ARNOLD, Circuit Judge.
This is a death-penalty case. It comes before us again on what is essentially a successive petition for federal habeas corpus. In
Simmons v. Lockhart,
814 F.2d 504 (8th Cir.1987),
cert. denied,
485 U.S. 1015, 108 S.Ct. 1489, 99 L.Ed.2d 717 (1988), we affirmed the District Court’s denial of habeas corpus. In due course, the District Court dissolved a stay of execution that had been in effect during the habeas proceeding. Simmons, the petitioner, then filed with us a motion to recall the mandate, grant rehearing, and stay his execution in order to permit him to assert grounds for constitutional relief that had not previously been raised. We granted his request, again stayed the execution, and remanded the case to the District Court for certain findings of fact.
Simmons v. Lockhart,
856 F.2d 1144 (8th Cir.1988). The District Court
has now made its findings and has certified them to us. 709 F.Supp. 1457 (E.D.Ark.1989). After fully considering these findings and the arguments that Simmons poses against them, we again hold that no federal constitutional error affecting Simmons’s rights has occurred in this case. Although our reasoning differs somewhat from that of the District Court, we agree with it that
Simmons is not entitled to federal habeas corpus relief. Accordingly, on reconsideration, our previous judgment, affirming the dismissal of Simmons’s petition, is reinstated. The stay of execution will remain in effect to allow Simmons to seek review of our holding on certiorari.
I.
The facts of the case and the details of prior proceedings are fully recounted in our two previous opinions, which we have cited above. We will not repeat them here. It will be sufficient, in order to set the stage for this opinion, to restate briefly the questions that we asked the District Court to answer on remand. These questions had to do with two grounds for habeas relief newly asserted by Simmons. The first ground is that he was convicted under a statutory scheme for murder which is so vague as to violate the Eighth Amendment. The claim is that the capital-murder and first-degree-murder statutes in Arkansas overlap im-permissibly, giving the jury unbridled discretion to select between these two penalties, in violation of Eighth Amendment principles requiring certainty and reliability in death cases. This claim was raised in the District Court in Simmons’s original habeas petition, but his appointed counsel did not press it on appeal to this Court. We instructed the District Court to determine the reasons for this failure to press the issue. We stated: "[The District] Court should make findings as to why previous counsel failed to press the statutory-overlap issue on appeal, and decide whether that choice constituted ineffective assistance of counsel.... The District Court must determine ... why counsel failed to raise the issue, and whether a reasonably competent lawyer could have made such a decision.” 856 F.2d at 1145-46.
The second issue has to do with a claim of ineffective assistance of counsel at Simmons’s trial in the Circuit Court of Crawford County, Arkansas. In our previous opinion, we described this issue as follows:
Next Simmons contends he was denied the effective assistance of counsel because his trial attorneys were hampered by a conflict of interest. One of the lawyers who represented Simmons also represented one of the State’s key witnesses against Simmons, James Davis, on an unrelated matter. Simmons claims that because the lawyer had earlier represented Davis, the lawyer could not vigorously cross-examine the witness on matters that had come to his attention through the attorney-client relationship.
Id.
at 1146. With respect to this issue, we directed the District Court to make findings of fact as to why it had not been raised before the Arkansas courts, and, in addition, as to why it had not been raised in Simmons’s original habeas petition. We explained that the attempt to raise the issue now would “constitute[ ] an abuse of the writ, ... unless Simmons can show a good enough reason why this issue was not presented before.”
Ibid
(citation omitted).
We further directed the District Court, if it should find that Simmons is not procedurally barred from pressing this new issue at this time, to decide the legal questions that would then be raised:
Should [the District] Court find that Simmons clears [these] ... procedural hurdles, then it should resolve whether Simmons’s lawyer was faced with an actual conflict of interest that prevented him from vigorously cross-examining Davis on his military record, psychiatric problems, and any deals Davis made with the prosecution. If an actual conflict is found, then Simmons was afforded ineffective assistance of counsel at his trial, and is entitled to habeas relief. In such a conflict-of-interest situation, prejudice is presumed.
