dissenting:
Today this court grants a writ of habeas corpus to Lisa Joyce Rubin, who, it concedes, unquestionably “shot and killed her husband,” and whom a Maryland jury convicted of first-degree murder. The court grants habeas relief to Rubin (an honors college graduate) on the theory that she received ineffective assistance of counsel at her trial. Yet for six months prior to that trial, a team of three lawyers specializing in criminal law planned her defense and formulated trial strategy; this team represented Rubin at trial, examining witnesses, making objections, and asserting legal arguments on her behalf. The majority does not fault the skill or ethics of these three lawyers, whom the state court described as “highly qualified and experienced criminal defense counsel.” Rather, the majority grants habeas relief because of a conflict of interest of two other attorneys, who advised Rubin in the hours and days immediately after she killed her husband. Although the two conflicted lawyers continued to meet with Rubin, they did not try her case, direct trial strategy, sit at counsel table, or even enter an appearance on her behalf in the trial court. The Supreme Court has never established — let alone clearly established — that a conflict of interest violates the Sixth Amendment under these circumstances.
This is not to say that the two conflicted lawyers — David Gavin and Darryl Longest — acted properly. They did not. Driven by their own financial interests rather than their client’s interests, they certainly violated the ethical codes that govern all lawyers, and they may have committed crimes. I join in the majority’s repro-bation of their conduct. Indeed, were I left to my own devices to determine when a conflict of interest yields constitutionally ineffective assistance of counsel, I might endorse the conclusion that the existence of case-related contact between the conflicted and the unconflicted lawyers tainted the latter. But we are not free to do that. 'Instead, we must apply controlling law, including the deferential standard of review in the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA), 28 U.S.C.A. § 2254(d)(1) (West Supp.2001), which requires us to deny habeas relief unless we find a state court’s decision is “contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court.” In this case, I cannot so find. •Accordingly, I must respectfully dissent from the majority’s contrary holding.
I.
The sole Supreme Court case cited as “the clearly established Federal law” authorizing habeas relief here is Cuyler v. Sullivan, 446 U.S. 335, 100 S.Ct. 1708, 64 L.Ed.2d 333 (1980). See ante at 402, and 406 n. 3. That case dealt with a conflict of interest involving dual representation — the same two attorneys represented three co-defendants, with conflicting interests, at three separate criminal trials. Id. at 337-40, 345-49. Sullivan provides no direct support for a finding of unconstitutional conflict of interest on the very different facts at issue here. Indeed, to the extent Sullivan has any specific relevance to this case, it suggests no conflict exists here. See infra n. 5. Under circuit precedent, however, Sullivan does establish the relevant standard for assessing conflict of in*408terest claims like the one at issue here. See United States v. Tatum, 948 F.2d 370, 375-76 (4th Cir.1991); but see Beets v. Scott, 65 F.3d 1258, 1272-(5th Cir.1995) (en banc) (holding that the standard set forth in Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), rather than Sullivan’s “more rigid rule,” applies to “attorney breaches of loyalty outside the multiple [or successive] representation context”). Sullivan requires that “a defendant who raised no objection at trial [like Rubin] must demonstrate that [1] an actual conflict [2] adversely affected his lawyer’s performance.” 446 U.S. at 348, 100 S.Ct. 1708. The majority iterates .and reiterates Gávin and Longest’s actual conflict, but offers only two reasons why this conflict “adversely affected Rubin’s representation.” Ante at 404. Neither is at all persuasive.
A.
First, the majority contends that “Longest and Gavin’s direct representation of Rubin was adversely affected by their conflict of interest.” Id. (emphasis added). Specifically, the majority agrees with the district court that Longest and Gavin should have testified at Rubin’s trial to “draw the spotlight on their own actions.” Rubin v. Gee, 128 F.Supp.2d 848, 869 (D.Md.2001).
To demonstrate “adverse effect,” however, a defendant must not only “identify a plausible alternative defense strategy his defense counsel might have pursued,” but also must: (i) show that this strategy was “objectively reasonable,” i.e., “clearly suggested by the circumstances,” “under the facts of the case known to the attorney at the time of the attorney’s tactical decision,” and (ii) establish a link between the conflict and the failure to pursue the strategy-. Mickens v. Taylor, 240 F.3d 348, 361 (4th Cir.2001) (en banc) (citation omitted), aff'd, — U.S. -, 122 S.Ct. 1237, 152 L.Ed.2d 291 (2002). The majority utterly fails to demonstrate that the proffered alternative defense strategy — calling Longest and Gavin to testify — was “objectively reasonable” or not “pursue[d]” because of Longest and Gavin’s conflict. Id. Thus, even if we were free to ignore the stringent standard of review mandated by AEDPA and to ask simply whether the decision not to call Longest and Gavin had an adverse effect on Rubin’s representation, Rubin could not prevail on this claim. Taking account of the limitation AEDPA places on our review, there is no way to conclude that the state court’s contrary determination was unreasonable.
