dissenting.
The Court today adopts the position that law-enforcement officers may renew contact with criminal suspects upon learning from third parties that the suspects are willing to waive their previously invoked right to counsel. In so holding, the majority concludes that neither reason nor established case law require suspects — who, by definition, are in jail surrounded at all times by law-enforcement personnel — to directly communicate to police their wish to waive their previously invoked constitutional rights. The Supreme Court has instructed us that we must view custodial waivers of rights with a high degree of suspicion. In my view, we must be doubly skeptical of a waiver of rights effected through the backdoor of a purported third-party agent, especially when all the suspect has to do is proclaim to the nearest guard, “I want to talk.”
Because I believe that only the suspect or his attorney may initiate discussions with the police (defined broadly to include other corrections officers) after a suspect has invoked his right to counsel, I respectfully dissent. Further, even assuming that third-party initiations are permissible, I disagree with the majority’s conclusion that the record in this case establishes that Van Hook reinitiated communication with the police through his mother.
*429I.
A.
Edwards v. Arizona, 451 U.S. 477, 101 S.Ct. 1880, 68 L.Ed.2d 378 (1981), creates the brightest of bright-line rules: If a suspect invokes his constitutional right to counsel, the police must immediately terminate their questioning. The interrogation may be renewed only if the suspect “himself’ initiates further conversation “with the police.” Id. at 485, 101 S.Ct. 1880. There is no doubt that police officers who ask a suspect hours or days after the suspect invoked his right to counsel whether he was now ready to talk would be guilty of the very “badgering” that Edwards proscribes. Oregon v. Bradshaw, 462 U.S. 1039, 1044, 103 S.Ct. 2830, 77 L.Ed.2d 405 (1983); see also United States v. Johnson, 400 F.3d 187, 193-94 (4th Cir.2005), cert. denied, 546 U.S. 886, 126 S.Ct. 134, 163 L.Ed.2d 133 (2005) (holding that police violated Edwards where an officer re-entered the interrogation room forty minutes after the suspect requested a lawyer to see if the suspect would waive his rights). Indeed, even if the police have reason to believe that a suspect is ready to waive his previously invoked rights, the police are prohibited from initiating contact with him. See United States v. Rodriguez, 993 F.2d 1170, 1174 (5th Cir.1993) (concluding that the police improperly renewed questioning of the suspect where a co-defendant contacted the police to say that “they” wanted to speak but it was not clear to whom “they” referred); Desire v. Attorney Gen. of Cal., 969 F.2d 802, 804-05 (9th Cir.1992) (holding that the police wrongly initiated contact where the officer overheard a conversation between the suspect and his co-defendant and in response, asked the suspect whether he “wanted to talk about anything”). Nonetheless, the majority holds that the police may do this very thing — ask a suspect whether he wants to talk after a suspect has invoked his right to counsel — without offending Edwards, so long as the police learn from a third party that the suspect is willing to submit to their questioning.
At first blush, there does not appear to be anything extraordinary or untoward about authorizing police to renew custodial interrogations when third parties, such as family or friends, tell the police that the suspect is prepared to answer questions without the assistance of counsel. But here we must pause and review what we know. The suspect has been taken into custody and has been read his Miranda rights. See Miranda v. Arizona, 384 U.S. 436, 444-45, 86 S.Ct. 1602, 16 L.Ed.2d 694 (1966). He has terminated law enforcement’s questioning by unambiguously invoking his right to counsel. He has preserved that right by maintaining his silence. At some point, according to the majority, the suspect changes his mind, perhaps while conferring with a family member or friend, and decides that he will waive his right to counsel after all. So committed to reversing course and so desirous of speaking is the suspect, that he does a strange thing: He does not tell the people that he wants to talk to, i.e., the police, that he wants to talk to them. He tells someone else, someone who is not a police officer.
The majority overlooks this logical flaw. Rather than concluding, as Edwards commands, that if a suspect genuinely wishes to waive his previously invoked right to counsel and answer law enforcement’s questions, the suspect will so inform the police directly, the majority endorses the counter-intuitive proposition that we may treat a suspect as willing to talk to the police despite his silence to the police. United States v. Michaud, 268 F.3d 728, 741 (9th Cir.2001) (Reinhardt, J. dissent*430ing) (“The whole point of Edwards is that a suspect who has invoked her right to counsel may remain silent, and may not be questioned unless she breaks that silence...").