Ibid
(citations omitted).
On remand, the District Court held an evidentiary hearing and filed a thorough and comprehensive opinion, setting out its findings on the issues that we had posed. As to the statutory-overlap question, it held that Simmons’s appointed counsel's failure to press it on the initial appeal to this Court was not ineffective assistance of counsel. It was, rather, simply the exercise of a reasoned professional judgment to limit the
appeal to issues that counsel considered more likely to succeed. The District Court, citing
Stokes v. Armontrout,
851 F.2d 1085, 1088 (8th Cir.1988),
cert. denied,
488 U.S. 1019, 109 S.Ct. 823, 102 L.Ed.2d 812 (1989), held that “[t]he selective process of winnowing and focusing is the hallmark of effective appellate advocacy.” 709 F.Supp. at 1463 (internal quotations omitted). Accordingly, there was no reason to allow new counsel to press the statutory-overlap issue now. It had been, in effect, competently waived by counsel on the previous appeal.
As to the other question, the alleged conflict of interest arising from Simmons’s trial counsel’s previous representation of a prosecution witness on an unrelated matter, the District Court held Simmons procedurally barred. His previous appointed counsel had not raised this issue in the state courts in a post-conviction proceeding under Ark.R.Crim.P. 37 (now repealed), and his failure to do so did not amount to the ineffective assistance of counsel. Further, according to the District Court, to entertain the argument now would amount to an abuse of the writ. The argument had not been included in Simmons’s original habeas petition, and counsel’s decision not to include it was a reasonable one, not amounting to a violation of the Sixth Amendment right to representation by effective counsel. In the alternative, the District Court found that there was no actual conflict of interest on the part of Simmons’s trial counsel. The fact that counsel had previously represented a prosecution witness did not at all affect the vigor of cross-examination. No actual adverse effect was present, and therefore there was no real conflict of interest. Accordingly, the District Court recommended that all relief be denied.
II.
We discuss first the statutory-overlap question. We agree with the District Court and adopt its reasoning on this point. Certainly previously appointed counsel might have chosen to press this issue on appeal, and such a choice would have been reasonable. It does not follow that the opposite choice — to drop the issue — was unreasonable. Law is an art, not a science, and many questions that attorneys must decide are questions of judgment and degree. Among the most difficult are decisions as to what issues to press on appeal. Lawyers have often been told that it is not good strategy to argue on appeal every conceivable point contained in a record. We believe Simmons’s habeas counsel, in deciding not to argue the statutory-overlap question, was merely following this advice. It is possible to criticize his choice in hindsight. Perhaps a choice to press the issue would have been better. But we are dealing, after all, with fallible human beings, and a demand for perfection, even in death-penalty cases, cannot be met. The statutory-overlap issue was an arguable one, but certainly by no means ironclad, and our sense is that the decision to omit it was reasonable. The chances that the argument would succeed were not great, though they were greater than negligible, and we cannot say that counsel was ineffective because he decided to concentrate his time and energy on other points.
It is worth noting, in addition, that even if Simmons were to succeed on the merits of this question, and we were to hold the Arkansas capital-sentencing scheme unconstitutional on the overlap ground asserted, Simmons would still be under sentence of death. He has been convicted of killing four people and is under four separate death sentences. The statutory-overlap issue pertains only to three of these death sentences — cases in which the first-degree-murder instruction that Simmons claims is indistinguishable from the capital-murder instruction was given. The allegedly offending instruction was not given in the fourth case, and the death sentence in this case would stand, whatever resolution we made of the overlap issue.
III.
The conflict-of-interest question is another matter. The parties have expended much time and energy — and we understand why — debating whether the argu
ment is procedurally barred, either by the failure to raise it in the state courts, or by the failure to include it in the original habe-as petition. We deem it appropriate and prudent to resolve these contentions. We explore the permutations and combinations of procedural doctrine so well debated by the parties in order to lend clarity to this complex and multi-layered area of the law. We disagree with the State’s contention, and the District Court’s alternative holding, that Simmons’s ineffective-assistance claim is beyond our reach. Simmons has a sufficient legal reason for failing to raise his claim before now, namely, the ineffectiveness of his post-conviction counsel.
