specially concurring:
This case comes before this en banc court as an appeal from the district court’s decision to deny petitioner a writ of habeas corpus. A divided panel of this court voted to vacate in part the district court’s decision and to remand the case. I agree with the panel that this case must be remanded, but do so for different reasons.
Petitioner’s main contention in support of his application for a writ, and the only contention with which this en banc court is concerned, is that he was denied his federal constitutional right, under the sixth, and fourteenth, amendments, to the effective assistance of counsel. In support thereof, petitioner alleges that his counsel incompetently failed to produce certain mitigating evidence at his state capital-sentencing trial. Petitioner also alleges, as he must to support his claim, that he was prejudiced by his counsel’s incompetent omission.
To prevail on his claim, petitioner must sustain both aspects of it: incompetence and prejudice arising therefrom. It is only the latter aspect of this claim with which this opinion is concerned. Because I believe petitioner cannot prevail on the prejudice aspect of his claim, I do not reach the issue whether counsel was incompetent.
My discussion proceeds as follows. First, I discuss the facts and the procedural history of this case. Second, I announce the proper standard for determining prejudice arising from counsel’s allegedly incompetent failure to produce mitigating evidence at a state capital-sentencing trial. Third, I discuss the way in which the standard I propose must be applied. And fourth, I *1265describe the errors the district court committed in this case.
Before beginning my discussion, I note the central themes running throughout this opinion. The first theme is the most basic and all the others follow from it — that prejudice must be determined as a matter of state law.1 Although the ultimate question of ineffectiveness must be decided based on federal constitutional standards, the threshold, and in this case dispositive, question of prejudice is a state law question. The second theme follows directly from the first — federal courts should not interfere in this state law area unless it is absolutely necessary to resolve the claim of ineffectiveness. For this reason, I propose a test for prejudice that minimizes federal court intrusion on state law. The third, and final, theme is that federal courts need not even engage in sensitive determinations of prejudice if state collateral attack courts clearly articulate the state law, in this case state sentencing policy, and if federal courts are aware of their duty to dismiss unexhausted habeas claims. I now begin my discussion.
I.
On September 20,1976, the petitioner and an accomplice robbed and murdered Daniel Pridgen, a minister; petitioner stated that they killed Pridgen because he believed that a minister who engaged in homosexual activities, as he alleges Pridgen did, is a “hypocrite.” On September 23, petitioner broke into the house of Katrina Birk with an intent to rob. Finding Ms. Birk and her three elderly sisters-in-law, and experiencing difficulty during the course of the robbery, petitioner shot and stabbed each victim, killing Ms. Birk and gravely injuring the others. On September 26, petitioner and two accomplices kidnapped Frank Meli. Petitioner killed Meli on September 29 when a ransom demand failed.2
A. The State Proceedings.
On October 1, 1976, petitioner surrendered to police after the apprehension of his accomplices in the Meli case. He confessed and was indicted in the Dade County Circuit Court for that murder and related, lesser offenses on October 7. William Tun-key, the counsel whose effectiveness at the sentencing phase of petitioner’s trial is in question here, was appointed at that time.3
Acting against the advice of counsel, petitioner confessed to the other crimes described above on November 5, 1976, and was indicted for them on November 17. Again acting against the advice of counsel, petitioner pled guilty to all charges in all three cases on December 1, 1976. The circuit judge conducted a thorough and extensive colloquy to ensure the voluntariness of the guilty plea and then accepted it. Petitioner then waived his right to a jury at the sentencing hearing.
At the sentencing hearing, the state opened by detailing the circumstances of the three murders. Tunkey waived an opening statement, relying on a sentencing memorandum filed with the court. The state then called nine witnesses who testified about the aggravated nature of the offenses. It also introduced fifteen exhibits that portrayed the aggravated nature of the crimes.
Tunkey did not introduce any new evidence in mitigation; instead, he adopted petitioner’s prior statement at the guilty plea colloquy, thus shielding his client from cross-examination. This previous statement dealt primarily with matters germane to the guilty plea; it also contained some very limited information petitioner volunteered about his dire economic situation, his remorse, and his emotional state. The cir*1266cuit judge foreclosed the petitioner’s attempt to make a more detailed explanation of his actions at the plea hearing, and stated that the court would consider such information at sentencing. Tunkey, however, declined this explicit invitation to introduce evidence in mitigation at the sentencing hearing.
The state then made a closing argument. Tunkey followed, briefly emphasizing the defendant’s honesty in admitting his guilt, stressing the existence of the possibility of life imprisonment without option for parole, and asking the court for mercy. The state then made a brief rebuttal argument.
The court imposed three sentences of death for the three murders.4 In imposing these sentences, the court followed the procedures the Florida death penalty statute, Fla.Stat. § 921.141, mandated. First, it arrived at petitioner’s sentencing profile. It did this by finding and then weighing statutory aggravating circumstances, id. § 921.-141(5), and statutory, id. § 921.141(6), and nonstatutory mitigating circumstances. It then combined these circumstances to arrive at a comprehensive profile of the defendant and his crime.5 Second, the court discerned the state’s sentencing policy applicable to petitioner’s case. This policy was expressed in sentencing decisions in cases presenting profiles similar to petitioner’s. Third, the court applied such policy to petitioner’s profile and determined that death was the appropriate sentence in all three cases. The Florida Supreme Court affirmed all three death sentences on statutorily-mandated direct appeal. Washington v. State, 362 So.2d 658 (Fla.1978), cert. denied, 441 U.S. 937, 99 S.Ct. 2063, 60 L.Ed.2d 666 (1979).
On March 19, 1981, petitioner, with different counsel, moved for post-conviction relief in the Dade County Circuit Court pursuant to Fla.R.Crim.P. 3.850.6 Since the sentencing judge, Richard Fuller, had retired, Circuit Judge Mario Goderich heard the motion. The core of this collateral attack on the death sentences was that Tun-key’s failure to investigate and adduce mitigating evidence at the sentencing hearing denied petitioner his constitutional right to the effective assistance of counsel.
Petitioner’s motion incorporated all the mitigating evidence that he alleged was incompetently omitted at his sentencing hearing. Although he did not attempt to show that sentencing profiles similar to the one he presented in his motion had received life imprisonment in the past,7 he did offer psychiatric evidence of his broken and violent home, one marked by extensive child abuse and incest; his panic, frustration, and depression at his economic circumstances; and his remorse for his crimes. He also offered affidavits from family, friends, former employers, and teachers. These affidavits portrayed a young man under intense emotional pressure because of his inability to provide for himself, his wife, and his infant. They described petitioner as responsible and nonviolent, active in his church, and devoted to his family. They also emphasized that there was an inexplicable difference between the person the affiants knew and the one who committed these crimes. The affiants stated that they would have testified at the sentencing hear*1267ing but were never contacted for that purpose.8
After reviewing the record and hearing arguments of counsel, but without holding an evidentiary hearing, the court denied all relief on March 27, 1981.9 In so doing, the court assumed arguendo that the allegations of petitioner’s motion and the affidavits he presented in mitigation were true but held, nonetheless, that he failed to establish a prima facie showing of prejudice arising from ineffective assistance of counsel, which showing is a necessary component of a claim of ineffectiveness.10 In effect, the court dismissed the motion for failure to state a claim on which relief could be granted. The court stated that
as a matter of law, the record affirmatively demonstrates beyond any doubt that even if Mr. Tunkey had [presented ' the new mitigating evidence] at the time of the sentencing, there is not even the remotest chance that the outcome would have been any different. The plain fact is that the aggravating circumstances proved in this case were completely overwhelming, and that even to this date the Defendant cannot show that any statutory mitigating circumstances existed. The non-statutory mitigating circumstances which he claims his attorney failed to investigate and present at the time of sentencing would as a matter of law, be insufficient to outweigh the multiple aggravating circumstances present in this case.