In addition to eviscerating Edwards, the majority’s holding deviates from the clear import of the Supreme Court’s jurisprudence on custodial interrogations. It is to that I now turn.
B.
As the majority correctly explains, Miranda and its “corollary,” Edwards, have dual purposes, including safeguarding a suspect’s exercise of his constitutional rights in the face of the state’s disproportionate power and providing clear guidance to law enforcement in how to handle custodial interrogations. Arizona v. Roberson, 486 U.S. 675, 680, 108 S.Ct. 2093, 100 L.Ed.2d 704 (1988) (describing Edwards as a “corollary” to Miranda). The majority further acknowledges that the Supreme Court has praised the “relativfe] rigid[ity]” and “clear and unequivocal” nature of the Miranda and Edwards rules as one of their chief benefits. Fare v. Michael C., 442 U.S. 707, 718, 99 S.Ct. 2560, 61 L.Ed.2d 197 (1979); Roberson, 486 U.S. at 682, 108 S.Ct. 2093. Indeed, “[t]he merit of the Edwards decision,” according to the Supreme Court, “lies in the clarity of its command and the certainty of its application.” Minnick v. Mississippi, 498 U.S. 146, 151, 111 S.Ct. 486, 112 L.Ed.2d 489 (1990); Roberson, 486 U.S. at 680, 108 S.Ct. 2093 (“As we have stressed on numerous occasions, one of the principal advantages of Miranda is the ease and clarity of its application.”) (internal quotation marks and citation omitted).
Although the majority pays lip service to these precepts, it never comes to terms with them central place in the Supreme Court’s custodial-interrogation jurisprudence. Coursing unmistakably through the Court’s doctrine is the insistence that the relationship between a criminal suspect and his official interrogator be governed by “bright-line” rules that precisely spell out the rights and duties of both parties. Roberson, 486 U.S. at 681, 108 S.Ct. 2093 (“We have repeatedly emphasized the virtues of a bright-line rule in cases following Edwards as well as Miranda.”)-, Minnick, 498 U.S. at 151, 111 S.Ct. 486 (“Edwards .. .implements the protections of Miranda in practical and straightforward terms.”). The Supreme Court has determined that in this area of the law, neither suspect, nor police officer, nor the public at large, is well served by rules that tolerate ambiguity or require nuanced decisionmaking. The Court’s unflagging commitment to “clarity” is motivated by the recognition that the coercive setting of custodial interrogation is ready-made for the infringement, whether intentional or inadvertent, of constitutional protections, such that suspects must be plainly advised of their rights so they may act on them. See Dickerson v. United States, 530 U.S. 428, 435, 120 S.Ct. 2326, 147 L.Ed.2d 405 (2000) (“Because custodial police interrogation, by its very nature, isolates and pressures the individual, we stated that ‘even without employing brutality, the “third degree” or other specific stratagems, custodial interrogation exacts a heavy toll on individual liberty and trades on the weakness of individuals.’ ”) (quoting Miranda, 384 U.S. at 455, 86 S.Ct. 1602); Bradshaw, 462 U.S. at 1044, 103 S.Ct. 2830 (describing Edwards as a “prophylactic rule” intended “to protect an accused in police custody from being badgered by police officers”). Furthermore, law enforcement must be advised about the limits of its authority, so that it is not subjected to judicial second-guessing as to its methods, or burdened by undue complexity in *431the discharge of its investigatory work. Davis v. United States, 512 U.S. 452, 461, 114 S.Ct. 2350, 129 L.Ed.2d 362 (1994) (describing Edwards as providing “a bright line that can be applied by officers in the real world of investigation and interrogation ... ”); Fare, 442 U.S. at 718, 99 S.Ct. 2560 (stating that the “relatively rigid requirement that interrogation must cease upon the accused’s request for an attorney ... has the virtue of informing police and prosecutors with specificity as to what they may do in conducting custodial interrogation ...”); Miranda, 384 U.S. at 441-42, 86 S.Ct. 1602 (describing one of the purposes of the decision as providing “concrete constitutional guidelines for law enforcement agencies and courts to follow”).
