specially concurring in part and dissenting in part:
I concur in Section IV of the majority opinion, affirming the judgment of the district court in granting petitioner a new sentencing trial because he was sentenced by an unconstitutionally selected jury. I also concur in Section 11(a) of the majority opinion, which holds that a Sandstrom violation occurred. However, I dissent from Section 11(b) of the opinion, which holds the Sandstrom violation to be harmless error, and Section III, which holds that the petitioner was not prejudiced by improper prosecutorial argument at the sentencing trial.
I. THE SANDSTROM ISSUE
I acknowledge that I am bound by the recent decisions of the en banc court in Davis v. Kemp, 752 F.2d 1515 (11th Cir. 1985) (en banc), and Tucker v. Kemp, 762 F.2d 1496 (11th Cir.1985) (en banc). However, the majority’s opinion unjustifiably extends the harmless error concept to a case where the intent of the admitted killer was clearly at issue.
In Davis, the Court identified two circumstances under which the harmless error rule might be applicable to Sandstrom violations: (1) where the evidence of the defendant’s guilt was overwhelming, and (2) where the invalid instruction concerned an element of the crime which was not at issue at trial. Id. at 1521; Lamb v. Jernigan, 683 F.2d 1332 (11th Cir.1982), cert. denied, 460 U.S. 1024, 103 S.Ct. 1276, 75 L.Ed.2d 496 (1983). In Davis, the faulty instruction concerned the intent of the defendant, but the Court held that the defendant did not contest the issue of intent. See Davis, supra, 752 F.2d at 1521. The Court held that under those circumstances the faulty instruction, which impermissibly shifted the burden of proof on the intent issue to the defendant, was harmless error. In reaching its conclusion, the Court relied heavily on Engle v. Koehler, 707 F.2d 241 (6th Cir.1983), aff'd by an equally divided court, — U.S. —, 104 S.Ct. 1673, 80 L.Ed.2d 1 (1984) (per curiam), in which the Sixth Circuit held that a Sandstrom violation could be harmless error where the defendant admits that an intentional, malicious killing occurred but claims non-participation in the crime; but that such a violation could not be harmless error where the defendant asserts lack of mens rea. See Davis, supra, 752 F.2d at 1521; Engle, supra, 707 F.2d at 246.
Although intent was not contested in Davis, intent was very much at issue in the instant case. The petitioner raised a temporary insanity defense, which called into question his capacity to form the requisite intent for malice murder. This is the same defense that was raised in Engle, in which the Sixth Circuit held that a Sandstrom violation was not harmless error. 707 F.2d at 246. Nevertheless, the majority holds that the Sandstrom violation, which may have given some jurors the impression that the petitioner had the burden of proving that he lacked the requisite intent for malice murder, was harmless error.
The majority points out that the faulty jury instruction in this case stated that one is presumed to intend the natural and probable consequences of his acts if he is a person of “sound mind”. The majority argues that this presumption could have become operative only if the jury first determined that the petitioner was sane, i.e., if the jury had already rejected the petitioner’s insanity defense. The majority concludes that if the presumption had become operative, and the Sandstrom violation thus occurred, the violation was rendered harmless by the fact that petitioner’s de*690fense on the issue of intent had already been rejected by the jury.
The majority’s analysis fails to address the fact that the State was required to prove not that petitioner had a sound mind but, rather, that he had the requisite intent for malice murder. Even if the jury rejected petitioner’s argument of temporary insanity, the issue of intent was not a defense that petitioner had to prove; intent was an element of the crime that the State had the burden of proving beyond a reasonable doubt. Davis, supra, 752 F.2d at 1528 (Johnson, J., dissenting). By arguing that he had a diminished capacity to form the requisite intent for malice murder, defendant did not thereby admit that he had the requisite intent if he had the capacity to form it. Since intent was a contested issue, holding a Sandstrom error harmless in the present case is an unwarranted extension of the “harmless error” rule.
II. THE PROSECUTORIAL MISCONDUCT ISSUE
In the present case, the prosecutor argued that a “noted justice” had stated that mercy would be an inappropriate consideration for the defendant. The prosecutor implored the jury to “fulfill the commands of the law,” and thus inflict the death penalty on the defendant. The majority holds that, despite this concededly improper argument, the prosecutor did not render the sentencing phase of the trial fundamentally unfair, and appellant is not entitled to habeas relief on this ground. The majority relies on the “prejudice” test adopted by this Circuit in the en banc decision Brooks v. Kemp, 762 F.2d 1383, 1413 (11th Cir. 1985) (en banc), which held that a court should grant habeas relief for improper prosecutorial argument at the sentencing phase only if there is a reasonable probability that, in the absence of such argument, the death penalty would not have been imposed.
In holding that the improper prosecutorial argument was not grounds for relief, the majority disregards the recent Supreme Court decision in Caldwell v. Mississippi, — U.S. —, 105 S.Ct. 2633, 86 L.Ed.2d 231 (1985), in which the Court vacated a death sentence because the prosecutor used the same kind of improper argument that was used in the present ease. Because the recent Supreme Court decision rejects the “prejudice” test adopted in Brooks, I dissent from Section III of the majority opinion.