Order Denying Post-Conviction Relief Filed Pursuant to Fla.R.Crim.P. 3.850 at 12 (emphasis in original).
On April 6, 1981, the Florida Supreme Court affirmed this denial of relief, concluding that “[appellant’s] claims are shown conclusively to be without merit so as to obviate the need for an evidentiary hear*1268ing.... [W]e can find no prejudice caused to appellant, even if we assume that every allegation he has made in his petition is true.” Washington v. State, 397 So.2d 285, 286 (Fla.1981).
B. The Federal Habeas Corpus Proceedings.
The same day the Florida Supreme Court handed down its decision, two days prior to his scheduled execution, petitioner applied to the federal district court for a writ of habeas corpus. He presented the same ineffective assistance of counsel claim he had presented in state court. On April 7, the district court conferred with counsel to determine whether an evidentiary hearing would be necessary and thus a temporary stay of execution. The court inquired whether the Dade County Circuit Court had held an evidentiary hearing on petitioner’s claim, and, if so, which issues had been resolved.11 The state replied that the state court had denied petitioner’s claims on the pleadings, without an evidentiary hearing, concluding as a matter of law that petitioner had failed to allege a sixth, and fourteenth, amendment violation. Petitioner’s counsel insisted that the court had to hold an evidentiary hearing to dispose of each element of petitioner’s claim. The state noted that petitioner’s habeas petition, including his allegation of prejudice, merely replicated the Fla.R.Crim.P. 3.850 motion, which the state court had denied on the pleadings. It urged the court therefore to dismiss the petition without a further hearing, for failure to state a claim on which relief could be granted. The district court concluded this conference with counsel without ruling on the sufficiency of the petition. Shortly thereafter, the court notified counsel that it would convene an evi-dentiary hearing on April 10.
When the hearing began on April 10, the court, announced that it was going to expedite the proceedings. As a starting point,12 the court stated that it would consider the mitigating evidence that petitioner contended his trial counsel should have produced at the sentencing hearing. The court stated that it would receive that evidence in the form of the affidavits and psychiatric reports attached to the habeas petition. Record, vol. 2, at 4-5. It is worth emphasizing that these affidavits and psychiatric reports were identical to those the state collateral attack court had considered and rejected. The court would not, however, permit the affiants to testify. Petitioner’s counsel objected to this procedure and argued that the affidavits were merely illustrative and did not contain everything the affiants and others would say in mitigation of the death penalty if permitted to testify in open court. The court overruled this objection, and instructed counsel to proceed with his case.
Counsel then called petitioner’s trial attorney, William Tunkey, to the stand. Tun-key testified that the court had appointed him in October 1976 to represent petitioner against charges that petitioner had kidnapped and murdered Frank Meli. He then testified that after seeing his client’s confessions he “had a hopeless feeling,” id. at 22, and that “the investigation ..., the work that was done to locate prospective witnesses to testify on his behalf [was] minimal and that is using hindsight.” Id. at 25-26. Tunkey stated that his knowledge of the sentencing judge, Judge Richard Fuller, dictated his trial strategy for the sentencing hearing. Tunkey believed that Judge Fuller respected a defendant who candidly admitted his guilt. Therefore, Tunkey presented his client’s case in a manner that he thought would convince the *1269judge that the defendant had pled guilty to all charges with candor and sincerity.13
Petitioner rested his case after Tunkey’s testimony, and the state moved for an involuntary dismissal, under Fed.R.Civ.P. 41(b). The court announced that it would defer ruling on the motion until the close of all the evidence and directed the state to proceed with its case.14 The state’s case consisted primarily of the testimony of Judge Fuller. The state presented Judge Fuller as “an expert.witness with regard to his experience on the Bench.” Record, vol. 2, at 80. Over petitioner’s objection on relevancy and other grounds, the state elicited Judge Fuller’s opinion that petitioner’s new mitigating evidence would have made no difference in the sentence imposed at trial. Judge Fuller stated that the murders petitioner committed were so aggravated that even if petitioner had produced the witnesses identified in his habeas petition at the sentencing hearing and they had testified as petitioner represented they would have, it “would not have changed my opinion then nor would it have changed my sentencing were I to give it today.” Id. at 96.
The district court denied the writ on the ground that petitioner had failed to prove prejudice arising from Tunkey’s failure to produce mitigating evidence at the state sentencing hearing. In its opinion, the court indicated that it did not consider Judge Fuller’s testimony as “determinative on the issue of prejudice” because his testimony had “the potential weakness of hindsight analysis”; nevertheless, the court reached the identical result that Judge Fuller said he would have reached. It concluded that “there does not appear to be a likelihood, or even a significant possibility that the balance of aggravating against mitigating circumstances under the Florida death penalty statute would have been altered in petitioner’s favor.” Record, vol. 1, at 63.
Having determined that petitioner’s claim failed for lack of prejudice, the court observed that it was not required to decide whether Tunkey’s performance during the sentencing phase of the criminal prosecution passed constitutional muster. Nonetheless, the court found that “Mr. Tunkey should have made an independent investigation of factors relevant to mitigation, and that such investigation would have produced generally favorable information from family, friends, former employers, and medical experts.” Id.15
An appeal was taken, and a divided panel of this court vacated in part the district court’s decision and remanded the ease. Washington v. Strickland, 673 F.2d 879 (5th Cir.1982). The majority held that the district court had applied an erroneous stan*1270dard to determine prejudice; the dissent argued that no prejudice could be shown under any test, and that counsel had made competent, strategic choices. The ease is now before us en banc.
II.
I now consider the degree of prejudice a petitioner must prove to obtain federal ha-beas corpus relief on a claim that his counsel incompetently failed to produce mitigating evidence at his state capital-sentencing trial. My discussion proceeds as follows. First, I note that there is no generally accepted test in this area. Therefore, I treat this ease as one of first impression. Second, I reject the test on which the district court relied, the outcome-determinative test, for two reasons: it results in the most onerous federal intrusion on state sentencing policy, and it is incongruous with the goal underlying the right to counsel — to ensure fairness in the criminal process. Third, I reject the test the panel proposed, whether the mitigating evidence counsel failed to produce would have “altered [petitioner’s trial] in a way helpful to [petitioner],” 673 F.2d at 902, because it is too vague. Finally, I propose a test that ensures fairness in the criminal process, and is neither too intrusive on state sentencing policy nor too vague: whether the mitigating evidence counsel failed to produce would have substantially or materially affected the decision-making process of a rational sentencer.