The Supreme Court has steadfastly guarded the “bright-line” quality of the Edwards rule. The Court has applied Edwards to new factual scenarios where it deemed doing so necessary to uphold the rule’s essential clarity, but has resisted applying Edwards where the result would be to muddy its dictates. Several cases illustrate the point.
In Minniek, the Court addressed whether Edwards is satisfied where the suspect has invoked his right to counsel and has been afforded an opportunity to consult with his lawyer prior to interrogation, but the lawyer is not physically present during the interrogation. The Court concluded that the ability to consult was insufficient to allow interrogation, in part because such an exception “would undermine the advantages flowing from Edwards’ ‘clear and unequivocal’ character.” 498 U.S. at 154, 111 S.Ct. 486. Suspects’ rights would be jeopardized because the quality of consultations would vary widely, ranging from “a telephone call to say that the attorney is en route, to a hurried interchange between the attorney and client in a detention facility corridor, to a lengthy in-person conference in which the attorney gives full and adequate advice....” Id. at 155, 111 S.Ct. 486. Further, approving a consultation-only rule would entail unnecessary trouble for law-enforcement personnel: “[E]ven with the necessary scope of consultation settled, the officials in charge of the case would have to confirm the occurrence and, possibly, the extent of consultation to determine whether further interrogation is permissible.” Id.
In Davis, the Court’s decision was influenced in part by a desire to maintain the “clarity and ease of application” of the Edwards rule. 512 U.S. at 461, 114 S.Ct. 2350. There, the Court held that a suspect’s request for counsel must be unambiguous to trigger law enforcement’s duty to terminate questioning. Id. at 459, 114 S.Ct. 2350. The Court refused to require police to cease interrogation upon a suspect’s ambiguous request for counsel because to do so would force police officers “to make difficult judgment calls about whether the suspect in fact wants a lawyer even though he has not said so, with the threat of suppression if they guess wrong.” Id. at 461, 114 S.Ct. 2350; see also Moran v. Burbine, 475 U.S. 412, 425, 106 S.Ct. 1135, 89 L.Ed.2d 410 (1986) (upholding a suspect’s waiver of his rights where law enforcement failed to tell him that his lawyer attempted to reach him, because not upholding the waiver would “mudd[y] Miranda’s otherwise relatively clear waters” and “spawn” a host of legal questions).
Perhaps most telling is the Court’s recent decision in Dickerson, reaffirming Miranda. The Dickerson Court confronted the question of whether Miranda had to yield to a Congressional statute that dispensed with Miranda’s required warnings by instead embodying a totality-of-the circumstances test for assessing the volun-tariness of a criminal defendant’s confes*432sion. Dickerson, 530 U.S. at 428, 120 S.Ct. 2326. In declining to overrule Miranda, the Court held that the popularly known “Miranda rights” had been broadly embedded in society. Id. at 443-44, 120 S.Ct. 2326. In addition, Miranda constituted a rule that could be more easily administered by law enforcement, compared to the one enacted by Congress: “[Experience suggests that the totality-of-the-eircum-stances test which [the statute] seeks to revive is more difficult than Miranda for law enforcement officers to conform to, and for courts to apply in a consistent manner.” Id. at 444, 86 S.Ct. 1602.
Thus, it is evident that the Supreme Court regards clarity and ease of application as necessary features of any legal rule governing custodial interrogations and that the Court eschews proposed modifications that fail this test.
C.
To hold as the majority does today, that a suspect may re-initiate interrogation through a third party, contravenes the reasoning of the Supreme Court’s custodial-interrogation jurisprudence by inviting the very uncertainty and complexity into the circumstances surrounding a suspect’s waiver of his rights that the Court has sought to banish. This uncertainty and complexity ensnares both the suspect and the law-enforcement officers in charge of his case.
