly relieve “any doubt” that he was the murderer. Zappulla had occupied Room 234 with Scarpati for more than twenty-four hours; the fact that he might have used a towel in that room or had sexual relations with Scarpati would hardly be surprising.
This is not to suggest that such forensics evidence — if it could have been procured— would have been irrelevant. I simply do not join in the majority's harsh criticism of the investigative efforts in this case. Law enforcement authorities did secure the forensics evidence that speaks loudest in identifying Scarpati’s murderer: her own blood at numerous locations on the clothing Zappulla was wearing at the time and site of her death.
. Although the majority conclusorily asserts that "a written confession can never truly be said to be merely cumulative,” supra at [474], case law is to the contrary. See United States ex rel. Moore v. Follette, 425 F.2d 925, 928 (2d Cir.1970) (Friendly, J.) (noting that “the case where admission of an improperly obtained confession can be considered harmless error is exceedingly rare,” the court, nevertheless, rules the defendant's transcribed confession harmless precisely because it was cumulative of other evidence); accord Campaneria v. Reid, 891 F.2d at 1022 (concluding that receipt of even video-recorded confession was harmless where cumulative of other evidence); see also Boles v. Foltz, 816 F.2d 1132, 1135-36 (6th Cir.1987) (holding that defendant’s "taped confession was merely a reiteration of his first [unrecorded] confession and therefore cumulative”); Martin v. Wainwright, 770 F.2d 918, 926, 932-33 (ruling admission of video-taped confession was "merely cumulative” of other evidence).
. To facilitate comparison with Zappulla's admission to Pabon, I reproduce the police confession, as read at trial, deleting exchanges between counsel and the witness during this reading:
On March 18, 1998, at 0115 hours, the undersigned [Detective McMahon and] Detective Delahunt, were present at the 61 Precinct Detective Squad Room. Guy [Zap-pulla] stated that he would rather tell his story to the District Attorney. He wanted to speak with them about pleading guilty and getting this over with as quickly as possible. He also stated that he wanted to know how long of a sentence he would be facing.
Guy was questioned further and stated that a male he knows as Teamo or Aaron had rented the room at the Golden Gate Motel on Sunday and Monday. Guy went down to the front desk sometime on Monday to pay $10 for the phone. He stated that he had been getting high for three days.
Guy stated that he had gone to the room with his girlfriend Margaret's fur coat and jewelry. Sometime on Monday he noticed that Aaron, who had left the room, returned with more drugs. Aaron said that he had used Margaret’s jewelry to get the drugs. Guy stated that he got mad because Aaron and Jennifer had taken advantage of him while he was high. He never gave them permission to sell Margaret’s jewelry. He got mad but stated that he still used the drugs.
Aaron left the room, and Guy states that he started to argue with Jennifer about the jewelry. Guy stated Guy threatened to call the police. He states that that's when Jennifer started to fight with him. He fought back and choked her. Guy didn't know if she was dead. She was unconscious when he left the room.
Trial Tr. at 495-98.
. Although Pabon testified that the inmate law clerk helped prisoners analyze their case, read police paperwork, and prepare legal motions, it is not clear that he ever prepared papers for Zappulla who, after all, was represented by counsel in connection with his pending trial. The testimony indicates simply that Zappulla told Pabon that the inmate was generally helping him.
. Nothing in the record indicates whether in July 1998, Zappulla's papers were limited to documents filed in court, such as the complaint, indictment, motions, and transcripts of appearances, or whether he had also, by that date, received materials usually provided closer to or at trial, for example, witness statements, forensics reports, documentary exhibits, etc. Zappulla did not first stand trial for Scarpati’s murder until February 1999.
. The majority cites as further evidence of the "weakness” of the prosecution case — even with the xm-Mirandized confession — the fact that Zappulla's first trial ended in a hung jury. See supra at [471]. While this fact is properly considered, it is by no means determinative. As this court recently observed, "[a] jury may hang for any number of reasons, including the idiosyncratic views of a single juror.” United States v. Newton, 369 F.3d at 680. In Zappulla’s case the record reveals that the first jury hung with eleven members in favor of conviction and one juror favoring acquittal. See Trial Tr. at 685, 697.