Testimony from Longest and Gavin simply would ■ not have helped Rubin and, ■therefore, offering this testimony at trial would not have been “objectively -reasonable.” Id. At. the state post-conviction hearing, Rubin’s two surviving trial lawyers (one had died before the post-conviction proceeding occurred) recounted their struggles to formulate a defense theory around Rubin’s shifting and incomplete account of the night she shot her husband. They explained, at length, the multiple problems with introducing testimony by Longest and Gavin, as well as the incompatibility between such testimony and their trial strategy. Their explanation establishes beyond doubt that calling Longest and Gavin, though superficially plausible, was not an “objectively reasonable” strategy that was “clearly suggested by the circumstances.” Id. (citation omitted).
First, testimony from Longest and Gavin, who knew a great deal about Rubin’s activities and statements after the killing and at other times, would have opened the door to much damaging evidence. After Rubin shot her husband, she first telephoned Robert Miller, a private detective *409with whom she had a “personal relationship” including late night dinners, assistance in finding a new home, and personal confidences. See Rubin v. State, 325 Md. 552, 602 A.2d 677, 679 (1992); Rubin v. Gee, 128 F.Supp.2d at 852 (quoting closing argument). Miller arrived at the scene with his assistant, another private detective named Robert Leopold, viewed the body, and telephoned Longest. Not until an hour after the detectives’ arrival did Longest and Gavin arrive at the scene.
If Longest and Gavin had testified, the jury would have heard that right after Rubin repeatedly shot her husband, reloaded, fired five additional bullets into his back as he lay face down, and reloaded again — all supposedly in a frenzy of terror — she developed the presence of mind to stay in the vicinity for an extended period to consult at length with her private detectives and lawyers. The jury would have heard details including the three hours Rubin spent at the scene without calling an ambulance. With this fodder, the prosecutor could have extensively elaborated on his brief comments on the point that the first person Rubin called was not a lawyer but Robert Miller, her intimate.1 We can ascertain this much even from the minimal information that Longest and Gavin have -disclosed to date. See Rubin v. State, 602 A.2d at 682-84; Answer to Petition for Habeas Corpus, Ex. 2 at 96 et seq. (containing the scant post-conviction testimony by Longest concerning the events of the night of the killing and no testimony by Gavin).
In addition, a substantial portion of the night’s activities has never emerged, because Rubin did successfully assert the privilege with respect to Longest and Gavin, and with respect to her communications with Miller and Leopold after Longest and Gavin arrived at the scene of the shooting. We cannot know what other damaging statements or observations, made after the privilege arose, Longest and Gavin (and Miller, Rubin’s intimate) would have had to relate, if Rubin had waived the privilege. We do know that without sacrificing the privilege as defined by the trial court,2 or otherwise harming Longest or Gavin’s interests in the slightest, the defense could have called Rubin’s private detective (and good friend), Miller, to describe his several lengthy one-on-one conversations with Rubin at the scene before Longest and Gavin arrived. See Answer to Petition, Ex. 2 at 173, Ex. 4 at 54-57, 70. If the defense had believed such testimony about Rubin’s behavior or conversation soon after the killing would have assisted her, Miller, an obviously friendly witness, was certainly available to testify. Yet the defense chose not to open that door, in a decision that no one does, or could, attribute to a desire to protect Longest and Gavin.
To be sure, the prosecution made good use of the missing pieces in Rubin’s account, pointing out that she had left the scene of the shooting and had checked herself into the hospital under a false name; and if Longest and Gavin had testified the jury would probably have heard that they told Rubin to leave and use an *410abas. But see supra n. 1. But the same prosecutor who used the missing pieces in Rubin’s story to the State’s advantage could certainly have made hay out of Longest and Gavin’s more detailed account of Rubin’s behavior — amplified by whatever additional facts a waiver of privilege might have revealed through testimony by Longest, Gavin, or Miller.
Moreover, even if calling Longest and Gavin had been a more attractive option, the decision not to cah them was part of the broader defense strategy aimed at allowing Rubin to escape conviction altogether. Rubin’s team of trial lawyers explained that they focused on asserting attorney-client privilege as a way to exclude the testimony of her other private detective, Robert Leopold, which, they believed, was so uniquely damaging to Rubin that if it could be excluded as privileged, Rubin would walk free.3 Of course, as it turned out, the trial court permitted Leopold to testify, and on appeal this was held to be only harmless error. See Rubin v. State, 602 A.2d at 684-90. But at the time, trial counsel hoped to exclude Leopold’s observations, leaving a far greater area of doubt as to events surrounding the shooting.