First, the majority’s interpretation of Edwards allows the police to approach a suspect even when the suspect does not initiate further communication through a third party. While in custody, a suspect may have occasion to confer with family, friends, or other third parties. The suspect will have no reason to know that he could be opening himself up to renewed police contact if the third party misconstrues or ignores what he says and tells the police that the suspect is prepared to waive his right to counsel and talk with them. It does not take much imagination to realize the “lost in translation” problems that inhere, in the majority’s preferred locution, in “person-to-person-to-person” communications. Maj. Op. at 418.
For example, during conversations with third parties, the suspect might equivocate in his decision not to speak with the police, might speculate about whether making a statement would ameliorate his situation, and might even express agreement with advice encouraging him to tell the police what he knows. Even though such statements reveal no more than the suspect’s current mental state, and hardly constitute a request for the police to re-start their interrogation, they could easily be passed on to the police, or interpreted by the police, as declarations of the suspect’s desire to talk. Similarly, third parties’ communications with the police may be colored by their own views as to what is in the suspect’s best interests, and may not accurately reflect the suspect’s own intentions on how to proceed.
The majority’s reliance on cases from the Eighth, Ninth, and Eleventh Circuits upholding third-party initiations as permissible under Edwards only serves to illustrate the danger of such a rule. In Owens v. Bowersox, 290 F.3d 960 (8th Cir.2002), United States v. Gonzalez, 183 F.3d 1315 (11th Cir.1999), and United States v. Gaddy, 894 F.2d 1307 (11th Cir.1990), there was no evidence about what the suspect said to the third party to make the latter believe that the suspect wanted to talk, nor was there any discussion of the accuracy and reliability of the third party’s representations to the police. In Michaud, there was no evidence that the defendant told her cellmate that she wanted to talk to the police; the cellmate apparently concluded on her own that Michaud “needed *433to talk to somebody” as a result of Mi-chaud’s distraught state, and further concluded that a police officer was the necessary “somebody.” Michaud', 268 F.3d at 739-40 (Reinhardt, J. dissenting).
While acknowledging the potential for errors in the transmission of information from the suspect to a third party to the police, the majority minimizes this concern by arguing that even if a suspect tells the police that he wants to talk, the corrections officer with whom the suspect initiates communication will seldom be the investigating officer assigned to his case. The majority therefore reasons that a suspect’s desire to talk will almost invariably travel through a third party, namely, a third-party corrections officer. I agree. The majority is quite right when it says that “[djetectives and investigating officers do not typically act as guards roaming all day the areas directly adjacent to holding cells.” Maj. Op. at 433.
Where I part company with the majority is in its assumption that a suspect’s communication to a law-enforcement officer, even one that is not assigned to his case, counts as an initiation through a third party. It makes no sense to say that a third-party corrections officer and a third-party family member, friend, or cell-mate are the same thing for purposes of applying Edwards. Indeed, a corrections officer is not a “third party” at all.
It goes without saying that the relationship between the suspect and the police— defined broadly to include corrections personnel beyond just the investigating officer — is adversarial. Is it possible, as the majority muses, that a suspect might make statements to a prison guard that the guard misinterprets as “evincing] a willingness and a desire for a generalized discussion about the investigation”? Brad-shaiv, 462 U.S. at 1045-46, 103 S.Ct. 2830. Of course. But that is the risk the suspect takes when he voluntarily chooses to discuss his case with a law-enforcement officer. Such a situation does not involve a “third party” in the relevant sense because the suspect “himself’ has initiated communication “with the police,” even if he has not initiated communication directly with the investigating officer responsible for his case, which Edwards, on its face, does not require. For these reasons, the majority misses the mark when it contends that a suspect might prefer to initiate communication through his mother, rather than through his prison guard, because “[f]rom the suspect’s perspective,” his mother is the more trustworthy party. A suspect who tells his guard that he wants to talk is not initiating communication to the police “through” anyone; he is speaking directly to them.