Lest there be any doubt, we hold that ineffective assistance of post-conviction counsel can be “cause” for purposes of lifting a procedural bar. Otherwise, a petitioner would be stuck with whatever his post-conviction lawyer had done, however incompetent. It is certainly true that the ineffectiveness of post-conviction counsel is not, in and of itself, a ground for habeas relief. That, however, is not the same thing as saying it cannot be cause to excuse a procedural default. The procedural default in question, of course, must be (as it is here) one that occurred in the post-conviction proceeding.
Our cases support this understanding. In
Stokes,
851 F.2d at 1091-93, after finding that a jury-instruction claim was procedurally barred, we examined an allegation (using the
Strickland v. Washington,
466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984) standard) that post-conviction counsel had been ineffective. Our inquiry points up the law. We were looking for cause for the post-conviction lawyer’s failure to raise the jury-instruction claim. While we did not find it in Stokes’s case— because we held his lawyer’s efforts to be constitutionally adequate — our search indicates at least that post-conviction ineffectiveness may be cause.
Bliss v. Lockhart,
891 F.2d 1335, 1342-43 (8th Cir.1989), echoes this conclusion. In
Bliss,
we excused a procedural default because of the ineffectiveness of (different) counsel at both the trial and post-conviction stages. We held that the trial counsel’s ineffectiveness so infected the collateral proceeding that the post-conviction counsel’s failure to raise certain issues should also be deemed constitutionally ineffective, thus excusing the procedural default that would otherwise have barred habeas review.
Has Simmons demonstrated this sort of cause? Has he shown that his post-conviction lawyer’s failure to make the conflict-of-interest argument, either in the state courts or in the original federal habeas corpus petition, fell below the standard of reasonable professional performance, and that this failure has operated to his prejudice? We first examine the failure of counsel to make the argument in the state courts. Simmons now argues that a state post-conviction proceeding raising the conflict-of-interest claim should have been filed. Faced with an imminent execution date, counsel first filed a motion for stay of execution with the Supreme Court of Arkansas, arguing that he needed a stay to allow him time to prepare a post-conviction Rule 37 petition. The Supreme Court denied the stay, at which point counsel went to the District Court with the original federal habeas petition, obtaining a stay and proceeding to litigate the issues pleaded in the petition — issues that did not include the conflict-of-interest point. The District Court, in its most recent findings now certified to us for review, held that counsel’s failure to assert the conflict-of-interest claim in a state post-conviction proceeding was not unreasonable. At the time that the motion for stay of execution in the Supreme Court was made, the District Court reasoned, counsel did not know about the conflict of interest. Later, when counsel discovered the existence of the issue, the federal habeas petition had already been filed and a stay obtained from the District Court. At that point, the District Court held, it would have been unreasonable for counsel to file a state-court post-conviction proceeding. The very filing of that proceeding would have demonstrated that state remedies had not been exhausted, thus causing the federal petition, the source of the stay of execution, to be dismissed.
We respectfully disagree with this reasoning. We accept the District Court’s finding of fact that counsel was unaware of the conflict-of-interest argument at the time of his initial filing in the state court. After the stay of execution had been secured from the federal court, however, counsel read the transcript of Simmons’s trial, 709 F.Supp. at 1466, and the conflict-of-interest claim is readily apparent from that trial transcript. On both direct and cross-examination, the witness in question, Davis, stated that Simmons’s lawyer, Settles, used to be his lawyer. Davis testified that this lawyer was the first person he told about witnessing Simmons’s kidnapping of the murder victims, and that he sought the lawyer’s advice about what to do. Tr. 2959, 2961, 2967-70. The potential conflict — when Simmons was tried for murder, he was represented by Settles, and Davis, Settles’s former client, was a key witness for the prosecution — fairly leaps off the page. At this point, we think Simmons’s appointed counsel should have at least attempted to file a Rule 37 petition with the state courts, raising the conflict-of-interest argument. Counsel must have known that claims not presented to the state courts are, with some exceptions, barred on federal habeas review. This bar could have been removed if a state post-conviction proceeding had been filed. Such a proceeding would have resulted, in all likelihood, in rejection of the conflict-of-interest argument on its merits. Such a rejection would have cleared the way for consideration of the merits of the claim by the federal habeas court.