Trial counsel’s lucid explanations, both of the trial strategy they followed and of the danger to Rubin in the strategy she now suggests, clearly show why testimony by Longest and Gavin was not an “objectively reasonable” alternative strategy that the circumstances at trial “clearly suggested.” Mickens, 240 F.3d at 361. In sum, the pitfalls involved in Longest and Gavin’s testimony about the night of the killing and the possibility of excluding Leopold’s testimony and leaving massive gaps in the State’s case demonstrate that under any standard, Rubin has not shown that the decision not to call Longest and Gavin had an adverse effect on her defense.4
Applying the stringent AEDPA standard of review, as we must, renders this conclusion inescapable. On habeas review we may only reverse the state post-conviction court if its determination was “contrary to, or involved an unreasonable application of’ clearly established federal law. 28 U.S.C.A. § 2254(d)(1). In this case, the state post-conviction appellate court concluded that “it is understandable why [Longest and Gavin] were never called to the stand by the defense team.” Considering the perils on the road not taken, I can only agree. In any event, the state post-conviction court’s conclusion *411that Rubin suffered no adverse effect from the decision not to call Longest and Gavin — that calling them was not an “objectively reasonable” strategy that was “clearly suggested by the circumstances” — was certainly not itself objectively unreasonable.
Finally, and most basically, the majority’s contention — that “Longest and Gavin’s direct representation of Rubin was adversely affected by their conflict of interest” because they determined they would not testify at trial — rests on a false premise. Ante at 404-05. In fact, Longest and Gavin did not determine that they would not testify. Rather, as the state court expressly held, the “choice to offer [their] testimony was neither Longest’s or Gavin’s to make.” Instead, like all “trial strategy” decisions, the state court held that the decision as to whether Longest and Gavin would testify was made by the three “outstanding attorneys,” Rubin v. Gee, 128 F.Supp.2d at 862, who represented Rubin at trial. The majority expressly states that it does “not call into question the state court’s factual finding that Rubin’s trial lawyers made the decision to not call Longest and Gavin as witnesses.” Ante at 405. Given this state-court finding, Rubin has not established and cannot establish the necessary “link[ ]” between Longest and Gavin’s “actual conflict” and the decision she challenges. See Mickens, 240 F.3d at 361.
Therefore, the majority’s first reason for concluding that Longest and Gavin’s conflict of interest "adversely affected Rubin's representation," ante at 404, unquestionably fails
w
The majority's only other rationale for concluding that Longest and Gavin’s conflict of interest adversely affected Rubin’s representation is that this conflict assert-edly.tainted Rubin’s three trial counsel, Barry Helfand, Alan Goldstein, and Fred Joseph. The majority does not suggest that any member of the three-man trial team, himself, had any conflict of interest with Rubin. Nevertheless, the majority concludes that the conflict of interest of non-trial counsel (Longest and Gavin) infected the trial team. Maybe so, but the Supreme Court has never endorsed such a theory.5
In holding to the contrary, the majority heavily relies on our precedent, United States v. Tatum, 943 F.2d at 373-79. Recently, however, the Second Circuit reached the opposite conclusion, finding that “because [a conflicted lawyer] had no input into trial strategy, his participation [at trial] does not require per se reversal.” Triana v. United States, 205 F.3d 36, 42 (2d Cir.2000). But see Tatum, 943 F.2d at 378 (“From the necessary conclusions that [the conflicted lawyer] was an acknowledged source of information to [the non-conflicted lawyer] and [the latter] ... felt the need to’ have [the conflicted lawyer] present throughout the trial, we conclude that the presentation of [the] defense was *412the product of both ..., even though their contributions may have varied in function and degree.”).
Although I certainly prefer our approach, in light of the Supreme Court’s silence on the point, it is impossible to conclude that it was “clearly established,” as a matter of Supreme Court law, that our rule must be followed. See 28 U.S.C.A. § 2254(d)(1); Williams v. Taylor, 529 U.S. 362, 412, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000). Indeed, the circuit split exhibited by Triana and Tatum testifies to the absence of any Supreme Court case that resolves the question. Nor can I conclude that the state court’s application of Sullivan, echoing that of the Second Circuit in Triana, was objectively unreasonable.