Initially, I note that “[t]he law of our circuit is as yet unclear as to the precise degree of prejudice that a defendant must demonstrate before he is entitled to habeas corpus relief on grounds that' he received ineffective assistance of counsel .... ” Washington v. Watkins, 655 F.2d 1346, 1362 n. 32 (5th Cir.1981), cert. denied, - U.S. -, 102 S.Ct. 2021, 72 L.Ed.2d 474 (1982) (emphasis in original).16 It is clear, however, that “some degree of prejudice must be shown.” Id. at 1362 (emphasis in original).17 I now proceed to determine what that degree of prejudice should be in the case before us.
In formulating the proper standard, it is helpful to recognize the deficiencies of the two tests the federal courts have used in this case. The first is the outcome-determinative test, which the district court used. The Court of Appeals for the District of Columbia Circuit first formulated the outcome-determinative test in United States v. Decoster, 624 F.2d 196 (D.C.Cir.1979): “[T]he accused must bear the initial burden of demonstrating a likelihood that counsel’s inadequacy affected the outcome of the tri*1271a 1. ” Id. at 208 (en bane; plurality opinion; emphasis added).18
A majority of the panel in this case rejected the outcome-determinative test because it
would require that the court hearing the ineffective assistance claim put itself in the place of the trial-court factfinder in an attempt to predict with some considerable degree of accuracy what that fact-finder would have done had it been presented with different evidence. We think that a framework for analysis which would inevitably require us, in determining whether the petitioner has made out a prima facie case for habeas relief, to engage in such highly speculative re-creations and revisions of trial court proceedings is to be avoided rather than embraced.
Washington v. Strickland, 673 F.2d at 901.19
An even more important reason for rejecting the outcome-determinative test is its invidious effect on the state sentencing process. As I discuss supra Part I, in Florida the decision whether to impose the death penalty involves a three-step process. First, the sentencer finds aggravating and mitigating circumstances, and weighs them to arrive at a sentencing profile.20 Second, the sentencer examines cases presenting profiles similar to petitioner’s to find the relevant state sentencing policy. Third, the sentencer applies the state policy to petitioner’s sentencing profile, and decides whether the death penalty should be imposed. This process results in a normative determination whether the circumstances of the crime and the characteristics of the defendant warrant a sentence of life imprisonment or of death. This norm, which embraces the given sentencing profile, then becomes a part of the state sentencing policy to which other sentencers must look in the future.
As applied in this context, the outcome-determinative test requires the federal ha-beas judge to determine whether the omission of certain evidence in petitioner’s state sentencing trial substantially affected the outcome of that trial. This determination requires the habeas judge to engage in the three-step process described above, and thus to forecast the appropriate sentence in light of the new evidence. The federal habeas judge must, therefore, promulgate state sentencing policy. Although the state would always be free to reject the federal judge’s determination of such policy, the *1272federal judge still would have entered a provisional sentencing norm. The existence of such provisional sentencing norms could lead to some truly anomalous results, as the following two examples show.
First, posit that a federal habeas judge denies the writ of a petitioner sentenced to death because he concludes that the new evidence would not have affected the outcome of petitioner’s sentencing trial. In so doing, the court must have found that the applicable state sentencing policy, which takes into account the new evidence, requires that petitioner receive the death sentence. Now, posit a second case, this one in state court, which presents a sentencing profile identical to the one the petitioner presented in federal court. Clearly, the state is free to reject the federal judge’s interpretation of state sentencing policy and sentence the defendant to life imprisonment. If, however, the state court imposes such a sentence, the unbelievable result is that a federal court has upheld a death sentence on the basis of a provisional norm which the state later rejects as an erroneous forecast of state sentencing policy. This court must not adopt a test that might produce such a result.
A second example further shows the anomalous results of using an outcome-determinative test. Posit a federal habeas judge who grants the writ of a petitioner sentenced to death because the new evidence requires the application of state sentencing policy that calls for a sentence of life imprisonment. Now, posit that a resen-tencing trial occurs in state court. There, the state court considers the identical sentencing profile that the habeas judge considered. As in the first example, the state sentencer is free to reject the federal court’s promulgation of a temporary state sentencing norm and to sentence petitioner to death. If the state sentencer does impose the death penalty, the federal court’s provisional decision may create needless confusion, and a perception of injustice arising from the conflicting decisions of two courts.
Because the outcome-determinative test could produce these inconsistent results, it must be rejected. This court cannot countenance any test for prejudice that requires the federal courts to intrude so deeply into the promulgation of state sentencing norms, an area in which the state courts have the primary authority.21 The result of allowing the federal judiciary to intrude on such an area is that the federal courts must make provisional forecasts of state sentencing norms. As the above examples illustrate, such provisional decision making must be avoided.22
*1273The outcome-determinative test must also be rejected because it is incongruous with the rationale underlying the right to counsel. The Supreme Court recently stated this rationale in United States v. Morrison, 449 U.S. 361, 101 S.Ct. 665, 66 L.Ed.2d 564 (1981): “The Sixth Amendment provides that an accused shall enjoy the right ‘to have the assistance of counsel for his defense.’ This right, fundamental to our system of justice is meant to assure fairness in the adversary criminal process.” Id. at 363, 101 S.Ct. at 667. The Supreme Court has thus intimated that the test for determining prejudice arising from the ineffective assistance of counsel should focus on whether the fairness of the “adversary criminal process” was affected. Thus, we should not adopt a test that measures prejudice by looking to the outcome of the criminal proceeding, as opposed to the fairness of the proceeding itself, because such a test focuses on the wrong issue and therefore is not suited to achieving the underlying goal of the right to counsel. For the reasons described above, we must reject the outcome-determinative test as the standard for determining prejudice in federal habeas proceedings.
In rejecting the outcome-determinative test, the panel suggested an alternative: “this prejudice requirement is satisfied by demonstrating that but for his counsel’s ineffectiveness his trial, but not necessarily its outcome, would have been altered in a way helpful to him.... [T]he change [in the trial] must be something more than insubstantial or de minimus. See Washington v. Watkins, 655 F.2d at 1362, n. 32 .... ” 673 F.2d at 902. This standard is, however, also problematic. It is vague in the extreme and, therefore, of little utility. It simply provides no standard at all to guide courts in assessing prejudice. Any additional evidence would alter a trial somewhat; what constitutes more than insubstantial or de minimus alteration is obscure. Therefore, this test too must be rejected.
Having described the shortcomings of the tests proposed thus far, it is clear that the proper standard for determining prejudice must satisfy the following concerns: it must provide the courts with some definitive guidance to assess prejudice; it must be responsive to the goal of ensuring “fairness in the adversary criminal process”; and it must not require the federal courts to engage in provisional decision making. I believe the following test satisfies these concerns: to show prejudice in a case like the one before us petitioner must prove that the decision-making process of a rational sentencer would have been substantially or materially altered had counsel properly produced the omitted mitigating evidence.23 If the district court determines *1274that the new evidence would not have made a substantial difference on the decision-making process of the state sentencer, the prejudice test is not satisfied.
This test provides the federal courts with definitive guidance because it requires them not merely to determine whether the trial process was “affected,” but whether it was “substantially affected.” Although such terms as “substantial” and “material” still do not attain the level of guidance desired, they are terms which courts apply daily in testing abstract concepts. Moreover, these terms become more meaningful when applied in light of the underlying goal of ensuring fairness in the criminal process. As such, this test provides the federal courts with as much guidance in assessing prejudice as is possible in this' area. This test also properly focuses on the trial process, rather than on its outcome. Therefore, it is responsive to the goal of ensuring a fair criminal process, the underlying purpose of the right to counsel. Finally, because this test does not require the court to concern itself with the outcome of a hypothetical case, the court need not engage in provisional decision making. In sum, the test remedies the shortcomings of the tests previously discussed.
