dissenting.
In late August, 1991, Petitioner Lonnie Wright Richie and a teenage friend, Danny Waller, committed a brutal and terrible crime. They kidnaped Mrs. Laura Laun-hardt and placed her in the closet of an abandoned house with hands and feet bound and a strap around her neck secured to the clothes rod. They took her credit and debit cards and went on a spending spree. Her dead body was found several days later. The cause of death was either asphyxiation or cardiac arrest. The question at trial was whether this was first degree, malice-aforethought murder, warranting the death penalty. The prosecution’s theory was that after securing Mrs. Launhardt in the closet, the defendant killed her by raising her legs, thus putting pressure on the blood vessels in her neck and bringing about the asphyxiation. Petitioner’s defense in both the guilt and the penalty phase of the trial was that he and his accomplice had left Mrs. Laun-hardt alive, and that death had resulted sometime later, perhaps by her slipping or fainting.
There was evidence in support of both theories. The defense relied primarily on *1129what is delicately called “entomological evidence” — namely, the size of the maggots found on her decomposing body — which indicated, according to the defense expert, that Mrs. Launhardt had not died until the evening of August 30, two days after the defendant and his accomplice left her in the closet. The prosecution relied primarily (or perhaps exclusively, as will be discussed below) on the expert testimony of Dr. Robert Hemphill, the state medical examiner, who gave the jury his opinion that “someone killed this person by partially suspending her by the neck while she was bound and face down in the position which she was found.” It turns out, however, that had Dr. Hemphill been asked, he would have testified that there was no medical evidence supporting the conclusion that Mrs. Launhardt was deliberately suspended by anyone, and that Mrs. Laun-hardt could have died as a result of passing out and falling into the strap, as the defendant contended. It is the failure of defense counsel to conduct a proper cross-examination, and thus elicit this information, that led the magistrate judge to recommend, and the district judge to hold, that Petitioner was denied effective assistance of counsel.
I. Deficient Performance
The standard for effective assistance of counsel is not demanding. The Constitution does not require optimal, or even commendable performance; it is enough if counsel provides assistance that is “reasonable considering all the circumstances.” Strickland v. Washington, 466 U.S. 668, 688, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). And we begin our analysis with a presumption that the reasonableness standard was met, that a particular representation fell “within the wide range of reasonable professional assistance.” Id. at 689, 104 S.Ct. 2052. But when counsel’s performance fails to “make the adversarial testing process work in the particular case,” we must find ineffective assistance. Id. at 690, 104 S.Ct. 2052; see Rompilla v. Beard, —U.S.-, 125 S.Ct. 2456, 162 L.Ed.2d 360 (2005). I cannot agree with the majority that defense counsel met that standard.
During trial preparation, defense counsel interviewed Dr. Hemphill and concluded that his testimony would be favorable to. the defense. According to the magistrate judge’s summary of the evidence, “From talking with Dr. Hemphill, [counsel] understood that Dr. Hemphill’s opinion was that no physical evidence supported the State’s theory that Ms. Launhardt had been killed by hanging her upside down by her feet.” R & R 9-10. As counsel stated at the evidentiary hearing, “I know when I walked out of that office I thought of him as our witness or as a defense witness.”1 She therefore- did not retain an independent expert witness to testify at trial regarding the position of the body and the manner of death.
After the trial began, but before Dr. Hemphill testified, Dr. Hemphill approached defense counsel during a recess to inform her of a change in his intended testimony, “because I didn’t want to blindside her, so-to-speak.” Hemphill Depo. 25, R. Vol. Ill 890. Dr. Hemphill had learned of a statement by Mr. Richie’s accomplice, Mr. Waller, later retracted and not admitted at trial, that death was caused by deliberately hanging the victim upside *1130down. Based on this dubious information (and not on any change of mind regarding the medical evidence) Dr. Hemphill informed defense counsel he would now testify that the probable cause of death was deliberate strangulation. Defense counsel described herself as “devastated” by this information. Rather than request a continuance, however, which would have enabled her to prepare for cross-examination, challenge the basis of Dr. Hemphill’s testimony on account of its reliance on the inadmissible statement by Petitioner’s co-defendant, or call an independent defense expert, defense counsel proceeded with trial.
On direct examination, Dr. Hemphill testified as follows:
[T]he fact that this lady was tied, her hands tied behind her back, her feet tied together, a ligature tied behind her neck and then tied to something above her and that she was partially suspended by it to me is overwhelming evidence that somebody did this to her.... So I believe this is a homicide. That is, I believe somebody else did this to her, she didn’t do it to herself.... All I can say is this evidence indicates that someone killed this person by partially suspending her by the neck while she was bound and face down in the position which she was found ... [I]t appears that — ’that as I said, someone partially suspended her, pulled her head up in this position while she was face down and accomplished the asphyxiation in that matter [sic].