Instead, no such state proceeding was filed, resulting in the invocation of the doctrine of procedural waiver, a doctrine that would completely bar consideration of the merits of Simmons’s claim by the federal courts. The prejudice to Simmons is clear, and counsel’s failure to avoid it by filing a state post-conviction proceeding was, in our view, professionally unreasonable. It is no answer to argue, as the District Court did, that the filing of such a proceeding would have resulted in the dismissal of the federal habeas petition and the dissolution of the stay of execution. The habeas court could have held the petition in abeyance, maintaining the stay of execution in effect, pending the determination by the state courts of the Rule 37 proceeding. See,
e.g., Collins v. Lockhart,
707 F.2d 341, 343-44 (8th Cir.1983),
cert. denied,
474 U.S. 1013, 106 S.Ct. 546, 88 L.Ed.2d 475 (1985);
Rogers v. Britton,
466 F.Supp. 397, 403 (E.D.Ark.1979),
rev’d on other grounds,
631 F.2d 572 (8th Cir.1980),
cert. denied,
451 U.S. 939, 101 S.Ct. 2021, 68 L.Ed.2d 327 (1981). We hold, accordingly, that counsel should have preserved Simmons’s conflict-of-interest claim by filing a Rule 37 petition with the Arkansas courts. Failing to do so rendered his assistance professionally unreasonable, and that ineffectiveness is sufficient cause to excuse Simmons’s failure to present the conflict-of-interest argument to the state courts.
Post-conviction counsel’s failure to make the conflict-of-interest argument in Simmons’s first habeas petition succumbs to the same analysis. Once counsel had read the transcript, he should have amended the habeas petition to make the conflict-of-interest allegation. We reach this conclusion because of the special nature of conflict-of-interest claims. If a real conflict is demonstrated, having an adverse effect on counsel’s representation of a defendant, prejudice is presumed, and the stricter
Strickland
standard for the demonstration of prejudice does not have to be met. A conflict-of-interest claim, in other words, has special advantages for a habeas petitioner. It is the kind of claim that should rarely be abandoned, and not the kind of claim (unlike the statutory-overlap issue discussed above) that can, in normal circumstances, reasonably be winnowed out in the process of selecting grounds counsel believes to be most promising. Therefore, Simmons has shown cause for the omission of the argument in his first federal petition, and he has also, necessarily, shown prejudice, because the omission of the argument, unless it can be excused, would result in dismissal of his claim at this stage without consideration of its merits. See
Fairchild v. Lockhart,
900 F.2d 1292, 1294 (8th Cir.1990). We hold, therefore, that Simmons has a
good legal reason for failing to press his claim in the original habeas petition: habe-as counsel’s defective performance. He therefore has overcome all the procedural obstacles that would bar our consideration of his claim of ineffective assistance of trial counsel based on his lawyer’s former representation of a prosecution witness.
IV.
We finally come, then, to the merits of Simmons’s claim. In general, as we have noted above, ineffectiveness of counsel requires two things: that a lawyer’s performance fall below a certain level of reasonable professional conduct, and that the client be prejudiced by this default. “Prejudice,” in this context, means that one’s confidence in the outcome of the trial is undermined — to put it another way, that there is a reasonable likelihood that, had the ineffectiveness of counsel not been present, the outcome would have been different. See
Strickland,
466 U.S. at 687, 104 S.Ct. at 2064. In a conflict-of-interest situation, however, a person challenging a conviction does not have such a high hill to climb. If there is an actual conflict of interest, prejudice is presumed, see
Strickland,
466 U.S. at 692, 104 S.Ct. at 2067, and habeas relief must be granted. The conflict of interest, however, must be actual, not merely theoretical. Under the cases, this appears to mean that there must have been some actual adverse effect on the defendant’s case. The mere fact, for example, that one of Simmons’s lawyers in the state trial court had previously represented a person later called as a prosecution witness against Simmons, does not suffice to entitle Simmons to relief. Simmons must show, in addition, that this dual representation made some difference, and that this difference was adverse to his defense. The difference, we take it, need not be so great as to meet the general
Strickland
standard (reasonable likelihood of a different result). But it must be an actual and demonstrable effect of the conflict, and not merely an abstract or theoretical one.