III.
Having articulated the proper standard for determining prejudice, I now explain how a federal habeas court must apply it. The overriding message of this explanation is that because prejudice must be determined as a matter of state law, the court must first identify the relevant state sentencing profiles and norms applicable to petitioner’s case. Furthermore, once it is understood that a court can determine prejudice only by discerning state sentencing policy, it becomes clear that if state courts clearly articulate such policy and if federal courts are sensitive to dismissing unex-hausted petitions, no room is left for federal courts to interfere with such policy. In fact, the prejudice test I propose need be applied only when the state collateral attack court has failed to articulate clearly the state sentencing policy applicable to the sentencing profile petitioner has presented to the federal court. Thus, a major concern the state has expressed in this case, that of federal court intrusion on state sentencing policy, is addressed. I now discuss these points in greater detail.
I begin my analysis by observing that the question of prejudice from counsel’s failure to produce mitigating evidence is a question of determining the effect the omitted evidence would have had on the original sen-tencer. Of course, no collateral attack court,can say with certainty what this effect would have been. It simply cannot read the mind of a hypothetical sentencer in a hypothetical case. The best the court can *1275do is look to those considerations the sen-tencer would have looked to had it been presented with the omitted evidence. By examining the same considerations the sen-tencer would have examined, the court is able to make a rough judgment how the new evidence would have affected the sen-tencer.
As I discuss in Parts I and II supra, in Florida the decision whether to impose the death penalty involves a three-step process. First, the sentencer considers all the relevant evidence presented concerning the offense and the offender. The sentencer evaluates this evidence and arrives at a sentencing profile.24 Second, the sentencer looks for other sentencing decisions presenting sentencing profiles similar to the one before it. From these decisions, it gleans the state sentencing norms promulgated in those cases. Third, having determined both the sentencing profile in the case before it and the relevant policy as expressed in norms, the sentencer applies the policy to the profile and arrives at its decision. This decision involves the sentencer’s judgment about how clearly state policy addresses defendant’s case. If state policy is clear, the sentencer has less discretion within which to impose its sentence. If state policy is unclear, however, because the sentencing profile before it is materially different from those presented to state sentencers in previous cases, the sentencer has more discretion within which to make its sentencing decision. In making its decision, the sen-tencer promulgates a new sentencing norm which becomes part of state sentencing policy.
Once the above description of the sentencing process is understood clearly, the task of federal courts in deciding the issue of prejudice in connection with claims of ineffective assistance of counsel based on counsel’s failure to produce evidence becomes simple. The federal court must go through the first two steps just as the original sentencer would have done. First, it identifies the sentencing profile petitioner claims he would have presented to the sentencing court had his counsel not been incompetent. It does this by starting with the profile the sentencer described when it imposed sentence, and then altering that profile to account for the omitted evidence.
Having identified the profile, the court next engages in the second step — finding those profiles similar to petitioner’s in order to find what sentencing norms were applied to those profiles. As I discuss in more detail in this Part infra, if petitioner has presented his new profile to the state collateral attack court, therefore exhausting his claim, the federal court should have to look no further than the state collateral attack court’s dispositive order to find the norm governing petitioner’s casé. Thus, if that state court has articulated this norm clearly, the federal court’s task in engaging in the third step becomes simple.
At the third step the similarity between the role of the sentencer and that of the federal court ends. The federal court has no role in determining whether existing state sentencing policy dictates the outcome in a petitioner’s case. This is a state policy determination and, as I discuss in Part II supra in rejecting the outcome-determinative test, an area into which federal courts must not intrude. Rather, under the test I propose, the federal court’s role is to determine only whether state sentencing policy is clear enough that the sentencer would have had no substantial discretion to impose a sentence of life imprisonment. If the state policy leaves the sentencer with substantial discretion to impose either a sentence of death or of life imprisonment, the federal court must find that prejudice exists. In practical terms, this is what I mean when I say that prejudice exists when the new evidence would have materially altered the decision-making process of a rational sen-tencer.
*1276Having described the way in which the federal court applies the test for prejudice in terms of three concrete steps, my next observation is that if the state courts clearly articulate state sentencing policy and if federal courts are sensitive to dismissing unexhausted habeas petitions, the federal court’s performance of the third step is simple. In approaching the third step, one of three things should become obvious to the federal habeas judge: he must dismiss petitioner’s claim because state policy, as promulgated by the state collateral attack court, is clear that death is the appropriate sentence; he must proceed to make the determination dictated by the third step because the state collateral attack court has not clearly articulated state sentencing policy; or he must dismiss the petition because petitioner has presented an unexhausted claim. Only in the second scenario is the federal habeas court forced to apply the prejudice test I propose and thus engage in the potentially difficult canvass of state sentencing policy the test dictates. The occurrence of this scenario is, however, within the state’s power to prevent. This becomes apparent when one realizes the function of the state collateral attack court in resolving the issue of prejudice. I review this function briefly before returning to a discussion of these three scenarios.
In deciding a claim of attorney incompetence based on counsel’s failure to produce evidence, the state collateral attack court must also engage in a three-step process to determine whether the petitioner was prejudiced. The first two steps are identical to those performed by the state sentencer and the federal habeas judge: the court must find the offense and offender characteristics and describe the new sentencing profile the petitioner presents and it must identify the sentencing norms promulgated in similar cases. The crucial distinction for our purposes between a state collateral attack court and a federal collateral attack court becomes apparent at the third step. At this stage, the state collateral attack court acts like the original state sentencer in the sense that it promulgates state sentencing policy by discerning the norm in the case before it. This norm becomes part of the state sentencing policy the federal court must look to in engaging in its third step.25
Once it is understood that state collateral attack courts play a crucial role in promulgating state sentencing policy, it becomes clear that the federal court should have to look no further than the state collateral *1277attack court’s promulgation of state policy to find the norm controlling petitioner’s case. Therefore, the federal court need not engage in the sensitive canvass of state policy the third step might otherwise have called for. The simplest way to demonstrate this point is to consider the three ways in which a claim such as the one petitioner here presents can come to federal court.
The first possible scenario is that the petitioner, in alleging prejudice, presents the same sentencing profile to the federal court as he did to the state collateral attack court, and the state collateral attack court has clearly articulated state sentencing policy in determining that petitioner had failed to prove prejudice. In making that determination, the state collateral attack court decided that the death penalty is required in a case presenting the petitioner’s new sentencing profile. This interpretation of state sentencing policy is binding on the federal court. Consequently, the petitioner cannot possibly sustain his allegation of prejudice in federal court. The federal court must deny the claim because it has identified the state sentencing policy that directly controls petitioner’s case and this policy dictates the penalty of death. As I discuss in Part IV infra, this scenario occurred in this case, but the district court failed to realize that its decision was dictated by clear state policy.