The majority also suggests that the "length and deliberative conduct” of the jury at Zap-pulla's second trial indicates that the evidence of guilt was not overwhelming. Supra at [471]. I cannot agree. The jury, which was charged on the law on the morning of March 10, 1999, returned a guilty verdict at 5:15 p.m. the same day. During that time, it made several requests for evidence. Given that the charge in the case, murder, could not have been more serious, neither the jury's interest in carefully reviewing the evidence nor the length of its deliberations can reasonably be interpreted as indicative of a weak case. When in United States v. Jean-Baptiste, 166 F.3d 102 (2d Cir.1999), we noted that a full day's deliberations indicated that the government's case was not overwhelming, we were dealing with a charge of making a false statement on a passport application, see 18 U.S.C. § 1542, and the "entire presentation of evidence at trial consumed little more than half a day,” Jean-Baptiste, 166 F.3d at 109. Given that Zappulla's trial spanned a week during *486which twenty witnesses testified, some persons might conclude that a verdict returned in a day's time was more indicative of a strong rather than a weak prosecution case.
. To support its conclusion that the confession was not cumulative evidence, the majority quotes the prosecutor's statement in colloquy: " 'how do we show any motive of why the killing happened, other than by the defen*487dant's own statement.’ ” Supra at [472] (quoting Trial Tr. at 167). The statements under discussion, however, were neither Zap-pulla’s confession to the police nor his admission to Pabon. Rather, the parties were arguing about whether Cohen should be allowed to testify to Zappulla’s statements to him in a March 16 conversation about jewelry. The defense objected to this evidence on hearsay grounds. See Trial Tr. at 164-68.
. The fact that Zappulla still had jewelry in his pocket at the time of arrest indicates that he did not give Cohen all the jewelry.
. The latter possibility is suggested by Zap-pulla’s confession that "he got mad because Aaron and Jennifer had taken advantage of him while he was high," and that “[h]e never gave them permission to sell [his girlfriend's] jewelry.” Trial Tr. at 497.
. The only ''circumstances” referenced in the majority opinion are three identified by the Appellate Division: (1) approximately twenty four hours had passed between Zap-pulla’s initial receipt of Miranda warnings and police questioning about the Scarpati murder, (2) Zappulla was not in continuous police custody in this interval, and (3) the later interrogation concerned a different crime from the one for which Zappulla was initially arrested. See People v. Zappulla, 282 A.D.2d at 697-98, 724 N.Y.S.2d at 435-36.
. Federal courts considering petitions for writs of habeas corpus are not "bound by a state appellate court's ruling in petitioner’s favor on a matter of constitutional law.” Pinkney v. Keane, 920 F.2d 1090, 1093 (2d Cir.1990) (affirming district court denial of habeas on the basis of evidence ordered suppressed by the state appellate court), cert. denied, 501 U.S. 1217, 111 S.Ct. 2824, 115 L.Ed.2d 995 (1991). Although AEDPA requires deference to state court rulings challenged by a petitioner seeking habeas relief, the statute nowhere requires parallel deference to rulings in his favor. Nor is such symmetry warranted. The observations made by Judge Oakes, writing for the panel in Pink-ney, are as pertinent after AEDPA as they ■ were before:
[W]e believe that symmetry, while aesthetically pleasing, is a particularly inappropriate principle to invoke in a habeas proceeding under 28 U.S.C. § 2254. Habeas corpus is not a neutral proceeding in which the petitioner and the State stand on an equal footing. Rather, it is an asymmetrical enterprise in which a prisoner seeks to overturn a presumptively valid judgment of conviction. Because of this, the petitioner generally bears the burden of proof ■throughout' the habeas proceeding.... In addition, issuance of the writ is an extraordinary remedy, ... granted only in the exercise of a sound judicial discretion.... To constrain the State merely to impose some symmetry on an otherwise one-sided proceeding would be to substitute aesthetics for legal reasoning.
Pinkney v. Keane, 920 F.2d at 1094 (internal citations omitted).
. This exchange suggests that the issue before us is technical Miranda compliance; there appears to be no question that Zappulla was, in fact, fully aware of his rights in speaking with the police.