To put it in concrete terms for purposes of the present case, how was Simmons’s defense hurt by the fact that one of his trial lawyers had previously represent-' ed the witness Davis? We are not convinced that his case was hurt at all. We have studied the transcript of the trial, with particular reference to counsel’s cross-examination of the witness Davis, and we are unable to fault that effort at discrediting Davis and his testimony. Simmons suggests that counsel, who had learned from confidential communications from Davis that he had been taking lithium, should have cross-examined Davis about this (and about his underlying mental condition) in order to try to show that his identification of Simmons as a person seen with the victims of the crime was incorrect. Counsel testified before the District Court that he decided not to raise the lithium issue as a matter of trial strategy. In light of his knowledge of Arkansas juries, he did not think that the strategy would be effective. This is the kind of judgment, right or wrong, that trial lawyers regularly make, and we are not convinced that it was incorrect. More to the point, we cannot see how the fact that counsel had previously represented Davis affected this judgment in the least. Counsel testified that he felt no constraints in his cross-examination of Davis. He believed that bringing up the lithium matter would have made it appear that he was trying to humiliate the witness, and might have alienated the jury. A counsel entirely free of conflict of interest, one who had not previously spoken with Davis on a confidential basis, might not have known about the lithium at all.
There is a further difficulty with Simmons’s contention that more vigorous cross-examination about Davis’s medical condition and medication would have made a difference. The District Court noted in its “Preliminary Statement” that “[djuring the pertinent period (prior to August 1981), there is nothing to indicate any problem with perception, hallucinations, orientation or memory.” 709 F.Supp. at 1460. That finding is not clearly erroneous. We have reviewed the contemporaneous medical evidence. It is at best indeterminate in terms of its usefulness for detracting from Davis’s testimony. It is true that Davis
was diagnosed as schizophrenic in 1980. It is also true that the records contain a repeated contrary diagnosis, by one of Davis’s treating physicians at the Veterans Administration hospital, of mood swings. As the District Court noted, this later diagnosis of mood swings specifically quarrels with the earlier conclusion. This medical dispute works against the petitioner’s claim here, since Davis’s supposedly impaired ability to perceive the world around him is rooted in his alleged schizophrenia. Moreover, it was not until six days after Davis testified that yet another doctor opined that Davis had a “relatively poor” ability to perceive reality. Though the petitioner places great weight on this determination, it would not have been available for cross-examination purposes. And the rest of the relevant medical evidence supports the District Court’s finding that neither Davis’s medical condition nor his medication had any appreciable effect on his ability to perceive events.
In the face of these medical disagreements, we cannot say that Settle’s failure to secure and use these records materially weakened his cross-examination. This is especially so given our overall agreement with the District Court’s conclusion that Settle’s efforts to discredit Davis were wide-ranging and vigorous. Just as choices must be made about what claims to raise on appeal, decisions must be made about what to investigate about witnesses. Settle’s choice not to delve into Davis’s medical background was not unreasonable, given his concerns about juror sympathy. The best cross-examiner is one who knows when to stop, and counsel’s decision was within the range of constitutionally acceptable lawyering. The opportunities for rehabilitation offered by the medical evidence as a whole — had it been developed before trial as we have it before us now — also support our conclusion that this game was probably not worth the candle. The substantive indeterminacy of this medical evidence, and its probable lack of impact, thus also justify the District Court’s decision not to receive expert testimony on this issue.