Even if we could rely on our circuit precedent (ignoring well-reasoned contrary precedent of a sister circuit) as indicative of “Federal law, as clearly established by the Supreme Court” or as limiting objectively reasonable applications of that law, numerous distinctions between Rubin’s case and Tatum prevent reliance on Tatum here. Given these factual and legal distinctions, it is doubly difficult to see how the state appellate post-conviction court’s conclusion, that Longest and Gavin’s conflict did not render the assistance of Rubin’s three trial counsel constitutionally ineffective, was an objectively unreasonable application of federal law as established by the Supreme Court.
First, the nature of Longest and Gavin’s work for Rubin dramatically differed from that of the conflicted lawyers in Tatum (and Triana) for their clients. The taint we considered in Tatum — from a conflicted lawyer who appeared at trial, “helping” trial counsel “throughout the presentation for this trial” to “bring [trial counsel] up to speed” — is simply not implicated in Rubin’s case. Tatum, 943 F.2d at 378; see also Sullivan, 446 U.S. at 342, 100 S.Ct. 1708 (noting that the conflicted lawyers involved “played important roles at all three [defendants’] trials”). Rather, the surviving two members of the trial team have unequivocally testified — and the state post-conviction court has found — that they, not Longest and Gavin, made Ml the strategic trial decisions, in accord with their own views of “the best interest of Lisa Rubin,” and that they made no decisions in order to protect Longest and Gavin.6 As one of them testified, “[N]ever, at any time, was there any indication whatsoever that any discussion of strategy was based upon any relationship ... [t]hat any of us had with anyone in the world other than [Rubin].” Consistent with this account, neither Longest nor Gavin planned the course of the trial, delivered an opening statement or closing argument, examined a witness, or took part in the trial in any way. Indeed, far from guiding and educating trial counsel in court as in Tatum or *413Triana, Longest and Gavin never appeared at trial or even noted their appearances in the trial court on behalf of their client.
Second, also unlike the conflicted lawyer in Tatum, who had allegedly recommended commission of the charged crime, Longest and Gavin gave relevant advice only after the crime had been committed. Cf Tar turn, 943 F.2d at 374 (noting that the conflicted attorney allegedly recommended the act in question, a bankruptcy fraud). Such advice, unlike that of the lawyer in Tatum, does not directly illuminate Rubin’s intent at the time of the act. More broadly, Rubin’s act — shooting her husband — differs greatly from the financial crime addressed in Tatum, in which advice of counsel might even have excused the entire act; in Tatum the defendant might have blamed the conflicted lawyer for recommending the very act that was the center of the trial. Factual testimony by Longest and Gavin could not have directly altered the law’s treatment of the shooting itself.
Third, Tatum, unlike the case at hand, did not involve federal review of a state conviction. The claim of the defendant in Tatum therefore simply did not implicate the “significant interest[ ] in ... federalism that underlie[s] the federal habeas scheme.” Evans v. Smith, 220 F.3d 306, 321 (4th Cir.2000). In contrast, when considering Rubin’s claim, we must bear in mind the “important interest! ] in ... respect for state court judgments that underlie[s] the statutory habeas scheme.” Id. at 323. Congress’s enactment of the AEDPA, which occurred after we decided Tatum, further heightens the importance of these federalism concerns. To overturn the reasoned decision of a state court on the basis of our precedent, rationally disputed by another circuit and neither addressed nor confirmed by the Supreme Court, would erode the balance that both our federalist Constitution and the AEEi-PA are meant to protect.
In sum, given the relevant circuit split on facts that more obviously support a finding of ineffective assistance — that is, in cases where the conflicted lawyer appeared at trial and where the facts concerning the lawyer’s role were more directly exculpatory — I cannot conclude that the state appellate post-conviction court unreasonably applied Supreme Court law by declining to rule that Longest and Gavin tainted the effectiveness of Rubin’s trial team. Indeed, scrupulous adherence to Mickens, 122 S.Ct. at 1245-46 (emphasizing the importance of actual representation of conflicting interests for the presumption of prejudice), would seem to compel denial of habeas relief in this case. For these reasons, the majority’s only remaining rationale for finding adverse effect also clearly fails.
H.
. In the months between April 24, 1990, the night Lisa Rubin shot her husband, and November 11, 1990, the day she was convicted of his murder, she employed five lawyers. Although two of them, Longest and Gavin, served her poorly on the night of the killing, three other talented lawyers took over within days of the killing and prepared her case for over six months before trial. The state court specifically found that those three “highly qualified and experienced criminal defense counsel” alone fashioned Rubin’s defense, and represented her at trial. Those three lawyers made the tactical decision to advise Rubin to assert privilege with respect to her conversations with Longest and Gavin. Those three lawyers reached this decision to avoid opening the door to everything Longest and Gavin knew about Rubin’s tangled relationship with her husband and her incriminating behavior after she killed him. Neither the majority nor the district *414court suggests that any one of those three lawyers did not skillfully represent Rubin or had an interest conflicting with Rubin’s.