As we have seen, the majority is not interested in examining the nature of the communication between the suspect and the third party, nor is it troubled by the way in which a suspect’s intentions could be misinterpreted or ignored by a third party. Instead, the majority contends that any concerns about the accuracy of the communication from the suspect to the third party to the police can be easily put to rest by having the police confirm the suspect’s “willingness and desire” to talk before renewing their interrogation. Bradshaw, 462 U.S. at 1045-46, 103 S.Ct. 2830 (holding that an initiation occurs when a suspect makes a statement that evinces “a willingness and a desire for a generalized discussion about the investigation”); United States v. Whaley, 13 F.3d 963, 966 (6th Cir.1994).
According to the majority then, once the police have learned from a third party that a suspect wishes to waive his rights and talk, the police may resume contact with the suspect to verify the third party’s representation. The majority fails to expli*434cate the content of this “limited inquiry” that law enforcement may conduct. More importantly, the majority turns the Edwards rule on its head by holding that a suspect initiates communication with the police only after the police ask him if he wants to talk: An initiation occurs “[w]hen the police receive information that a suspect wants to talk; when there is a sufficient basis for believing its validity; and when the police confirm with the suspect the validity of that information.” Maj. Op. at 424-25 (emphasis added). As explained above, the police cannot approach a suspect to discuss the case unless an initiation by the suspect has already occurred. Oregon, 462 U.S. at 1039, 103 S.Ct. 2830; Johnson, 400 F.3d at 187. It is the suspect’s initiation that gives the police license to ask him if he is prepared to waive his rights and answer questions; police contact prior to an initiation, or for purposes of sparking an initiation, is a clear violation of Edwards. Thus, contrary to the majority’s assertion, the better reasoned view is that a third-party initiation occurs, if at all, when the third party conveys to the police the suspect’s alleged desire to talk without the aid of counsel.
Here we encounter the second layer of uncertainty and complexity that plagues third-party initiations, this time for law enforcement. The majority says that the police will have to have “a sufficient basis for believing” the third party’s representations before they may resume contact with the suspect. We are left to guess about what constitutes a “sufficient basis,” and so are the police. Even though the Court’s opinion lands law enforcement on uncertain terrain, the majority chooses not to spell out any rule to guide police conduct in responding to an alleged third-party initiation. Of course, any attempt to do so immediately runs afoul of the Supreme Court’s admonition that the rules governing custodial interrogation should not put law-enforcement officers in the position of making “difficult judgment calls.” Davis, 512 U.S. at 461, 114 S.Ct. 2350. But that is exactly what will happen.
To overcome the “heavy” presumption against a custodial suspect’s waiver of his rights, Moran, 475 U.S. at 450-51, 106 S.Ct. 1135 (Stevens, J. dissenting), the prosecution will have to show that the police had a reasonable and good-faith basis for believing that the suspect communicated an intention to talk through the third party. This, in turn, will compel the police to verify that a conversation between the suspect and the third party actually occurred, and to assess the credibility and motivations of the third party. In addition, the police will have to examine the reliability of the purported waiver by probing the third party about exactly what the suspect said and in so doing, exercise their discretion in deciding whether the suspect’s statements sufficiently express a “willingness and desire” to talk such that police may approach him. Among the more obvious questions that the majority’s opinion will relegate to police officers include: Who may be regarded as a proper third party agent? What words must a suspect use to show that he is prepared to waive his rights? Is it enough for the suspect to agree with a third party’s recommendation that he talk, or must he affirmatively say that he “will” talk? What if the suspect says he will talk but never instructs the third party to so advise the police?
The Supreme Court instructed us in Davis that it is asking too much of law-enforcement officers to require them to assess whether a suspect is invoking his right to counsel even when the suspect is in the presence of police officers who can directly observe him and hear his statements first-hand. If that is the case, then *435I fail to see how it is not asking too much of law enforcement to assess a suspect’s intention to waive his rights where that purported intention is not communicated openly to the police, but is relayed through a third party.
Ultimately, of course, the complicated questions surrounding third-party initiations will end up in the courts. To sort out the mess, courts will be compelled on a case-by-case basis to develop new standards for adjudicating the propriety of third-party initiations. In the end, we can safely conclude that the hallmark “clarity” and “certainty of [] application” of the Edwards rule will be lost. Minniek, 498 U.S. at 161, 111 S.Ct. 486.
D.