R. Yol. VII 2083-84. According to the district court, this was “[t]he only evidence supporting the State’s theory” of malice-aforethought murder. Order 6.2
Since the entire defense ease as to guilt and mitigation depended on the viability of the left-alive theory, it was singularly important for defense counsel to ask Dr. Hemphill if that theory, too, was consistent with the medical evidence. And indeed, if such a question had been asked, Dr. Hemphill’s response would have been highly favorable to the defense. Notwithstanding his trial testimony, Dr. Hemphill believed there was no “good, measurable, hard, objective evidence” to indicate how Mrs. Launhardt died. Hemphill Depo. 16-17, R. Vol. Ill 881-82; see also id. at 33, R. Vol. Ill 898. He testified in his deposition that his opinion at trial, rather than being based strictly on the medical evidence, was influenced by the (now-discredited) statement by Mr. Waller that Mr. Richie had deliberately killed her by lifting her legs. As the magistrate judge summarized Dr. Hemphill’s deposition testimony at the evidentiary hearing: “Dr. Hemphill, in his deposition taken on June 27, 2003, maintained that there was no scientific evidence to support a conclusion that the compression occurred by someone holding the victim’s legs to cause death.... Dr. Hemphill informed [defense counsel] and continues to maintain in his most recent deposition, that no medical evidence existed to support the State’s theory of death. According to Dr. Hemphill, absent the evidence of the co-defendant, there was no evidence that the victim was deliberately suspended by anyone.” R & R 13-14.
Yet on cross-examination, defense counsel asked Dr. Hemphill only four questions: ■
Q First of all, Doctor, you have no medical evidence to show that the death occurred anywhere but Pawnee County; is that correct?
A That’s correct.
Q Secondly, the larvae or maggots that you’ve testified about from your report, *1131they vary in length from 2 millimeters to 1 centimeter?
A Let me just double check my notes. Yes, that is correct.
Q And you did not collect any of these; is that right? Preserve any of them? A That’s correct.
Q And finally, Doctor, you have no medical evidence that is inconsistent with Ms. Launhardt being tied, restrained in that closet in a sitting or standing position, correct?
A That’s correct.
R. Vol. VII 2103-04.
Defense counsel did not ask whether the evidence was consistent with the possibility that the victim had been left alive, and the jury therefore did not receive the information that Dr. Hemphill viewed the scientific evidence as supporting the defendant’s theory as well as the prosecution’s. As the district court noted, “Even Dr. Hemphill expressed surprise during his deposition testimony that [defense counsel] had not asked him about the reasonableness of the defense theory during cross-examination. He indicated a willingness to testify, if he had been asked, that the findings at the scene of the crime were consistent with the defense’s theory.” Order 6-7.
The district judge had no difficulty in concluding that this performance was constitutionally deficient, calling counsel’s cross-examination of Dr. Hemphill “woefully inadequate.” Order 8. “The inadequacy of the cross-examination,” the court found, “is evident ... from the trial record itself.” Id. The court elaborated:
There was no measurable scientific evidence to support either the State’s theory or the defense theory. Dr. Hemphill was willing to testify that the defense’s theory provided a reasonable explanation for the circumstances under which the victim was found. [Defense counsel], however, failed to present this significant fact to the jury through the testimony of Dr. Hemphill. The State’s presentation of its “hanging” theory through Dr. Hemphill’s testimony was allowed to go unchallenged.
Order 7-8 (internal citations and footnote omitted). This was consistent with the magistrate judge’s evaluation:
The trial proceeded, and [defense counsel] asked virtually no questions of Dr. Hemphill on cross-examination. [Defense counsel] completely failed to ask if the Petitioner’s theory of the case was consistent with the medical evidence. Although the sole premise of Petitioner’s argument was that when he left Ms. Launhardt she was alive, [defense counsel] presented no medical evidence that this theory was consistent with the medical evidence.
R & R 24. The district court thus held that defense counsel provided ineffective assistance in violation of the Sixth Amendment.
The majority reverses that decision. In doing so, the majority maintains: (1) that Dr. Hemphill’s testimony was not the only support for the State’s theory of deliberate strangulation (Op. at 1123-24); (2) that Dr. Hemphill’s testimony was less absolute, and hence less damaging, to the defense than the district court allowed (Op. at 1124); (3) that defense counsel’s failure to question Dr. Hemphill more extensively was a strategic choice to which the court must defer (Op. at 1124); and (4) that far from being constitutionally deficient, defense counsel’s cross-examination was “aggressive and successful” (Op. at 1124). I must respectfully disagree on all points.
The majority begins by holding that the district court “clearly erred in finding that Dr. Hemphill provided the only evidence supporting the state’s theory.” Op. at *11321124. The majority points to the testimony of a police officer regarding the position of the victims’s dress and to pictures of the crime scene, showing that when the body was found, part of the victim’s skirt had fallen upward, obscuring the ligatures that bound her hands behind her back. The majority hypothesizes that this evidence regarding the position of the victim’s dress “also provided evidence supporting the state’s theory.” Op. at 1124.
