SULLIVAN v. LOUISIANA

Justice Scalia, delivered the opinion of the Court.

The question presented is whether a constitutionally deficient reasonable-doubt instruction may be harmless error.

I

Petitioner was charged with first-degree murder in the course of committing an armed robbery at a New Orleans bar. His alleged accomplice in the crime, a convicted felon named Michael Hillhouse, testifying at the trial pursuant to a grant of immunity, identified petitioner as the murderer. Although several other people were in the bar at the time of the robbery, only one testified at trial. This witness, who had been unable to identify either Hillhouse or petitioner at a physical lineup, testified that they committed the robbery, and that she saw petitioner hold a gun to the victim's head. There was other circumstantial evidence supporting the conclusion that petitioner was the triggerman. 596 So.2d 177, 180-181 (La. 1992). In closing argument, defense counsel argued that there was reasonable doubt as to both the identity of the murderer and his intent.

In his instructions to the jury, the trial judge gave a definition of "reasonable doubt" that was, as the State conceded below, essentially identical to the one held unconstitutional in Cage v. Louisiana, 498 U.S. 39 (1990) (per curiam). See 596 So. 2d, at 185, and n. 3. The jury found petitioner guilty of first-degree murder and subsequently recommended that he be sentenced to death. The trial court agreed. On direct appeal, the Supreme Court of Louisiana held, consistent with its opinion on remand from our decision in Cage, State v. Cage, 583 So.2d 1125, cert. denied, 502 U.S. 874 (1991), that the erroneous instruction was harmless beyond a reasonable doubt. 596 So. 2d, at 186. It therefore upheld the conviction, though remanding for a new sentencing hearing because of ineffectiveness of counsel in the sentencing phase. We granted certiorari, 506 U.S. 939 (1992).

II

The Sixth Amendment provides that "[i]n all criminal prosecutions, the accused shall enjoy the right to a speedy and public trial, by an impartial jury . . . ." In Duncan v. Louisiana, 391 U.S. 145, 149 (1968), we found this right to trial by jury in serious criminal cases to be "fundamental to the American scheme of justice," and therefore applicable in state proceedings. The right includes, of course, as its most important element, the right to have the jury, rather than the judge, reach the requisite finding of "guilty." See Sparf v. United States, 156 U.S. 51, 105-106 (1895). Thus, although a judge may direct a verdict for the defendant if the evidence is legally insufficient to establish guilt, he may not direct a verdict for the State, no matter how overwhelming the evidence. Ibid. See also United States v. Martin Linen Supply Co., 430 U.S. 564, 572-573 (1977); Carpenters v. United States, 330 U.S. 395, 410 (1947).

What the factfinder must determine to return a verdict of guilty is prescribed by the Due Process Clause. The prosecution bears the burden of proving all elements of the offense charged, see, e. g., Patterson v. New York, 432 U.S. 197, 210 (1977); Leland v. Oregon, 343 U.S. 790, 795 (1952), and must persuade the factfinder "beyond a reasonable doubt" of the facts necessary to establish each of those elements, see, e. g., In re Winship, 397 U.S. 358, 364 (1970); Cool v. United States, 409 U.S. 100, 104 (1972) (per curiam). This beyond-a-reasonable-doubt requirement, which was adhered to by virtually all common-law jurisdictions, applies in state as well as federal proceedings. Winship, supra.

It is self-evident, we think, that the Fifth Amendment requirement of proof beyond a reasonable doubt and the Sixth Amendment requirement of a jury verdict are interrelated. It would not satisfy the Sixth Amendment to have a jury determine that the defendant is probably guilty, and then leave it up to the judge to determine (as Winship requires) whether he is guilty beyond a reasonable doubt. In other words, the jury verdict required by the Sixth Amendment is a jury verdict of guilty beyond a reasonable doubt. Our per curiam opinion in Cage, which we accept as controlling, held that an instruction of the sort given here does not produce such a verdict.* Petitioner's Sixth Amendment right to jury trial was therefore denied.