The second possible scenario occurs when the federal court is presented with the identical sentencing profile presented to the state collateral attack court, and the state court has denied petitioner’s claim without clearly articulating state sentencing policy. For example, the state court may have denied petitioner’s claim without giving reasons therefor.26 Thus, the federal court would be unable to discern whether the petitioner failed to prove prejudice or incompetence or both. Because it could not tell whether petitioner proved prejudice, it would have to engage in its own determination of prejudice without the benefit of a clearly controlling state norm. Obviously, this determination should be avoided if possible. Fortunately, it can be avoided easily enough if state collateral attack courts carefully and clearly articulate sentencing policy in their decisions. This second scenario is, therefore, within the state collateral attack court’s power to prevent.
The final scenario occurs when the petitioner comes to federal court with a sentencing profile materially different from that presented to the state collateral attack court.27 The state collateral attack court has therefore not been given the opportunity to apply state sentencing policy to the profile petitioner presents to the federal court. Because the state has not been given this opportunity, the federal court must dismiss the claim for want of exhaustion. Thus, if the federal court is sensitive to dismissing unexhausted claims, this third scenario can be avoided.28
In sum, I have described how the state sentencing court, the state collateral attack court, and the federal-habeas court all engage in a three-step inquiry, to impose sentence in the case of the first court, and to determine prejudice in the case of the latter two courts. Although the first two steps are the same for all three courts, the third step varies among them. It is this crucial third step which may present difficulty for federal courts. This difficulty can be, however, avoided. Federal courts can dismiss unexhausted claims. State collateral attack courts can do their part by clearly articulat*1278ing state sentencing policy.29 Through this cooperative effort among federal and state courts much needless and detrimental litigation can be avoided.
I now turn to the errors the district court committed in this case.
IV.
In addition to applying the wrong test for prejudice, the outcome-determinative test, the district court committed three fundamental errors, each of which alone constituted reversible error. Before examining these errors, I note that each error arose from the court’s failure to apply the three-step process I have described in Part III supra. To reiterate, first the district court should have identified petitioner’s sentencing profile. Second, it should have discerned the relevant state sentencing policy. Third, it should have determined whether state policy was such that the sentencer would have had substantial discretion to impose a life sentence. Had the district court engaged in this process, it most likely would not have committed the errors described below.
The first ruling in which it was essential that the district court engage in the above process, but failed to do so, was its ruling, implicit in the scheduling of an evidentiary hearing, that the petitioner stated a claim of prejudice arising from counsel’s alleged incompetence.30 Had the court engaged in *1279the three-step process described above, it would have realized that petitioner alleged the same sentencing profile that the state collateral attack court had held warranted the death penalty. As I discuss supra Part I, the state collateral attack court stated that
as a matter of law, the record affirmatively demonstrates beyond any doubt that even if Mr. Tunkey had [presented the new mitigating evidence] at the time of the sentencing, there is not even the remotest chance that the outcome would have been any different. The plain fact is that the aggravating circumstances proved in this case were completely overwhelming, and that even to this date the Defendant cannot show that any statutory mitigating circumstances existed. The non-statutory mitigating circumstances which he claims his attorney failed to investigate and present at the time of sentencing would as a matter of law, be insufficient to outweigh the multiple aggravating circumstances present in this case.
Order Denying Post-Conviction Relief Filed Pursuant to Fla.R.Crim.P. 3.850 at 12 (emphasis in original). The Florida Supreme Court affirmed this denial of relief, concluding that “[appellant’s] claims are shown conclusively to be without merit so as to obviate the need for an evidentiary hearing. ... [W]e can find no prejudice caused to appellant, even if we assume that every allegation he has made in his petition is true.” Washington v. State, 397 So.2d 285, 286 (Fla.1981).
Because the state collateral attack court’s normative decision, as affirmed by the Florida Supreme Court, that death was the appropriate sentence for petitioner based on his sentencing profile, is binding on federal courts, the district court should have, held that petitioner could not possibly sustain his claim of prejudice. The norm as promulgated by the state collateral attack court governed petitioner’s alleged sentencing profile and dictated a sentence of death. The district court should have, therefore, dismissed the petition for failure to state a claim for relief.31 Its holding of an eviden-tiary hearing was its first error.
Having improperly determined that an evidentiary hearing was necessary, the court committed its second error during the hearing when it prevented petitioner from introducing all of his evidence on prejudice. Instead, the court relied only on the affidavits and psychiatric reports petitioner provided in support of the petition. As I noted in Part I supra, the petitioner’s counsel objected to this procedure and argued that the affidavits were merely illustrative and that he would call additional witnesses to testify on mitigation. Petitioner thus needed to introduce all of his evidence to establish his sentencing profile, from which he would then argue prejudice. Without en*1280gaging in the first step, determining petitioner’s sentencing profile, the court’s search for the relevant sentencing policy from which it could evaluate the petitioner’s claim of prejudice, was impossible. Because the court did not understand the three-step process and thus the importance of the evidence petitioner sought to introduce. to the proper resolution of the first step, it excluded the evidence. In so doing, the court ruled on the issue of prejudice by looking at an incomplete sentencing profile presumably different from that which petitioner alleged competent counsel would have established at his state sentencing trial. This clearly was error; the court should have provided petitioner the opportunity to establish his alleged sentencing profile.32
The court committed its third error when it allowed the original state sentencing judge, Judge Fuller, to testify at the hearing on the issue of prejudice. The inadmissibility of Judge Fuller’s testimony is obvious when one considers the three-step process described above. Each step of the process involves questions the federal habeas court is competent to decide independently. A federal habeas court can hear evidence and find the sentencing profile petitioner alleges his counsel should have presented to the state sentencer. It can examine the findings of fact and conclusions of law of sentencers, see Fla.Stat. § 921.141(3), and collateral attack courts, and the decisions of the Florida Supreme Court, to identify the relevant sentencing policy. Finally, it can compare this policy with petitioner’s profile and determine whether the new evidence would have materially altered the decision-making process of the sentencer.
I can see no reason why Judge Fuller’s testimony would be admissible in the district court on any of these issues. The first two questions are factual, and the district court can find these facts from the record evidence the parties adduce without the assistance of an expert witness. The third question is one of mixed law and fact, and the district court is capable also of making this determination. Judge Fuller’s “expert” testimony would be admissible only if the decision maker, here, the federal habeas court, is untrained or unqualified to determine the issues presented to it. See Fed.R. Evid. 702. Obviously, the federal habeas court is trained in the law and qualified to make the decisions required to determine prejudice. Because Judge Fuller’s opinion testimony was unnecessary, I agree with the majority's holding that it was inadmissible under Federal Rule of Evidence 702.
V.
In light of the errors described above, I would remand this case to the district court for proceedings not inconsistent with this opinion. Specifically, if the district court determines that petitioner presents the same sentencing profile to it as it did to the state collateral attack court, it must deny petitioner’s claim of ineffective assistance of counsel. If, on the other hand, it determines that petitioner advances a sentencing profile materially different from that which he presented to the state collateral attack court, it must dismiss the petition for want of exhaustion.
.I note immediately that because the issue of prejudice arising from counsel’s failure to produce evidence at a state capital-sentencing trial must be determined within the context of the state’s death penalty scheme, my discussion of prejudice in this opinion may pertain only to death sentences imposed under Florida law.
. For a fuller account of these crimes, see Washington v. State, 362 So.2d 658, 660-61 (Fla.1978).
. I note that Tunkey’s general competence as a lawyer, or the quality of his representation at , any point until the sentencing phase of the trial, is not at issue here.