There is also the matter of Davis’s military record. In our previous opinion, we instructed the District Court to make findings as to whether a conflict of interest prevented counsel “from vigorously cross-examining Davis on his military record.” 856 F.2d at 1146. The District Court made the following findings with respect to this question: “Settle pulled no punches. On cross-examination he brought out that Davis was convicted of a felony by court martial in Vietnam for ‘destruction of villages with civilians.’ Apparently, Davis had participated in a My Lai-type massacre. It is hard to conceive how more damaging testimony could have been elicited.” 709 F.Supp. at 1470-71. We agree fully with this reading of the record.
Finally, there is the matter of a “deal” that the prosecution may have made with Davis in order to get him to testify against Simmons. Counsel had originally represented Davis on a charge of theft by deception, and this charge was still pending at the time of the Simmons trial. (Shortly after being appointed to represent Simmons, counsel had withdrawn from the Davis case, with leave of court.) If counsel, feeling perhaps some solicitude on account of his previous representation of Davis on the theft charge, had failed to bring this charge up on cross-examination, Simmons would have a real point here. But this is not what happened. Counsel did get before the jury the fact that a charge of theft by deception was pending against Davis, thus raising the possibility that Davis might be testifying in hopes of getting favorable treatment in his own case, either because of a deal that had already been made or because of a deal that he hoped would be made. If there had actually been a “deal,” and if counsel had in fact negotiated this deal for Davis, then an obvious and serious conflict would exist, but there is no proof of this. Simmons’s brief does not even suggest it.
In short, although counsel might have been better advised to get out of both the Simmons and Davis cases, we do not believe that his representation of Simmons was harmed in any way by the fact that he had previously been Davis’s lawyer. He cross-examined Davis every bit as vigor
ously as a lawyer without an arguable conflict would have done. In short, although Settle’s previous representation of Davis creates a theoretical conflict of interest, one justifying the extra analysis that has now been performed by the District Court and this Court, it had no actual adverse effect on Simmons’s defense. Prejudice, therefore, cannot be presumed, and Simmons’s claim of ineffective assistance of counsel based on a conflict of interest must fail.
V.
In sum, we reject both of Simmons’s newly asserted grounds for habeas relief. We deal with what amounts to a second petition for habeas corpus. The first issue, the question of statutory overlap, is barred as an abuse of the writ: The decision not to press it on the appeal of the denial of the first habeas petition was a reasonable one, and Simmons is bound by this decision. The question, in any event, would not affect one of the four death sentences to which Simmons is subject. As to the alleged conflict of interest, even though it is procedurally open at the present time, the argument is without merit. No actual adverse effect on Simmons’s defense has been shown. Therefore, on rehearing, the previous judgment, affirming the denial of the writ of habeas corpus, is reinstated, and our mandate (subject to the right of appellant to petition for rehearing, with or without a suggestion for rehearing en banc) will issue in due course.
In previous cases involving second or successive petitions for habeas corpus, it has been our custom, if we hold that an abuse of the writ has occurred, to dissolve any stay of execution at the time that our mandate is issued. In this situation, any further stay would have to be obtained from this Court en banc or from the Supreme Court or a Justice thereof. We follow a different course here, because in our view this case is in a distinguishable procedural posture. Here, one of Simmons’s arguments — the question of conflict of interest — has been dealt with on its merits. This argument was not previously disposed of by this Court, not having been raised at the time of the previous appeal. It could therefore not have been included in Simmons’s petition for certiorari, filed after our previous judgment. We therefore believe it appropriate to follow our ordinary practice, applicable to initial habeas petitions, of continuing the stay of execution in effect pending petition for certiorari to the Supreme Court of the United States.
Judgment of affirmance reinstated; stay of execution to remain in effect pending final disposition by the Supreme Court of the United States, in the event that a timely petition for certiorari is filed.
We are grateful to present appointed counsel for their zealous and effective representation of Simmons. Their performance and that of their law firm has been exemplary.