Finally, it cannot be ignored that the majority’s opinion creates a glaring asymmetry in the law of custodial interrogations. A suspect cannot invoke his right to counsel through a third party. It would be impossible for a suspect to satisfy Davis’s “unambiguous” requirement for invoking the right to counsel without the suspect himself telling the police that he wants a lawyer. Davis, 512 U.S. at 459, 114 S.Ct. 2350 (stating that a suspect “must articulate his desire to have counsel present sufficiently clearly that a reasonable police officer in the circumstances would understand the statement to be a request for an attorney”). Moreover, the Supreme Court has held that a suspect’s own lawyer cannot protect his right to counsel for him. Moran, 475 U.S. at 425, 106 S.Ct. 1135 (holding that the police were not required to refrain from questioning the suspect, even where the suspect’s attorney contacted the police to ensure that he would not be questioned outside the presence of counsel, because the suspect himself did not tell the police that he wanted a lawyer).
Here, the majority adopts the opposite approach: A third party who could not invoke the accused’s right to counsel may nonetheless play a crucial role in bringing about the waiver of that right. It makes no sense to apply different standards to the invocation of constitutional rights and the waiver of those same rights. Nonetheless, such is the paradox the majority has wrought.
The majority rejects this characterization on the grounds that “[t]he symmetrical opposite of ‘invoking the right to counsel,’ is not initiating a discussion but rather waiving the right to counsel.” Maj. Op. at 424 (emphasis in the original) (internal citation omitted). True, initiation alone does not amount to a waiver that enables the police to re-start their questioning. A suspect must both initiate communication with the police and knowingly and intelligently waive his right to counsel. Although initiation and waiver are separate inquiries, a proper initiation (i.e., where the suspect initiates a dialogue with the police in which he expresses “a willingness and a desire for a generalized discussion about the investigation”) is indispensable to finding a valid waiver. Edwards, 451 U.S. at 486, n. 9, 101 S.Ct. 1880 (stating that the validity of the waiver depends on whether it was knowing and intelligent under the totality of the circumstances, “including the necessary fact that the accused, not the police, reopened the dialogue with the authorities”). Indeed, a knowing and intelligent waiver will not cure an improper initiation, i.e., an initiation at the behest of the police. As the Supreme Court stated in McNeil v. Wisconsin, 501 U.S. 171, 177, 111 S.Ct. 2204, 115 L.Ed.2d 158 (1991), “[i]f the police do subsequently initiate an encounter in the absence of counsel ..., the suspect’s statements are presumed involuntary and therefore inadmissible as substantive evidence at trial, even where the suspect *436executes a waiver and his statements would be considered voluntary under traditional standards.”
Thus, because initiation remains bound up with the suspect’s waiver, and because the majority holds that the initiation can occur through a third party, the majority’s opinion does indeed give third parties an authority to act on the suspect’s behalf in bringing about the waiver of his constitutional rights that those same third parties are not accorded when it comes to invoking the suspect’s rights in the first place. Under the majority’s holding, the police may renew contact with a suspect based on a third party’s representations that the suspect wants to talk without the aid of counsel. In contrast, the police need not terminate their questioning based on a third party’s representations that the suspect wants a lawyer. If the law demands that the suspect act on his own in requesting a lawyer, then I cannot fathom why he is any less competent to act on his own in rescinding that request, the necessary first step of which is initiating communication with the police.
II.
Having explained why allowing third-party initiations constitutes a departure from Edwards and the Supreme Court’s custodial-interrogation jurisprudence generally, I turn now to the question of whether, even if Edwards did not prohibit third-party initiations, there is sufficient evidence to conclude that Van Hook initiated communication with the police through his mother.
The majority correctly notes that on ha-beas review, we presume that the factual findings of the Ohio Supreme Court are correct. We may set aside those findings only if one of the eight exceptions in 28 U.S.C. § 2254(d)(l)-(8) (the pre-AEDPA statute that governs our review) is satisfied. Van Hook argues that subsection (8) applies because the record does not support the Ohio Supreme Court’s factual determination that he initiated communication with the police through his mother. I agree.