. The judge also sentenced Washington to consecutive terms of imprisonment for the other offenses involved.
. Both aggravating and mitigating circumstances fall into two categories: those that aggravate, Fla.Stat. § 921.141(5)(c-i), or mitigate, id § 921.141(6)(c-e), the offense, and those that aggravate, id. § 921.141(5)(a) & (b), or mitigate, id. § 921.141(6)(a), (b), (f), & (g), the offender.
. An earlier Fla.R.Crim.P. 3.850 motion was denied October 2, 1980, without prejudice to refile, due to a lack of verification. Clemency proceedings ensued, but Governor Graham signed petitioner’s death warrant on March 13, 1981, setting execution for the week April 6-10. Washington then refiled his rule 3.850 motion, appropriately verified.
.Thus, petitioner argued neither that his “new” sentencing profile required a sentence of life imprisonment as a matter of state sentencing policy nor that it did as a matter of federal constitutional law. For a discussion of the crucial difference between the two, see note 21 infra.
. For a more extensive discussion of these affidavits, see Washington v. Strickland, 673 F.2d at 888 n. 4.
. The state’s claim that the Florida state court’s rejection of the ineffective assistance claim is entitled to 28 U.S.C. § 2254(d)’s (1976) presumption of validity requires no extensive discussion. Ineffective assistance claims are mixed questions of law and fact, and section 2254(d) is therefore wholly inapplicable. Baty v. Balkcom, 661 F.2d 391, 394 n. 5 (5th Cir.1981), cert. denied, - U.S. -, 102 S.Ct. 2307, 73 L.Ed.2d 1308 (1982); Harris v. Oliver, 645 F.2d 327, 330 n. 3 (5th Cir.), cert. denied, 454 U.S. 1109, 102 S.Ct. 687, 70 L.Ed.2d 650 (1981); Mason v. Balcom, 531 F.2d 717, 721-22 (5th Cir.1976).
Moreover, the absence of an evidentiary hearing in state court on this claim disposes of the state’s contention. The presumption referred to, 28 U.S.C. § 2254(d) (1976), applies only to
determination(s) after a hearing on the merits of a factual issue, made by a State court of competent jurisdiction in a proceeding to which the applicant for the writ and the State or an officer ... thereof were parties, evidenced by a written finding, written opinion, or other reliable and adequate written indicia
The Florida court disposed of this claim on the pleadings for failure to state a claim on which relief could be granted; therefore, section 2254(d)’s presumption is irrelevant.
.The state collateral attack court applied the standards the Florida Supreme Court enunciated in Knight v. State, 394 So.2d 997 (Fla.1981), to analyze the prejudice issue. Knight v. State, in ton, relied heavily on a plurality opinion of the Court of Appeals for the District of Columbia Circuit. See United States v. Decoster, 624 F.2d 196, 208 (D.C.Cir.1979) (en banc) (opinion of Leventhal, J., for four members of the court). The “outcome determinative” test set forth in Decoster may not now command a majority of the Court of Appeals for the District of Columbia Circuit, and it does not command a majority of this court. See note 18 infra.
I do not mean to criticize, however, the Florida Supreme Court for adopting the outcome-determinative test. As I discuss in Parts II & III infra, there are reasons why a federal court should not adopt such a test that do not apply to the state courts. In particular, state collateral attack courts may promulgate state sentencing policy, but federal courts have no power to do so. Because the outcome-determinative test requires federal courts to make provisional state sentencing policy, we must reject it. Nevertheless, the state is free to adopt any test for prejudice it chooses, subject to federal constitutional limitations. No claim is made in this case that the Florida courts may not constitutionally adopt an outcome-determinative test to determine whether a lawyer’s failure to introduce mitigating evidence at the sentencing hearing prejudiced his client, and I intimate no opinion thereon.
. This is the crucial information required to determine the relevance of 28 U.S.C. § 2254(d) and the need for an evidentiary hearing. See note 9 supra.
. The foundation for the hearing, as in all habeas cases, was the record of the sentencing and collateral attack proceedings in the Dade County Circuit Court, and the two opinions of the Florida Supreme Court reviewing those proceedings, together with copies of petition*1269er’s briefs. See Federal Habeas Rules 2, 4, and 5.
. My resolution of the prejudice issue makes it unnecessary for me to decide the question of incompetence. I intimate no opinion thereon.
. When a district court presented with a sixth amendment claim such as the one in this case elects to rule on a Fed.R.Civ.P. 41(b) motion at the close of the petitioner’s case, it must decide: (1) whether petitioner has produced mitigating evidence that would have caused the sentencing court to reconsider its decision, see Part II infra; and (2) whether the evidence indicates that lawyer incompetence — that is, ineffectiveness of constitutional magnitude— was the cause of the omission of this evidence at sentencing rather than, for example, a competent strategic choice. Both issues present mixed questions of law and fact. In answering these questions at the close of petitioner’s case, the district court of course does not make credibility choices, weigh the evidence, and decide facts; it only determines whether the petitioner has established a prima facie case.
.The court also summarily rejected fourteen additional alleged deficiencies in the state proceedings and the Florida sentencing statute, noting that although the issues had not been fully briefed, “my independent review of these [claims] reveals them to be meritless.” Record, vol. I, at 20. The district court did not explain why these additional claims were mer-itless; consequently, the panel was unable to carry out its appellate review function. The panel directed the district court, on remand, to address and dispose of each claim separately. Washington v. Strickland, 673 F.2d at 907. I concur in the panel’s instruction to the district court.
. Although the Watkins court made that observation, it had no reason to clarify the law because in that case petitioner failed to demonstrate any prejudice whatsoever. 655 F.2d at 1362-63.
. Accord Beavers v. Balkcom, 636 F.2d 114, 116 (5th Cir.1981); Mendiola v. Estelle, 635 F.2d 487, 491 (5th Cir.1981); Lovett v. Florida, 627 F.2d 706, 709-10 (5th Cir.1980); Davis v. Alabama, 596 F.2d 1214, 1221 (5th Cir.), vacated as moot, 446 U.S. 903, 100 S.Ct. 1827, 64 L.Ed.2d 256 (1979), vacated on remand, 623 F.2d 366 (5th Cir.1980); Buckelew v. United States, 575 F.2d 515, 521 (5th Cir.1978); United States v. Doran, 564 F.2d 1176, 1177-78 (5th Cir.1977), cert. denied, 435 U.S. 928, 98 S.Ct. 1498, 55 L.Ed.2d 524 (1978); see United States v. Morrison, 449 U.S. 361, 364, 101 S.Ct. 665, 668, 66 L.Ed.2d 564 (1981):
The premise of our prior cases is that the constitutional infringement identified has had or threatens some adverse effect upon the effectiveness of counsel’s representation or has produced some other prejudice to the defense. Absent such impact on the criminal proceeding, however, there is no basis for imposing a remedy in that proceeding, which can go forward with full recognition of the defendant’s right to counsel and to a fair trial.