It is not obvious why this point, even if valid, is relevant to whether defense counsel’s cross-examination of Dr. Hemphill was inadequate. If there was other evidence supporting the verdict, that might go to the question of prejudice; but it would hardly refute the district court’s conclusion regarding the quality of defense counsel’s performance.
But putting that aside, this line of argument regarding the position of the victim’s clothing is an appellate concoction, never argued to the jury. In the pictures discussed by the witnesses, Mrs. Launhardt’s skirt had been moved to display the ligatures on the victim’s wrists. Neither witness was asked about the significance of the position of her clothing, and neither commented on it independently. The prosecutor did not argue in closing that the position of the dress was significant. The prosecution’s apparent lack of interest in this evidence is most clearly reflected in its passivity when the judge excluded pictures showing the dress in its upward position. Had evidence of the position of the dress been significant to the government’s theory, the prosecutor would have argued as much when the trial judge determined which pictures to admit and which to exclude as duplicative. Since the argument on which the majority now relies was not presented to the jury, the majority is wrong to hold that the district court committed “clear error” — or any error at all— in finding Dr. Hemphill’s testimony to be the only evidence supporting the government’s theory.
Second, the majority downplays the significance of Dr. Hemphill’s testimony, stating that “Dr. Hemphill’s witness-stand assertions were hardly absolute, leaving both sides room to maneuver.” Op. at 1124. The majority goes so far as to state that:
Dr. Hemphill’s sole testimony relating to the manner of death occurred in the following exchange.
Q: (By Mr. Gillert [the prosecutor]) Let me ask you this hypothetical: If someone were to bind her hands behind her and bind her feet as she was found, tie the ligature around her neck and tie it to the poll with the length of the rope, would it be consistent with your findings that they suspended the person by some lower extremity while they hung to affect the death in the manner in which you found, is that consistent?
A: Yes. As far as I can see, it would be consistent.
Op. at 1122.
Once again, it is not clear why this goes to the quality of defense counsel’s performance rather than to the question of prejudice. But more importantly, I do not think the majority’s characterization of Dr. Hemphill’s testimony is accurate. His answer to Mr. Giller’s hypothetical was most certainly not his “sole testimony relating to the manner of death.” Well before Mr. Giller posed his hypothetical, Dr. Hemphill testified flatly that “this evidence indicates that someone killed this person by partially suspending her by the neck while she was bound and face down in the position which she was found.” R. Vol. YII 2084. This was in direct contradiction to the defense theory that death occurred some 48 hours after she was left in the closet in an upright position, when she slipped or *1133fainted. He then reiterated: “[I]t appears that — that as I said, someone partially suspended her, pulled her head up in this position while she was face down and accomplished the asphyxiation in that matter [sic].” Id. If that was not enough, Dr. Hemphill told the jury: “[T]he fact that this lady was tied, her hands tied behind her back, her feet tied together, a ligature tied behind her neck and then tied to something above her and that she was partially suspended by it to me is overwhelming evidence that somebody did this to her.... So I believe this is a homicide. That is, I believe somebody else did this to her, she didn’t do it to herself.” R. Vol. VII 2083. By the time the jurors heard Dr. Hemphill’s assent to the prosecutor’s hypothetical, they had heard him testify over and over again that someone had killed the victim. Far from “leaving both sides room to maneuver,” Dr. Hemphill’s testimony, if unchallenged, left little room for the defendant’s left-alive theory.
The majority next argues that defense counsel’s “line of questioning pursued during cross-examination objectively represents a responsive, tactical choice.” Op. at 1124. Courts must review the strategic choices of counsel with considerable deference because there are many ways to render effective assistance in a particular case. Not only do we defer to strategic decisions, but we begin with the presumption such a strategy existed, that individual choices were guided by an overarching, sound trial strategy:
[W]e always start the analysis that an attorney acted in an objectively reasonable manner and that an attorney’s challenged conduct might have been part of a sound trial strategy.... [Wjhere it is shown that a particular decision was, in fact, an adequately informed strategic choice, the presumption that the attorney’s decision was objectively reasonable becomes “virtually unchallengeable.”
Bullock v. Carver, 297 F.3d 1036, 1046 (10th Cir.2002) (quoting Strickland v. Washington, 466 U.S. 668, 690, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984)). In determining how much we are to defer to trial counsel’s decisions, however, a court must first determine whether those decisions were strategic.