A. Caldwell Rejects the Brooks Prejudice Test
In vacating the defendant’s death sentence because of improper prosecutorial argument, the Caldwell Court applied a prejudice standard that differs significantly from the 11th Circuit standard formulated in Brooks and relied upon by the majority. In Brooks, the en banc court held that to determine whether to grant habeas relief for improper prosecutorial arguments the reviewing court must decide whether there is a reasonable probability that, had the remarks not been made, the sentencing outcome would have been different. Brooks, supra, at 1402. A “reasonable probability” is a probability sufficient to undermine confidence in the outcome. Drake v. Kemp, 762 F.2d 1449, 1458 (11th Cir.1985) (en banc); Strickland v. Washington, 466 U.S. 668, _, 104 S.Ct. 2052, 2068, 80 L.Ed.2d 674, 698 (1984).
The basis for the standard adopted in Brooks was the notion that habeas relief should be available only when an error has affected the “fundamental fairness” of the challenged proceeding. This “fundamental fairness” standard originated in Donnelly v. DeChristoforo, 416 U.S. 637, 94 S.Ct. 1868, 40 L.Ed.2d 431 (1974). In that ease, the Supreme Court set forth the standard for reviewing habeas corpus petitions raising the impropriety of a state prosecutor’s argument in the guilt phase of a noncapital offense. In holding that the relevant inquiry was whether the remark violated due process, the majority stated that “not every trial error or infirmity ... constitutes a ‘failure to observe that fundamental fairness essential to the very concept of *691justice.’ ” Brooks, supra, at 1400, citing Donnelly, 416 U.S. at 642, 94 S.Ct. at 1871.
In order to elaborate on the fundamental fairness standard, the Brooks Court adopted the prejudice standard used by the Supreme Court in Strickland v. Washington, supra. The Brooks Court held that the Court in Strickland, “while addressing a specific Sixth Amendment violation, recognized that ‘fundamental fairness’ is the central concern of the writ of habeas corpus.” Brooks, supra, at 1401, citing Strickland, 466 U.S. at _, 104 S.Ct. at 2070, 80 L.Ed.2d at 700. The Brooks Court continued, “Thus, the [Strickland ] Court acknowledged that fundamental fairness, the same standard adopted in Donnelly, is the governing principle in reviewing errors of counsel. The [Strickland Court’s] use of the ‘reasonable probability’ test to elaborate the underlying principle suggests its applicability to other areas in which fundamental fairness is the guide.” Brooks, supra, at 1401. In short, Brooks held that the standard used to test fundamental fairness in the context of ineffective assistance of counsel claims is equally applicable in the context of prosecutorial misconduct claims in capital cases.
In support of its adoption of the “reasonable probability” test, the Brooks Court argued that that test was “consistent with the standards discussed in Donnelly and with subsequent eases applying the fundamental fairness standard.” Id. The Court claimed that those cases indicated that, in determining whether improper argument had a prejudicial impact on the sentencing proceeding, it was necessary to look beyond the types of argument used and to look at the strength of the evidence as well. The Court concluded, “Even argument greatly exceeding the bounds of propriety will not be fundamentally unfair in the guilt phase of a case with overwhelming evidence because of the low probability of the argument’s impact.” Id. at 1401-02 (footnote omitted).
The majority in the present case relies on the standard adopted in Brooks in determining that the improper prosecutorial argument was not a ground for vacating the death sentence. The majority holds that there was no reasonable probability that, absent the improper statements of the prosecutor, Bowen would not have been sentenced to death. Supra at 682.
In concluding that the prosecutor’s remarks required reversal of the death sentence, the Caldwell Court did not refer to a “reasonable probability” standard. The Court consistently emphasized that the Eighth Amendment requires a heightened standard of reliability in the capital sentencing decision. See Caldwell, supra, — U.S. at —, —, 105 S.Ct. at 2639, 2646. The Court concluded, “Because we cannot say that this effort had no effect on the sentencing decision, that decision does not meet the standard of reliability that the Eighth Amendment requires.” Id. at —, 105 S.Ct. at 2646.
The “reasonable probability” test adopted by the en banc court in Brooks is incompatible with Caldwell for three reasons. First, the “reasonable probability” test is inconsistent with the Caldwell Court’s language in the final section of the majority opinion. The Court held that, for the sentencing phase to satisfy the Eighth Amendment’s heightened standard of reliability, the reviewing court must be able to say that the improper argument “had no effect on the sentencing decision.” Id. This is a far more difficult standard for the State to satisfy than the Brooks standard. Unlike the Brooks standard, which places the burden on the defendant to show prejudice, the Caldwell standard places the burden of proving the absence of prejudice on the State. Prejudice will be presumed unless the reviewing court can say that the improper argument had no effect on the jury’s decision. See id. In addition, the State cannot overcome this presumption merely by showing that there is a “reasonable probability” that the improper argument did not affect the outcome. The State must prove that the improper argument “had no effect.”
