* I must also write, however, to point out an aspect of the trial judge's sentencing decision that violates indispensable principles contained in the prevailing death penalty jurisprudence of this Court. The trial judge in this case overturned a jury recommendation of life and sentenced the defendant to die, in part out of a belief as to the law that is wholly inconsistent with the constitutional principles of Eddings v. Oklahoma, 455 U.S. 104, 102 S. Ct. 869, 71 L. Ed. 2d 1 (1982), and Lockett v. Ohio, 438 U.S. 586, 98 S. Ct. 2954, 57 L. Ed. 2d 973 (1978).

In reviewing the jury recommendation against death, the sentencing judge explained that he saw no mitigating circumstances, and that he believed the jury had based its recommendation on an "invalid" mitigating circumstance stemming from residual feelings of doubt as to guilt. Because he saw this as a legally improper mitigating circumstance he felt little hesitancy in putting their recommendation to the side. Since the sentencing here, the Florida Supreme Court seems to have added its voice in support of the proposition that lingering doubts as to guilt cannot be a ground for mitigating a death sentence.

"A convicted defendant cannot be 'a little bit guilty.' It is unreasonable for a jury to say in one breath that a defendant's guilt has been proved beyond a reasonable doubt and, in the next breath, to say someone else may have done it, so we recommend mercy." Buford v. State, 403 So. 2d 943, 953 (1981), cert. denied, 454 U.S. 1164, 102 S. Ct. 1039, 71 L. Ed. 2d 320 (1981).

The error of the sentencing judge in this case thus seems to have been enshrined in Florida law.

This Court, in Lockett and then more decisively in Eddings, held that any aspect of the case that could rationally support mitigation must be deemed a legally valid basis for mitigation. There is certainly nothing irrational—indeed, there is nothing novel—about the idea of mitigating a death sentence because of lingering doubts as to guilt. It has often been noted that one of the most fearful aspects of the death penalty is its finality. There is simply no possibility of correcting a mistake. The horror of sending an innocent defendant to death is thus qualitatively different from the horror of falsely imprisoning that defendant. The belief that such an ultimate and final penalty is inappropriate where there are doubts as to guilt, even if they do not rise to the level necessary for acquittal, is a feeling that stems from common sense and fundamental notions of justice. As such it has been raised as a valid basis for mitigation by a variety of authorities.

The wisdom behind mitigating death sentences in the face of doubts as to guilt led the drafters of the Model Penal Code to include that factor in their model death penalty statute as a mitigating factor so strong that its presence would exclude the possibility of death as a matter of law.

"Death Sentence Excluded. When a defendant is found guilty of murder, the Court shall impose sentence for a felony of the first degree [i.e., a noncapital offense] if it is satisfied that:

* * * * *

"(f) although the evidence suffices to sustain the verdict, it does not foreclose all doubt respecting the defendant's guilt." ALI, Model Penal Code § 210.6(1), p. 107 (Off.Draft, 1980).

In its explanatory comments the authors made their intentions clear:

"This provision is an accommodation to the irrevocability of the capital sanction. Where doubt of guilt remains the opportunity to reverse a conviction on the basis of new evidence must be preserved, and a sentence of death is obviously inconsistent with that goal." § 210.6(1), Comment 5, p. 134.

The Fifth Circuit has also acknowledged the validity of the mitigating factor:

"The fact that jurors have determined guilt beyond a reasonable doubt does not necessarily mean that no juror entertained any doubt whatsoever. There may be no reasonable doubt—doubt based upon reason—and yet some genuine doubt exists. It may reflect a mere possibility; it may be but the whimsy of one juror or several. Yet this whimsical doubt this absence of absolute certainty—can be real.

"The capital defendant whose guilt seems abundantly demonstrated may be neither obstructing justice nor engaged in an exercise in futility when his counsel mounts a vigorous defense on the merits. It may be proffered in the slight hope of unanticipated success; it might seek to persuade one or more to prevent unanimity for conviction; it is more likely to produce only whimsical doubt. Even the latter serves the defendant, for the juror entertaining doubt which does not rise to reasonable doubt can be expected to resist those who would impose the irremedial penalty of death." Smith v. Balkcom, 660 F.2d 573, 580-581 (1981), cert. denied, 459 U.S. 882, 103 S. Ct. 181, 74 L. Ed. 2d 148 (1982).

Once again, the factor was presented as a valid one in Justice STEVENS' partial dissent in Spaziano v. Florida, 468 U.S. 447, 467, 104 S. Ct. 3154, 3166, 82 L. Ed. 2d 340 (1984). The dissent noted that the Spaziano jury "may well [have been] sufficiently convinced of [Spaziano's] guilt to convict him, but nevertheless also sufficiently troubled by the possibility that an irrevocable mistake might be made . . . that [it] concluded that a sentence of death could not be morally justified in this case." Id., at 488, n. 34, 104 S. Ct., at 3177, n. 34. We did not then realize, however, that Florida seems to have denigrated this wholly proper mitigating factor as a matter of law.

Standing in sharp contradistinction to the Eddings and Lockett holdings, the refusal by Florida courts to consider this valid factor should be reviewed.

Although this legal issue might arise in a sentencing scheme that respected the finality of a jury decision against death, it is not necessarily an accident that this case came from Florida's system, where judges may override jury decisions for life. In Spaziano, which upheld the Florida system, the dissent cited data showing that judges are usually more likely to impose death than are juries. 468 U.S., at 488, n. 34, 104 S. Ct., at 3177, n. 34. Where fully informed and rational juries decide that death is inappropriate, a judge's override rarely if ever shows that, in spite of the jury, there was really no reasonable doubt but that death was appropriate. It is far more likely that the override decision simply fails to account for the more intangible or less traditionally "legal" elements of mitigation that informed a jury's decision. This case seems a clear illustration. Indeed, the Spaziano dissent correctly suspected that a system of death sentencing by judges—"trained to distinguish proof of guilt from questions concerning sentencing"—would be less likely to recognize this valid mitigating circumstance. Ibid.

Eddings and Lockett entitle a defendant to a sentencer who can consider all mitigating circumstances, whether or not they conform to traditional legal categories. The weight assigned to any element can only be a function of the values of the community, for certainly there is no "objective" formula. Once a mitigating circumstance is considered, assigned weight, and determined to be sufficient to preclude death, the Constitution should allow no "superior" authority to remove that circumstance from the equation and impose death. To allow a judge to override a jury decision for life—as this Court did in Spaziano —not only places the defendant in the fundamentally unfair position of having to repetitively justify the appropriateness of one's continued life, it also facilitates the denigration of a variety of mitigating circumstances.

Only a narrow reliance on legal formalism could lead one to question whether doubts as to guilt can be insufficient to preclude acquittal but sufficient to preclude death. The judge here concluded that the jury confronted the horror of sending to his death a defendant who might be innocent. But rather than accept that this was weighty logic in favor of life, he assumed that it was a form of logic that should be declared invalid by Florida courts. I dissent from the denial of the petition in this case.