The only conceivable reason anyone would think of faulting the performance of .trial counsel is some kind of taint from the two conflicted lawyers who took no part in the trial. But the Supreme Court has never held that a non-participating lawyer’s conflict could so infect otherwise un-conflicted lawyers’ representation as to render the latter’s representation constitutionally ineffective. Even when a conflicted lawyer appears at trial and handles trial work, the circuits have divided as to whether a conflicted lawyer can taint the representation of other, otherwise uncon-flicted lawyers.
Against this legal and factual background, I cannot conclude that the state court’s decision rejecting Rubin’s ineffective-assistance claim was “contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court.” 28 U.S.C.A. § 2254(d)(1). I therefore respectfully dissent.-
. The state courts never fully resolved the question whether Longest or Miller gave the instruction to use an alias in checking. Rubin into the hospital. Direction by Miller, Rubin's intimate who was not a lawyer, see Rubin v. State, 602 A.2d at 679, would obviously not have resonated to her benefit.
. Over defense objections, the trial court ruled that the attorney-client privilege arose not when Rubin's private detectives, Miller and Leopold, arrived at-the crime scene, but only later, when her attorneys, Longest and Gavin, arrived. See Rubin v. State, 602 A.2d at 683-84 (describing and affirming this ruling).
. Leopold was the only witness who testified that Rubin's purse contained bullets fitting the gun she said her husband had brought to the scene.
. The district court also held that Longest and Gavin denied Rubin effective counsel during the pretrial phase. See Rubin v. Gee, 128 F.Supp.2d at 865-68. However, the court pointed to no specific adverse effect of Longest and Gavin’s out-of-court activities after trial counsel entered the case, see id. at 866-67, and it is difficult to see how their presumed desire to ■ avoid testifying could have affected trial preparation, given the very limited out-of-court work that even the district court attributed to them. Id. at 867. I note that a grand jury indicted Rubin, so her initial waiver of indictment had no effect whatsoever on her trial. Nor is it plausible that prosecu-torial focus on Longest and Gavin before trial would have diverted the prosecutors' attention from Rubin. The prosecutors were investigating Longest and Gavin well prior to Rubin's trial, see ante at 404, making it clear to Rubin’s three other, non-conflicted lawyers that they had the option of throwing Longest and Gavin to the wolves. Moreover, in view of the fact that Rubin alone shot her husband, the prosecutors were hardly likely to immunize her or allow her to plead to an equivocal account of the killing in order to pursue her counsel for obstruction of justice. Perhaps for these reasons, the majority tellingly does not hold that Longest and Gavin denied Rubin effective counsel during the pretrial phase.
. Indeed, in Sullivan the Court specifically noted that both conflicted lawyers "played important roles at all three [defendants'] trials,” as a predicate for its holding that the defendant had proved an actual conflict of interest. See Sullivan, 446 U.S. at 342, 100 S.Ct. 1708 (emphasis added). Although for clarity's sake, I'follow the majority's lead in treating taint as an "adverse effect” question, the context of the Sullivan Court's comment suggests that actually it may be more properly-analyzed as a question of whether an "actual conflict” exists. See also Beets, 65 F.3d at 1277-78. Viewed that way, I would conclude, for the reasons set forth in text, that no "clearly' established Federal law as determined by the Supreme Court,” 28 U.S.C.A. § 2254(d), holds that counsel in the position of Rubin's three-man trial team had an actual conflict of interest.
. The state post-conviction court found that Longest and Gavin's only contribution after Rubin turned herself in was " 'paralegal/investigative' work.” The district court rejected this finding, see Rubin v. Gee, 128 F.Supp.2d at 869-70, largely on the basis of Longest and Gavin’s self-serving letter justifying their attorney fees. On examination, however, that letter simply shows that Longest and Gavin scurried around to keep Rubin happy and to retain the substantial fees she had paid them, not that they did more than "paralegal/investigative work." In any event, the majority (although discussing the letter at length) does not reject the state-court finding as to the limited paralegal nature of Longest and Gavin's work or the state-court finding that Longest and Gavin had no effect on trial strategy. See ante at 404-05, 405-06. Accordingly, we are bound by those findings. See 28 U.S.C.A. § 2254(d)(2), (e)(1). In light of them, the majority’s taint discussion rests on a sandy foundation indeed.