The second reason the Brooks standard is incompatible with Caldwell is that *692Brooks treated the “reasonable probability” test as a monolithic standard to be applied in every habeas corpus proceeding, regardless of whether the error complained of occurred under the Fifth, Sixth, or Eighth Amendment. See Brooks, supra, at 1399-1401. In specifically alluding to the heightened need for reliability in the capital sentencing phase required by the Eighth Amendment, the Caldwell Court suggested that different prejudice tests are required for the various types of constitutional challenges to the fundamental fairness of a state trial. Caldwell, supra, — U.S. at —, 105 S.Ct. at 2645.
The third reason the Brooks standard should no longer be followed is that, although that standard requires a reviewing court to take into account the strength of the evidence against the defendant, Brooks, supra, at 1401-02, the Caldwell Court looked solely at the type of argument used, in the context of the entire argument and instructions, in determining whether the sentencing phase might have been rendered unfair. The Caldwell Court did not take the strength of the evidence into account at all in determining whether the improper argument affected the fairness of the sentencing phase. See Caldwell, supra, — U.S. at —, 105 S.Ct. at 2644-45. On the contrary, the Court suggested that a reviewing court could not accurately determine whether a jury might have returned a different sentence in the absence of improper argument. The Court said, “It is beyond question that an appellate court, performing its task with a presumption of correctness, would be relatively incapable of evaluating the ‘literally countless factors that [a capital sentencer] consider[s]’ ... in making what is largely a moral judgment of the defendant’s desert.” Id. at - n. 7, 105 S.Ct. at 2645 n. 7 (citation omitted). In short, the Caldwell Court applied a prejudice standard more closely resembling the harmless error test of Chapman v. California, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967), than the prejudice test of Strickland.
In addition to indicating that the Brooks prejudice standard was wrongly adopted, Caldwell indicates specific flaws in the way the Brooks standard has been applied. The dissent in Caldwell argued that the impact of the prosecutor’s argument regarding appellate review was mitigated by later prosecutorial comments that the jury played an important role in the sentencing process. Caldwell, supra, — U.S. at —, 105 S.Ct. at 2650 (Rehnquist, J., dissenting). However, the majority explicitly rejected this argument of the dissent, on the grounds that, “even if the prosecutor’s later comments did leave the jury with [the] view that they had an important role to play, the prosecutor did not retract, or even undermine, his previous [improper comment].” Caldwell, supra, at — n. 7, 105 S.Ct. at 2645 n. 7. The Court also emphasized the fact that the trial judge failed to issue strong curative instructions directed specifically at the improper argument. Id. at —, 105 S.Ct. at 2645.
The majority in the present case places heavy emphasis on the fact that “[a]ny prejudice from the improper remarks was for the most part alleviated by other statements of the prosecutor and the judge’s charge.” Supra at 682. The majority ignores the significance of the fact that neither the trial judge nor the prosecutor made any attempt to cure or retract the specific remarks that the majority concedes were improper. Yet, Caldwell clearly holds that general remarks to the effect that the jury bears responsibility for the sentencing decision are not sufficient to cure the prejudice caused by improper prosecutorial arguments aimed at diminishing the jury’s sense of responsibility. Caldwell, supra, at —n. 7, 105 S.Ct. at 2645 n. 7.
B. Caldwell Controls the Outcome in the Present Case
In Caldwell, the prosecutor argued that the jury would not bear responsibility for the imposition of a death sentence, because a death sentence would automatically be reviewed by an appellate court. The Court said that a capital sentencing jury, made up of individuals placed in a very unfamiliar *693situation and called on to make a very difficult and uncomfortable choice, might welcome the opportunity to diminish the importance of its role and delegate its decision making authority to others. The Court said,
This problem is especially serious when the jury is told that the alternative decision makers are the justices of the state supreme court. It is certainly plausible to believe that many jurors will be tempted to view these respected legal authorities as having more of a “right” to make such an important decision than has the jury.
Id. at —, 105 S.Ct. at 2642.
The essence of the impropriety in Caldwell was the implication in the prosecutor’s argument that the jury did not have the primary responsibility for deciding whether the death sentence was appropriate but, rather, that the jury had a duty to impose the death sentence in order to give effect to the decision of other authorities who were better able to judge the appropriateness of that penalty. Such authorities might include not only the justices of a state supreme court but also the prosecutor, grand jury, or police.
In the present case the prosecutor argued not only that a noted state supreme court justice had already determined that the death penalty was appropriate for the defendant, but also that the law itself commanded such a sentence. This argument went beyond what the Court condemned in Caldwell. Since the danger that the jury might choose to minimize the importance of its role was at least as high in the present case as it was in Caldwell, the analysis the Supreme Court used in deciding to vacate the death sentence in Caldwell should control the outcome in the present case. Because we cannot say that the improper arguments in the present case had no effect on the jury’s sentencing decision, I dissent.