Neither the magistrate judge nor the district judge detected an ounce of strategy in defense counsel’s “woefully inadequate” cross-examination of Dr. Hemphill, and I am similarly at a loss to see it. The majority, however, holds that “[njothing in the record would lead” to the conclusion that trial counsel’s cross-examination of Dr. Hemphill was not shaped by her presumptively sound trial strategy. Op. at 1124. But there is something in the record that would lead to this conclusion: counsel’s explanation of her own behavior. Counsel (who was engaged in her first capital murder trial) testified that Dr. Hemphill’s change in testimony “thr[ew] a grenade in the middle of our first and second stage defense.” R. Vol. Ill 615. Aghast at what she regarded as the destruction of the defense theory both as to guilt and mitigation, defense counsel testified that she did not cross-examine Dr. Hemphill as she ought to have done: “I was completely at a loss, given what he had just said out in the hall, and I did not handle it, did not touch it on cross-examination as I should have.” R. Vol. Ill 614-15. Counsel acknowledged that she should have asked for a continuance to re-interview Dr. Hemphill and secure other expert testimony as to the cause and manner of death, and to “just have a chance to step back and think and strategize about how much the case had changed right there in that few moments.” R. Vol. Ill 615. Had there been but world enough and time, counsel would have taken advantage of the break to make strategic choices; without a *1134continuance, counsel floundered through a fleeting and inscrutable cross-examination.
To say that “nothing in the record” supports the district court’s finding that defense counsel’s failure to conduct a more penetrating cross-examination was not strategic, Op. at 1124, thus requires us to ignore counsel’s testimony at the eviden-tiary hearing. The majority hints that this disregard of the evidence may be rooted in its view that the “testimony produced in evidentiary hearings ... benefits in its acuity from the passage of time.” Op. at 1123. But this is precisely the sort of evidence on which we are supposed to rely in cases of this sort. We permit criminal defendants to raise ineffective assistance of counsel claims for the first time in collateral proceedings precisely because of the utility of evidence as to counsel’s trial tactics:
A factual record must be developed in and addressed by the district court in the first instance for effective review. Even if evidence is not necessary, at the very least counsel accused of deficient performance can explain their reasoning and actions, and the district court can render its opinion on the merits of the claim.
United States v. Galloway, 56 F.3d 1239, 1240 (10th Cir.1995) (en banc) (footnote omitted). It seems hardly right, having insisted on the importance of developing the record to include counsels’ explanations for their performance, then to ignore the very explanations for which we have asked. It may be that the majority dislikes the position in which evidentiary hearings place trial counsel, forcing them to impugn their own performance if they wish to advance their former client’s position. I share that concern. But so long as Strickland and our precedent require consideration of whether a decision was strategic or not, we must use whatever evidence of strategy may be at hand.
The magistrate judge who witnessed trial counsel’s testimony at the evidentiary hearing was no doubt as aware as we can be of the incentives such hearings create, but the magistrate judge nevertheless credited trial counsel’s testimony as to her devastation. R & R 21-24. Indeed, the state objected to the magistrate judge’s report and recommendation on the ground that the magistrate had relied too heavily on trial counsel’s evidentiary hearing testimony. Order 8. The district court rejected this argument, approving the magistrate’s use of the testimony to “confirm[] that [trial counsel’s] failure to elicit critical testimony from Dr. Hemphill was clearly not a strategic decision.” Id. This was a credibility determination, which commands “even greater deference to the trial court’s findings” than do other findings of fact. Anderson v. City of Bessemer, 470 U.S. 564, 575, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985). I can agree that where counsel’s post hoc testimony contradicts the record or is otherwise implausible in light of the record, the factfinder should take it with a grain of salt. But in this case, counsel’s testimony is consistent with and corroborated by Dr. Hemphill’s testimony regarding his change of testimony and her reaction to it, and is confirmed by the utter lack of any incisive cross-examination. An appellate court’s suspicion that a witness had an incentive to shade her testimony regarding her past thoughts and emotions is insufficient to justify overruling a lower court’s credibility determination. In light of the especially deferential clearly erroneous standard of review due such determinations, I cannot accept the majority’s finding that counsel’s cross-examination performance was the product of a strategic decision.
*1135In response to this argument, the majority observes that the district court found “counsel’s performance was deficient ... on the face of the trial record,” Op. at 1124 n. 5, and thus maintains that there was no credibility determination to which we must defer. But the majority conflates the district court’s determination that defense counsel’s cross-examination of Dr. Hemp-hill was deficient (which was based on the trial record) with its determination that the cross-examination was not guided by any particular strategy, which was a credibility determination based on defense counsel’s testimony at the evidentiary hearing. See Order 8 (“[t]he inadequacy of the cross-examination is evident ... from the trial record itself. [Defense counsel’s] evidentiary hearing testimony simply confirmed that her failure to elicit critical testimony from Dr. Hemphill was clearly not a strategic decision.”). The majority disregards the second of these determinations, and holds defense counsel’s course of action to be strategic, claiming there is no record evidence to the contrary. Op. at 1124. That improperly disregards the district court’s fact-finding.