III

In Chapman v. California, 386 U.S. 18 (1967), we rejected the view that all federal constitutional errors in the course of a criminal trial require reversal. We held that the Fifth Amendment violation of prosecutorial comment upon the defendant's failure to testify would not require reversal of the conviction if the State could show "beyond a reasonable doubt that the error complained of did not contribute to the verdict obtained." Id., at 24. The Chapman standard recognizes that "certain constitutional errors, no less than other errors, may have been `harmless' in terms of their effect on the factfinding process at trial." Delaware v. Van Arsdall, 475 U.S. 673, 681 (1986). Although most constitutional errors have been held amenable to harmless-error analysis, see Arizona v. Fulminante, 499 U.S. 279, 306-307 (1991) (opinion of Rehnquist, C. J., for the Court) (collecting examples), some will always invalidate the conviction. Id., at 309-310 (citing, inter alia, Gideon v. Wainwright, 372 U.S. 335 (1963) (total deprivation of the right to counsel); Tumey v. Ohio, 273 U.S. 510 (1927) (trial by a biased judge); McKaskle v. Wiggins, 465 U.S. 168 (1984) (right to self-representation)). The question in the present case is to which category the present error belongs.

Chapman itself suggests the answer. Consistent with the jury-trial guarantee, the question it instructs the reviewing court to consider is not what effect the constitutional error might generally be expected to have upon a reasonable jury, but rather what effect it had upon the guilty verdict in the case at hand. See Chapman, supra, at 24 (analyzing effect of error on "verdict obtained"). Harmless-error review looks, we have said, to the basis on which "the jury actually rested its verdict." Yates v. Evatt, 500 U.S. 391, 404 (1991) (emphasis added). The inquiry, in other words, is not whether, in a trial that occurred without the error, a guilty verdict would surely have been rendered, but whether the guilty verdict actually rendered in this trial was surely unattributable to the error. That must be so, because to hypothesize a guilty verdict that was never in fact rendered—no matter how inescapable the findings to support that verdict might be—would violate the jury-trial guarantee. See Rose v. Clark, 478 U.S. 570, 578 (1986); id., at 593 (Blackmun, J., dissenting); Pope v. Illinois, 481 U.S. 497, 509-510 (1987) (Stevens, J., dissenting).

Once the proper role of an appellate court engaged in the Chapman inquiry is understood, the illogic of harmless-error review in the present case becomes evident. Since, for the reasons described above, there has been no jury verdict within the meaning of the Sixth Amendment, the entire premise of Chapman review is simply absent. There being no jury verdict of guilty-beyond-a-reasonable-doubt, the question whether the same verdict of guilty-beyond-areasonable-doubt would have been rendered absent the constitutional error is utterly meaningless. There is no object, so to speak, upon which harmless-error scrutiny can operate. The most an appellate court can conclude is that a jury would surely have found petitioner guilty beyond a reasonable doubt—not that the jury's actual finding of guilty beyond a reasonable doubt would surely not have been different absent the constitutional error. That is not enough. See Yates, supra, at 413-414 (Scalia, J., concurring in part and concurring in judgment). The Sixth Amendment requires more than appellate speculation about a hypothetical jury's action, or else directed verdicts for the State would be sustainable on appeal; it requires an actual jury finding of guilty. See Bollenbach v. United States, 326 U.S. 607, 614 (1946).

Insofar as the possibility of harmless-error review is concerned, the jury-instruction error in this case is quite different from the jury-instruction error of erecting a presumption regarding an element of the offense. A mandatory presumption—for example, the presumption that a person intends the ordinary consequences of his voluntary acts—violates the Fourteenth Amendment, because it may relieve the State of its burden of proving all elements of the offense. Sandstrom v. Montana, 442 U.S. 510 (1979); Francis v. Franklin, 471 U.S. 307 (1985). But "[w]hen a jury is instructed to presume malice from predicate facts, it still must find the existence of those facts beyond a reasonable doubt." Rose v. Clark, supra, at 580. And when the latter facts "are so closely related to the ultimate fact to be presumed that no rational jury could find those facts without also finding that ultimate fact, making those findings is functionally equivalent to finding the element required to be presumed." Carella v. California, 491 U.S. 263, 271 (1989) (Scalia, J., concurring in judgment). See also Pope, supra, at 504 (Scalia, J., concurring). A reviewing court may thus be able to conclude that the presumption played no significant role in the finding of guilt beyond a reasonable doubt. Yates, supra, at 402-406. But the essential connection to a "beyond a reasonable doubt" factual finding cannot be made where the instructional error consists of a misdescription of the burden of proof, which vitiates all the jury's findings. A reviewing court can only engage in pure speculation—its view of what a reasonable jury would have done. And when it does that, "the wrong entity judge[s] the defendant guilty." Rose, supra, at 578.