When Van Hook was apprehended by the Florida authorities on April 1, 1985, he invoked his right to counsel. That same day, the Florida police officers alerted Cincinnati Detective William Davis, who was in charge of the investigation into Selfs murder, that they had taken Van Hook into custody. Detective Davis and his Homicide Squad colleague, Kerry Rowland, immediately flew to Fort Lauderdale and went straight to the Broward County jail. There is no dispute that before he began his interrogation, Detective Davis knew that (1) Van Hook had invoked his right to counsel, (2) Van Hook himself did not contact either Cincinnati or Florida law-enforcement personnel to tell them that he wanted to talk, and (3) Van Hook did not ask anyone else to contact Detective Davis on his behalf to ask Detective Davis to come to the Florida jail. Despite these uncontroverted facts, the majority concludes that it was entirely consistent with Edwards for Detective Davis to have arranged for Van Hook’s transfer to an interrogation room at the Florida jail, and once there, to have advised Van Hook that he and Detective Rowland “had a lot to talk to him about.” No Edwards violation occurred, according to the majority, because on the day he flew to Fort Lauder-dale, Detective Davis spoke to Van Hook’s mother, who said something that made Detective Davis believe that Van Hook wished to speak with him.
Here the problems with third-party initiations discussed in Part I come alive. All the record establishes is that Detective Davis believed, after talking with Van *437Hook’s mother, that Van Hook was willing to make a statement. We do not know what Van Hook’s mother said to Davis that convinced Davis that Van Hook wanted to talk. We do not know, for instance, whether Van Hook’s mother suggested to Davis that Van Hook would talk because (1) she had urged Van Hook to cooperate with the police, (2) her familiarity with her son and his mental state led her to believe that he wanted to talk, or (3) Van Hook unambiguously told her that he wanted to confess. Van Hook’s mother may have misconstrued his comments, she may have been speaking out of her own subjective beliefs and motivations, or she may have accurately reported Van Hook’s statements, which may or may not have been sufficient to show Van Hook’s “willingness and [ ] desire for a generalized discussion about the investigation.” Bradshaw, 462 U.S. at 1045-46,103 S.Ct. 2830. The point is that we do not know because the record is silent on these issues.
Similarly, we do not know what Van Hook said to his mother that prompted her to tell Detective Davis whatever she told him. Although Van Hook told Detective Davis in the Florida interrogation room, “I talked to my Mom today an’ she jus’ told me, you know, be cooperative an’ jus’ tell the truth,” this statement tells us only what Van Hook’s mother recommended that he do. It in no way establishes that Van Hook told his mother that he wanted to talk with the police, nor does it establish that he asked his mother to inform the police of his desire to talk.
The faulty factfinding of the Ohio Supreme Court in concluding that the record supported the proposition that Van Hook initiated communication through his mother did not go unnoticed by the dissenting members of that court. Justice Wright aptly summarized his colleagues’ reasoning:
There is not a scintilla of evidence in the record to show that the appellant initiated further discussion. The court justifies its holding by drawing one inference from another. It infers that the appellant initiated further discussion from the inference about what appellant’s mother supposedly stated to Officer Davis. This is guesswork about facts which are not in the record. Such inference building undermines the constitutional foundations upon which Bradshaw, Fields, and Edwards ... are built.
Ohio v. Van Hook, 39 Ohio St.3d 256, 530 N.E.2d 883, 894 (1988) (Wright, J., dissenting).
I believe that the Ohio Supreme Court’s factual finding is not entitled to any deference from this Court because the evidence is not sufficient to show that Van Hook initiated communication with the police through his mother. Accordingly, I would grant Van Hook’s petition for habeas relief.
III.
In holding that custodial suspects may initiate communication with the police through a third party other than his attorney, the majority today repudiates the only interpretation of Edwards that is consistent with the Supreme Court’s custodial-interrogation jurisprudence. Applying its new rule, the majority infers, based on evidence that is not in the record, that Van Hook initiated communication with the police through his mother. Because I believe that the majority’s endorsement of third-party initiations impermissibly heightens the risk of constitutional error, and that such an error has occurred in this case, I respectfully dissent.