The majority suggests that counsel’s cross-examination was a reasonable tactical choice because “a longer cross-examination of Dr. Hemphill may have resulted in more resolute testimony, and a redirect examination reinforcing the state’s theory of the case.” Op. at 1125. But we know what Dr. Hemphill would have said if asked whether the medical evidence supported the left-alive theory, because he said so in his deposition. Such a question would not have elicited “more resolute” testimony reinforcing the prosecution’s position, but an affirmation that the defendant’s “left-alive” theory was consistent with the evidence. R & R 24. As Dr. Hemphill testified:
All [defense counsel] would have to do to counter any of [my testimony] would be to say, but is there any scientific evidence that indicates that indeed what Mr. Waller said happened did happen in this case? And the answer is, well, no.
Hemphill Depo. 30, R. Vol. Ill 895. But if there were any doubt about what Dr. Hemphill would say if he were seriously cross-examined, as the magistrate judge pointed out, defense counsel could have sought — but did not — a continuance to interview the witness and find out. R & R 11.
Rather than there being “[n]othing in the record” to undermine the presumption that counsel’s decisions were strategic, in actuality nothing in the record points the other way. The cross-examination was exactly what trial counsel testified it was: a debacle guided by no strategy at all. If there was no strategy, there is no reason to extend the special deference we accord strategic decisions, and no reason to convert an objectively unreasonable performance into an acceptable one.
Finally, the majority maintains that defense counsel’s performance was objectively reasonable. Indeed, the majority comments: “Although defense counsel understandably may have felt ‘devastated’ by her conversation with Dr. Hemphill, her conduct during his testimony was aggressive and successful.” Op. at 1124. The majority elaborates:
We disagree with the district court’s conclusion that counsel’s cross-examination was deficient on the face of the record. Experience teaches that rarely can performance be measured by the length of cross-examination alone. In this case, counsel engaged in a succinct line of questioning going to the heart of her case. Questions were posited regarding the size of the maggots on Mrs. Launhardt’s body to support the anticipated testimony of Dr. Greenburg. *1136Moreover, defense counsel established through the testimony of Dr. Hemphill that the victim could have been tied or restrained in a sitting or standing position. While it is true that defense counsel did not explicitly ask whether the evidence was consistent with the victim being left alive, we cannot say that this single omission, if it was that, under all the circumstances present renders counsel’s performance constitutionally deficient.
Op. at 1124-25 (internal footnotes omitted). With these few rather cursory sentences, the majority overturns the thorough and careful opinions to the contrary by the magistrate and district court judges below. I cannot agree with the majority’s assessment.
It is true that defense counsel did something on cross-examination, succeeding in eliciting support for the defense estimate of the time of death, based on the entomological evidence. But doing something is not always enough: “The duty of counsel is not merely to do ‘something’ rather than nothing on behalf of one’s client, but to act as the client’s effective advocate during each critical stage of the defense.” Fisher v. Gibson, 282 F.3d 1283, 1307 (10th Cir.2002). When a matter is central to the theories of both the prosecution and defense, counsel’s obligation to meet the prosecution’s evidence is heightened. See Rompilla v. Beard, — U.S. -, -, -, 125 S.Ct. 2456, 2462, 2467, 162 L.Ed.2d 360 (2005). With a few questions, defense counsel could have ensured that the defense theory was left alive at the end of Dr. Hemphill’s testimony. Failing to ask those few questions was not a “single omission,” Op. at 1125, but a glaring failure to counter evidence “at the very heart of the prosecution’s case.” Id. at 2470 (O’Connor J., concurring) (emphasis omitted).
It is necessary to compare counsel’s performance not to an ideal, Perry Mason-style defense, and not to a standard of “doing nothing,” but to what a reasonably competent counsel could accomplish under the circumstances of the case. His deposition for the evidentiary hearing reveals that, if asked, Dr. Hemphill would have testified that the medical evidence was consistent with the defense theory:
Q: ... So is it fair to say, Doctor, that Mrs. Launhardt could have passed out from medications and fallen into this strap and that could be the mechanism that caused this?
A: With emphasis on the word “could have,” yes.
Q: All right. Is it fair to say that Miss Launhardt could have experienced an arrhythmia from relatively mild constriction of the neck and that’s what could have caused or contributed to this death?
A: Could have, yes.
Q: All right. Is it fair to say that Miss Launhardt could have passed out from an arrhythmia or even died from it directly and that’s what could have caused this death?
A: Yes.
Q: Is it fair to say that Miss Launhardt could have been left alive and tied up in that closet for a period of time?
A: Yes.
Q: And is it fair to say that without the co-defendant’s statement there was no evidence she was deliberately suspended by anyone? I think that’s what you’ve told us already, Doctor.