But cf. Holloway v. Arkansas, 435 U.S. 475, 98 S.Ct. 1173, 55 L.Ed.2d 426 (1978) (when court improperly requires the same attorney to represent two defendants with conflicting interests at same trial, reversal is automatic without a showing of prejudice); Geders v. United States, 425 U.S. 80, 96 S.Ct. 1330, 47 L.Ed.2d 592 (1976) (when trial court prohibits defendant from consulting his attorney during overnight recess separating his direct testimony from his cross-examination, reversal is automatic); Herring v. New York, 422 U.S. 853, 95 S.Ct. 2550, 45 L.Ed.2d 593 (1975) (refusal to permit counsel to make a closing argument is a denial of effective assistance of counsel, and no showing of prejudice is required).
.It is significant to note that the outcome-determinative test advocated by the plurality in Decoster v. United States, 624 F.2d at 208, was developed to assess the effectiveness of counsel in the guilt phase of an armed robbery trial. I have found no case, other than the one before us, in which a federal court applied the outcome-determinative test to the sentencing phase of a capital trial. Furthermore, the plurality opinion in Decoster may not even be the controlling standard in the District of Columbia Circuit today. See United States v. Wood, 628 F.2d 554, 559 (D.C.Cir.1980) (en banc; per cu-riam) (“In order to secure a reversal, appellant must establish some basis for believing that a different kind of preparation would have resulted in the presentation of a contrary line of testimony for the jury’s consideration”). Subsequent panel opinions fail to resolve the question precisely. Compare United States v. Hinton, 631 F.2d 769, 782 (D.C.Cir.1980) (“likely prejudice” required) with United States v. Patterson, 652 F.2d 1046, 1048 (D.C.Cir.), cert. denied, 454 U.S. 904, 102 S.Ct. 412, 70 L.Ed.2d 223 (1981) (suggesting an “outcome determinative” test might be controlling).
. Cf. Holloway v. Arkansas, 435 U.S. 475, 98 S.Ct. 1173, 55 L.Ed.2d 426 (1978):
In the normal case where a harmless-error rule is applied, the error occurs at trial and its scope is readily identifiable.... But in a case of joint representation of conflicting interests the evil — it bears repeating — is in what the advocate finds himself compelled to refrain from doing .... It may be possible in some cases to identify from the record the prejudice resulting from an attorney’s failure to undertake certain trial tasks, but even with a record of the sentencing hearing available it would be difficult to judge intelligently the impact of a conflict on the attorney’s representation of a client.... Thus, an inquiry into a claim of harmless error here would require, unlike most cases, unguided speculation.”
435 U.S. at 490-91, 98 S.Ct. at 1182 (first emphasis in the original, second emphasis added;, citations omitted).
. When I refer to the sentencer I mean the sentencing circuit judge, but I note that he takes into account the recommendation of the advisory jury. Fla.Stat. § 921.141(2).
. I must emphasize that although state courts are the primary promulgators of state sentencing policy, they are, of course, subject to the Federal Constitution. Federal courts are the ultimate interpreters of that venerable document. Therefore, if a claim is made that a state sentencing decision is unconstitutional, because, for example, it is arbitrarily disparate from other sentences, the federal courts must decide that claim. Petitioner raises no such claim in this case, and I intimate no opinion thereon.
In this case, petitioner claims only that he was prejudiced as a matter of state law because he may have received a life sentence had his counsel produced the omitted mitigating evidence at his sentencing hearing. Although petitioner’s incompetency claim is based on the sixth and fourteenth amendments to the Federal Constitution, he attempts to show prejudice only as a matter of state law. Because we are dealing with a claim one aspect of which, incompetency, is based on federal law, and the other aspect, prejudice, on state law, unfortunately we may have to concern ourselves with state sentencing policy somewhat. Aside from this intrusion, which, as I discuss in Part III infra, is within the state’s power to prevent, we as federal courts have no business involving ourselves with state sentencing policy unless, of course, a constitutional claim arises.
. The use of the outcome-determinative test creates a situation analogous to the situation the Supreme Court sought to avoid by fashioning the abstention doctrine in Railroad Commission v. Pullman Co., 312 U.S. 496, 61 S.Ct. 643, 85 L.Ed. 971 (1941):
In this [case] a federal court of equity is asked to decide an issue by making a tentative answer which may be displaced tomorrow by a state adjudication.... The reign of law is hardly promoted if an unnecessary ruling of a federal court is thus supplanted by a controlling decision of a state court. The resources of equity are equal to an adjustment that will avoid the waste of a tentative decision ....
*1273312 U.S. at 500, 61 S.Ct. at 645 (citations omitted).
. The petitioner must carry his burden of proof by a preponderance of the evidence. See Walker v. Johnston, 312 U.S. 275, 286, 61 S.Ct. 574, 579, 85 L.Ed. 830 (1941) (a habeas case in which the Supreme Court required petitioner to bear “the burden of sustaining his allegations by a preponderance of the evidence”). This traditional civil burden of proof and its allocation are used because habeas corpus proceedings are civil proceedings. See Browder v. Director, Department of Corrections, 434 U.S. 257, 269, 98 S.Ct. 556, 563, 54 L.Ed.2d 521 (1978) (“It is well settled that habeas corpus is a civil proceeding”).
Judge Vance disagrees with this traditional burden and its allocation as applied in this case. He would change the model in habeas corpus cases based on an ineffective assistance of counsel claim to the following: the petitioner bears the burden of proving by a preponderance of the evidence that the ineffective assistance of counsel worked to his actual and substantial disadvantage. If petitioner satisfies this initial burden, then the burden of proof shifts to the state who must prove that counsel’s ineffectiveness was harmless beyond a reasonable doubt.
I reject Judge Vance’s proposed burden and its allocation for several reasons. First, Judge Vance’s proposal clearly flies in the face of the traditional standards of proof used in civil proceedings. In addition, Judge Vance’s proposed burden and its allocation are totally inconsistent — if either the petitioner or the state satisfies their respective burdens, then the other party cannot, as a matter of law, satisfy its burden. For example, if at the conclusion of the habeas proceeding, the petitioner has carried his burden of proving prejudice (i.e., that the ineffective assistance of counsel worked to his actual and substantial disadvan*1274tage), then it is logically impossible also to find that this prejudice was harmless beyond a reasonable doubt. Similarly, if the court finds that petitioner’s alleged prejudice was harmless beyond a reasonable doubt, then it cannot also find that petitioner proved this prejudice by a preponderance of the evidence.
I also reject Judge Vance’s proposed allocation of the burden of proof for strong policy reasons. Generally, the burden of proof is allocated to that party who has control of the evidence required to prove the claim raised in the action. In an action based on a claim of ineffective assistance of counsel due to a failure to produce evidence, this mitigating evidence is peculiarly within the control of the petitioner. Therefore, it is appropriate in such cases to allocate the burden of proof to the petitioner.
Finally, I note that Judge Vance’s reliance on Chapman v. California, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967), to shift the allocation of the burden of proof is misplaced. Chapman arose from a direct appeal of a decision by the California Supreme Court. This case arises from a collateral attack of a sentence of death by a state trial court. The different proceedings require different burdens as well as different allocations of proof. See United States v. Frady, 456 U.S. 152, 165, 102 S.Ct. 1584, 1593, 71 L.Ed.2d 816 (1982) (differences between collateral attack and direct review require differing standards of review); cf. Henderson v. Kibbe, 431 U.S. 145, 154, 97 S.Ct. 1730, 1736-37, 52 L.Ed.2d 203 (1977) (burden of proving that an erroneous instruction was prejudicial greater on collateral attack than on direct review). Thus, Judge Vance is wrong when he suggests that federal district courts in their trial of habeas corpus cases should use the standards the Supreme Court uses in its appellate review.