A: Well, I want to be sure that I’m— since this is such a sticky part with everybody I want to be sure I get it right this time. That without the co-defendant’s statement—
*1137Q: There was no evidence she was deliberately hanged by anyone.
A: Okay. You’re only referring to the suspension. Not anything else in this scenario. No evidence that someone else intentionally suspended her.
Q: That’s what I’m asking you.
A: As far as I know there is no other evidence that directly indicates that.
R. Vol. Ill 937-38. The contrast between what could have been accomplished on cross-examination and what was accomplished is stark.3
Although defense counsel’s attempt to set the stage for the entomological expert was commendable, the task at hand was to counter Dr. Hemphill’s testimony that “someone killed this person by partially suspending her by the neck while she was found and face down in the position which she was found.” Defense counsel did nothing of the sort, despite Dr. Hemphill’s willingness to acknowledge that the scientific evidence was entirely consistent with the defendant’s theory of the case. Questions about maggots must have seemed to the jury a digression. Counsel’s fourth question — whether the victim could have been tied or restrained in the closet in a sitting or standing position — could have served as the opening salvo in an effective line of questioning, but without more, it must have seemed to the jury more a confirmation of guilt than a step in a line of defense.
II. The Theory of the Concurrence
Putting forward an argument not made in the briefs or clearly argued below, the concurrence suggests a possible strategy behind the cross-examination. According to the concurrence, defense counsel “had little room to maneuver,” given the danger that any question about the bases of Dr. Hemphill’s opinion or the recent change in it would have led to the admission of Mr. Waller’s otherwise inadmissible hearsay statements, which would have been highly damaging to the defense. It was better, the concurrence seems to suggest, to lay low. Con. 1. This assessment of defense counsel’s position, however, seriously overestimates the danger of opening the door to Mr. Waller’s statement, and glosses over substantial barriers to its admission.
The concurrence takes an unduly narrow view of the options available to trial counsel, presupposing that any questions about the medical support for the left-alive theory would have prompted Dr. Hemphill to relate Mr. Waller’s statement. According to the concurrence, “[e]ven one question regarding the consistency of the physical evidence with the defense theory could open the door.” Con. 2. The concurrence assumes that trial counsel could not avoid asking Dr. Hemphill why he changed his opinion before trial. It is far from self-evident, however, that merely establishing that the medical evidence was consistent with the left-alive theory — without mentioning Dr. Hemphill’s about-face — would have opened the door to Mr. Waller’s statement. Similarly, the concurrence supposes that if defense counsel had called a defense expert on the cause of death, Dr. Hemphill could have introduced Mr. Waller’s statement on rebuttal to explain his disagreement. Id. But it is not at all clear *1138why calling another expert to establish the left-alive theory would have opened the door to Mr. Waller’s statement on rebuttal. Rebuttal evidence is evidence which attempts to “disprove or contradict” the evidence to which it is contrasted. Black’s Law Dictionary 579 (7th ed.1999). “[WJhether or not rebuttal evidence is admissible depends on ‘whether the initial proof might affect the case and whether the rebuttal evidence fairly meets the initial proof.’ ” Tanberg v. Sholtis, 401 F.3d 1151, 1166 (10th Cir.2005) (quoting Christopher B. Mueller & Laird C. Kirkpatrick, 1 Federal Evidence § 12 n. 2 (2d ed.2004)); see also United States v. Stitt, 250 F.3d 878, 897 (4th Cir.2001) (“[WJhen otherwise inadmissible, rebuttal evidence must be reasonably tailored to the evidence it seeks to refute.... [T]here must be a nexus between the purported rebuttal evidence and the evidence that the purported rebuttal evidence seeks to rebut.”). Expert testimony exploring the medical evidence is not refuted by a layman’s purported eyewitness account.
Even if Dr. Hemphill attempted to introduce Mr. Waller’s statement, it seems quite unlikely that Oklahoma law would have permitted it to come in.4 First, there is serious doubt whether Mr. Waller’s statement was reliable. According to the concurrence, Oklahoma has “adopted a relatively relaxed view” of permitting experts to testify about the bases of their opinions, even when they have relied on “unsubstantiated hearsay and improper character evidence.” Con. 2 (quoting Lewis v. Oklahoma, 970 P.2d 1158, 1166 (Okla.Crim.App.1998)). But Lewis does not support the inference that Oklahoma courts will freely admit otherwise inadmissible evidence as long as it is filtered through an expert. The defendant in Lewis objected to the admission of “unsubstantiated hearsay and improper character evidence” by an expert witness as part of the basis of his opinion. The Oklahoma Court of Criminal Appeals agreed that some of the evidence should not have been admitted because it might not have come from “a reliable source.” Lewis, 970 P.2d at 1166-67. Because the expert had properly relied on “substantial admissible evidence,” the court found the error in admitting the unreliable evidence to be harmless on a highly deferential plain error standard of review. Id. at 1167. The Lewis court made clear that the opportunity to admit evidence underlying an expert’s opinion is not “a license to parade a mass of inadmissible evidence before the jury.” Id. (quoting Sellers v. State, 809 P.2d 676, 685 (Okla.Crim.App.1991)). However relaxed Oklahoma’s stan*1139dard for admitting hearsay and character evidence as the basis of an expert’s opinion, Lewis does not stand for the proposition that an expert may shepherd patently unreliable testimony into the record.
Second, if experts rely on otherwise inadmissible evidence, it must be “of a type reasonably relied upon by experts in the particular field in forming opinions or inferences upon the subject.” 12 Okla. Stat. Ann. § 2703. It is unlikely that Dr. Hemphill’s reliance on Mr. Waller’s retracted statement meets that standard.5 I find it very difficult to believe that medical experts asked to testify about the physical evidence routinely discount scientifically plausible theories on the basis of inadmissible statements by participants in the crime. Furthermore, there was a mechanism available to probe the appropriateness of Dr. Hemphill’s reliance on Mr. Waller’s statement: Lewis holds that, “upon request of either party, [the trial court must] hold an in camera hearing to determine whether an expert’s reliance on particular information is reasonable.” Lewis, 970 P.2d at 1167. Trial counsel thus had an opportunity to determine — outside the presence of the jury — not only whether Mr. Waller’s statement could be admitted, but whether Dr. Hemphill’s reliance on it was reasonable. Not to take advantage of this mechanism falls short of effective assistance.
Finally, the fact that Dr. Hemphill relied on the Waller statement does not obviate the basic prerequisites to admissibility. Even if evidence is sufficiently reliable and of an appropriate type to justify an expert’s reliance, it may be inadmissible if it is unduly prejudicial. The Lewis court admitted the evidence on which the expert properly relied only after determining that “it was probative and its probative value was not outweighed by the danger of unfair prejudice.” Id. Defense counsel would have had a strong argument that the danger of prejudice substantially outweighed whatever probative value there was in Mr. Waller’s statement.
Thus, I cannot agree with the concurrence that a carefully crafted cross-examination of Dr. Hemphill would inevitably, or even probably, have opened the door to introduction of Mr. Waller’s damaging statement. That defense counsel feared Dr. Hemphill would introduce the Waller statement does not make it so. See Con. 2. Counsel’s unwarranted fear only confirms the real reason for her failure to present the defense theory: she panicked. To abandon one’s position for fear of a chimera hardly meets the standard of effective representation. The game was far from up, but trial counsel abandoned the crucial element of the defense. The concurring opinion, like the majority, supplies no adequate reason to dispute the district court’s conclusion that this cross-examination was “woefully inadequate.”
III. Prejudice
Deficient performance is only one aspect of ineffective assistance; in order to violate the Sixth Amendment right to counsel, the weaknesses in counsel’s performance must also prejudice the defendant. In order to demonstrate prejudice a defendant must show “a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different. A reasonable probability is a probability sufficient to undermine confidence in the outcome.” Strickland, 466 U.S. at 694, 104 S.Ct. 2052. The likelihood that prejudice resulted from counsel’s defi*1140cient performance alters with the strength of the government’s case: “a verdict ... only weakly supported by the record is more likely to have been affected by errors than one with overwhelming record support.” Id. at 696, 104 S.Ct. 2052.
The jury’s presumed finding that Petitioner did not leave his victim alive, but actively caused her death, was only weakly supported by the evidence. The government relied primarily, if not exclusively, on Dr. Hemphill’s testimony to establish that Petitioner actively killed the victim. Defense counsel’s failure to subject that evidence to meaningful adversarial testing allowed the government’s equivocal evidence to be presented to the jury as unequivocal. Moreover, the facts of the case suggested a plausible alternative: that the defendants left their victim alive and upright in the closet, and that she died as the result of slipping or fainting up to 48 hours later. It would surely have been helpful to the defense to establish that the medical evidence regarding the position of the body was consistent with both scenarios.
The fact that at the time of the trial the jury could have accepted the left-alive theory and still have convicted on felony murder makes it even more likely that the jury might have acquitted on the malice aforethought murder charge had defense counsel properly tested the government’s evidence supporting this charge.6 Defendant has therefore satisfied both prongs of the Strickland test.