. As I discuss supra in Part I, in making its findings of fact and conclusions of law the sentencer classifies the offense and the offender in terms of the statutory aggravating and the statutory and nonstatutory mitigating circumstances. It then combines these characteristics into a sentencing profile that presents a comprehensive picture of all aspects relevant to the sentence.
. I note that the state collateral attack court promulgates state sentencing policy no matter what test for prejudice it uses in engaging in the third step. For example, if the state court uses an outcome-determinative test, and finds no prejudice, it determines that state policy is such that the new evidence would not have affected the outcome of the sentencer’s decision. Therefore, state policy dictates the death sentence. If the state court uses a test such as the one I propose, and finds no prejudice, it determines that state policy is such that the new evidence would not even have materially affected the decision-making process of a rational, sentencer. Again, state policy dictates the death sentence. In either case, the state collateral attack court has clearly articulated state sentencing policy, and that is all the federal court is looking for. Therefore, it makes no difference to the federal court for purposes of determining state policy which prejudice test the state court uses. As I discuss at note 10 supra, no constitutional attack is made on the state’s test for prejudice and I intimate no opinion thereon.
I note parenthetically one possible scenario, however, in which the state court’s test for prejudice may affect the third step of the federal court’s process. This situation would occur when the state court finds prejudice under a test more nebulous than the federal test, such as the panel’s “altering the trial” standard, but finds no ineffectiveness. Thus, the state court would deny the claim. Again, assuming the federal test for prejudice is clearer than the state court’s test, the state court’s promulgation of state policy might be unclear to the federal court. If the federal court, is true to its test, it must question the state court’s finding of prejudice, and this involves a sensitive canvass of state sentencing policy. I note that this problem does not arise strictly from the state court test itself, but rather from the possible unclear articulation of state policy resulting from the state court’s use of the test. I raise this scenario parenthetically, but I note that it, involves the unlikely confluence of three factors: a state court finding of prejudice, a state court finding of no ineffectiveness, and a state test for prejudice more nebulous than the federal test.
. See also the situation discussed in the second paragraph of note 25 supra.
. I note that a fourth possible scenario exists. The petitioner, in alleging prejudice, presents the same sentencing profile to the federal court as he did to the state collateral attack court, which held that petitioner proved prejudice but did not prove ineffective assistance of counsel. I do not discuss this scenario in the text because it is identical to the first scenario in that the state collateral' attack court’s finding of prejudice is binding on the federal court. The federal court need concern itself only with the incompetency aspect of the claim.
.I note that the state, as party to the habeas petition, can do its part by raising petitioner’s failure to exhaust his claim to the federal court.
. State sentencing courts can also do their part by ensuring that all relevant sentencing evidence is before them before they impose sentence. The state sentencing court can ensure this by performing two simple tasks. First, after receiving the advisory jury’s recommended sentence but before convening the parties for sentencing, it can request a presentence investigation report. This will enable the court to impose, sentence with considerable knowledge of the crime and the defendant’s background. Second, the sentencing court, prior to imposing sentence, can ask both the defendant and his counsel whether any investigation of the defendant’s background remains to be done, and whether there is additional relevant information in mitigation of the death penalty that might be presented to the court.
This discussion is not meant to suggest, however, that the sentencing court bears the primary responsibility for ensuring that all relevant sentencing evidence is before it prior to imposing sentence. In an adversary sentencing hearing, this responsibility remains with defense counsel who should look for guidance to the American Bar Association Standards for Criminal Justice, More specifically, defense counsel should consider American Bar Association Standard for Criminal Justice 18 — 6.3(f)(ii) (2d ed. 1980), which provides that counsel
should take particular care to make certain that the record of the sentencing proceedings will accurately reflect all relevant mitigating circumstances relating either to the offense or to the characteristics of the defendant which were not disclosed during the guilt phase of the case and to ensure that such record will be adequately preserved ....
. In a typical federal habeas corpus proceeding, a state prisoner begins the procedure by filing a petition pursuant to Federal Habeas Rules 2 and 3. The district court must examine this petition and any attached exhibits to determine whether it states a claim for relief. Federal Habeas Rule 4. If the facts as pled do not state a claim for relief, the court must sua sponte dismiss the petition. Id.; 28 U.S.C. § 2243 (1976). If the facts as pled do state a claim for relief, the court should then examine whether dismissal is appropriate on procedural grounds, e.g., failure to exhaust state remedies, a decision pending in state court, petitioner not in custody within meaning of 28 U.S.C. § 2254 (1976), etc. If dismissal is appropriate on procedural grounds, the court must request the state to move for dismissal on such grounds. See Advisory Committee Note to Federal Habe-as Rule 4.
If the facts as pled do state a claim for relief and dismissal is not appropriate on procedural grounds, the court must order the state to file an answer. Federal Habeas Rule 5. In its answer, the state must rebut petitioner’s allegations, indicate whether state remedies have been exhausted, and append relevant portions of the transcripts of the state proceedings. Id. With the petition, answer, and supporting documents before it, the district court must redetermine whether a claim for relief is stated. If the court now determines that a claim for relief is not stated, it must sua sponte dismiss the petition. If the court determines that a claim for relief is stated, it then must determine, pursuant to 28 U.S.C. § 2254(d), which factual issues, if any, raised by the petition require an evidentiary hearing. Therefore, when a district court orders an evidentiary hearing in a habeas case, it implicitly holds that from its examination of the pleadings and exhibits before it, petitioner has' stated a claim for relief and only factual issues need be resolved.
The district court did not adhere to this procedure in the instant case because of time constraints. As I discuss in Part I supra, petitioner filed his habeas petition in the district court on *1279April 6, 1981, just two days prior to his scheduled execution. The district court, therefore, was unable initially to examine the petition to determine (1) if it stated a claim for relief, (2) if it should be dismissed on procedural grounds, (3) if the state should file an answer, and (4) if an evidentiary hearing should be held. Instead, the court had to combine these functions in an expedited conference of counsel held on April 7, 1981. There, counsel argued whether the petition stated a claim for relief and whether an evidentiary hearing was required. Based on these arguments and the state court record subsequently supplied to the district court, the district court ordered an evidentiary hearing for April 10, 1981. In ordering this hearing, the district court implicitly held that the petition stated a claim for relief and only factual issues remained to be resolved. The district court later made this holding explicit when it stated: “It is a close question, but I was unable to find from the records before me prior to the hearing that petitioner’s allegations raised legal questions only, or that assuming all factual allegations were true, that petitioner could not have prevailed as a matter of law.” Record, vol. I, at 52.
. If petitioner had modified his petition by reciting “new” sentencing facts that materially changed the sentencing profile he had presented to the state collateral attack court, or if he had recited in his petition other relevant sentencing norms not recited to the state courts that allegedly demonstrated that his sentence was disparate, the state would no doubt, have moved to dismiss his petition for want of exhaustion and the district court would have been compelled to dismiss on that ground. See the discussion of the third scenario in Part III supra.
. Alternatively, the court could have instructed counsel to proffer his mitigating evidence but did not do so. With a proffer in the record, the court could have engaged in the first step I have described above and found the petitioner’s new sentencing profile.