IV. Conclusion
A few weeks ago, in Rompilla v. Beard, — U.S. -, 125 S.Ct. 2456, 162 L.Ed.2d 360 (2005), the Supreme Court reversed the lower court’s denial of habeas relief on account of ineffective assistance of counsel in a case in every way weaker than this one. In Rompilla, the state court reviewed the conviction on the merits, and the decision was therefore entitled to deference under the Antiterrorism and Effective Death Penalty Act, 28 U.S.C. § 2254(d). See id. at 2462 (noting that the Court could grant habeas relief only if the state court’s decision was “[not only] incorrect or erroneous [but] objectively unreasonable”) (quoting Wiggins v. Smith, 539 U.S. 510, 520-21, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003)). In this case, the state court improperly found that Mr. Richie’s ineffective assistance claim was procedurally barred, and as the majority holds, our review is therefore de novo. Op. at 1119 n. 1. Moreover, in Rompilla, defense counsel supplied facially reasonable justifications for their failure to undertake a more thorough search for mitigating evidence. See id. at 2462 (“the merits of a number of counsel’s choices in this case are subject to fair debate”). In this case, defense counsel acknowledged that her inadequate cross-examination was a product of her surprise and “devastation” at learning of Dr. Hemphill’s intended change of testimony. Far from being “subject to fair debate,” the district court *1141found defense counsel’s performance in this case “woefully inadequate.” Finally, the prejudice in Rompilla — the possibility that the jury might have been moved by evidence of the defendant’s egregious upbringing to mitigate its appraisal of his culpability — was far more speculative than the prejudice from the inadequate cross-examination in this case, which concealed from the jury the fact that the prosecution’s expert witness believed the medical evidence equally supported the defendant’s theory that the victim was left alive. If the jury had known that the prosecution’s medical evidence was equivocal, that knowledge, coupled with the entomological evidence suggesting that death occurred some 48 hours after Mr. Richie and Mr. Waller left Mrs. Launhardt in the closet, might well have resulted in acquittal of malice aforethought murder.
In short, this is the unusual case where the performance of defense counsel was not only objectively deficient, but likely contributed to a wrongful conviction for a capital crime. I respectfully dissent.
. Although, as the majority notes, Dr. Hemp-hill's recollection does not completely jibe with defense counsel’s, Dr. Hemphill confirmed in his deposition that he had informed defense counsel that the "left-alive” theory of the defense was "at least as reasonable as any other explanation” and that he “couldn’t really think of anything that was any more reasonable than that.” Hemphill Depo. 23, R. Vol. Ill at 888.
. The majority challenges that conclusion, an issue I will address below at 8-9.
. Defense counsel exacerbated the poor cross-examination by summarizing it inaccurately in closing arguments. Counsel argued to the jury that Dr. Hemphill testified that the victim could have been "left there restrained, but alive ... there's nothing [in the medical evidence] inconsistent with that.” R. Vol. VIII 2229. But this is precisely what Dr. Hemp-hill did not testify to on cross-examination, because he was never asked to testify on this point. An attentive juror would be likely to discount an argument that inaccurately summarized the testimony on which it was based.
. The concurring opinion states that defense counsel "vigorously and repeatedly object[ed] that such use of hearsay was improper. The trial judge, however, overruled the objections.” Con. 1. I do not read the transcript that way. When the issue first arose, defense counsel objected that Dr. Hemphill should not be permitted to testify which mode of death was "most likely,” because that judgment would be based on Mr. Waller’s hearsay statement rather than the medical evidence. R. Vol. VII 2079-80. That objection was sustained, but the court allowed Dr. Hemphill to testify as to "how this [the death] could have occurred.” Id. at 2080, 2082. During this exchange, the judge stated: "I didn’t say [Waller’s statement is] hearsay. I didn’t say that's hearsay or not.” Id. at 2082. Defense counsel then made a series of objections to statements and questions regarding how the death could have occurred, all of which were overruled. Id. at 2083-84. When Dr. Hemp-hill started to allude to information he had access to from "someone who purported to have been there” — a reference to Mr. Waller — defense counsel again objected, and the objection was sustained, and the jury was admonished to disregard the answer in which Dr. Hemphill alluded to Mr. Waller's statement. Id. at 2085. The trial court thus found at least some uses of the Waller statement improper, though whether the basis for the rulings was hearsay or something else is not clear.
. Since the State did not advance on appeal the argument made in the concurrence, we have not had the benefit of the parties' briefing on whether Dr. Hemphill’s reliance on Mr. Waller’s statement was proper under the Oklahoma law of evidence.
. The government argues that the jury could have found Petitioner guilty of malice aforethought murder even if it had accepted Defendant’s left-alive theory. This argument is unpersuasive. The government's assertion that one can be guilty of malice aforethought murder when leaving a victim alive but in unpropitious circumstances is supported by a lone, unpublished case from the Tennessee Court of Criminal Appeals, a fact which does not inspire confidence that this is in fact the state of the law in Oklahoma. Moreover, the mere possibility that a jury, although not urged to do so by the prosecutor, might nevertheless have convicted on malice aforethought murder while believing the victim was left alive, does not mean that Defendant was not prejudiced. A defendant need only show a "reasonable probability that ... the result of the proceeding would have been different” to demonstrate prejudice. Strickland, 466 U.S. at 694, 104 S.Ct. 2052.