Another mode of analysis leads to the same conclusion that harmless-error analysis does not apply: In Fulminante, we distinguished between, on the one hand, "structural defects in the constitution of the trial mechanism, which defy analysis by `harmless-error' standards," 499 U. S., at 309, and, on the other hand, trial errors which occur "during the presentation of the case to the jury, and which may therefore be quantitatively assessed in the context of other evidence presented," id., at 307-308. Denial of the right to a jury verdict of guilt beyond a reasonable doubt is certainly an error of the former sort, the jury guarantee being a "basic protectio[n]" whose precise effects are unmeasurable, but without which a criminal trial cannot reliably serve its function, Rose, supra, at 577. The right to trial by jury reflects, we have said, "a profound judgment about the way in which law should be enforced and justice administered." Duncan v. Louisiana, 391 U. S., at 155. The deprivation of that right, with consequences that are necessarily unquantifiable and indeterminate, unquestionably qualifies as "structural error."

The judgment of the Supreme Court of Louisiana is reversed, and the case is remanded for proceedings not inconsistent with this opinion.

It is so ordered.

Chief Justice Rehnquist, concurring.

In Arizona v. Fulminante, 499 U.S. 279 (1991), we divided the class of constitutional violations that may occur during the course of a criminal proceeding, be it at trial or sentencing, into two categories: one consisting of "trial error[s]," which "may . . . be quantitatively assessed in the context of other evidence presented," id., at 307-308 (opinion of Rehnquist, C. J., for the Court), and are amenable to harmlesserror analysis; the other consisting of "structural defects," which "affec[t] the framework within which the trial proceeds," id., at 310, and require automatic reversal. There is a "strong presumption" that any error will fall into the first of these categories. Rose v. Clark, 478 U.S. 570, 579 (1986). Thus, it is the rare case in which a constitutional violation will not be subject to harmless-error analysis. See Fulminante, supra, at 309-310 (listing examples of structural errors).

The Court holds today that the reasonable-doubt instruction given at Sullivan's trial, which (it is conceded) violates due process under our decision in Cage v. Louisiana, 498 U.S. 39 (1990) (per curiam), amounts to structural error, and thus cannot be harmless regardless of how overwhelming the evidence of Sullivan's guilt. See ante, at 281-282. It grounds this conclusion in its determination that harmlesserror analysis cannot be conducted with respect to error of this sort consistent with the Sixth Amendment right to a jury trial. We of course have long since rejected the argument that, as a general matter, the Sixth Amendment prohibits the application of harmless-error analysis in determining whether constitutional error had a prejudicial impact on the outcome of a case. See, e. g., Rose, supra, at 582, n. 11. The Court concludes that the situation at hand is fundamentally different, though, because, in the case of a constitutionally deficient reasonable-doubt instruction, "the entire premise of Chapman [harmless-error] review is simply absent." Ante, at 280.

Where the jury views the evidence from the lens of a defective reasonable-doubt instruction, the Court reasons, there can be no factual findings made by the jury beyond a reasonable doubt in which an appellate court can ground its harmless-error analysis. See ante, at 280-281. The Court thus distinguishes our cases in which we have found jury instructions that create an unconstitutional presumption regarding an element of the offense subject to harmless-error review. In Rose v. Clark, supra, for example, we held that harmless-error analysis may be applied in reviewing instructions that violate the principles of Sandstrom v. Montana, 442 U.S. 510 (1979), and Francis v. Franklin, 471 U.S. 307 (1985). The "malice instruction" in Rose shifted the burden of proof on the issue of intent, in violation of due process under our decision in Sandstrom. Because the jury was instructed to presume malice from certain predicate facts, and it was required to find those facts beyond a reasonable doubt, we held that the Sandstrom error was amenable to harmlesserror analysis. 478 U. S., at 580. See also Connecticut v. Johnson, 460 U.S. 73, 96-97 (1983) (Powell, J., dissenting).

There are many similarities between the instructional error in Rose and the one in this case. In the first place, neither error restricted the defendants' "opportunity to put on evidence and make argument to support [their] claim[s] of innocence." 478 U. S., at 579. Moreover, "[u]nlike [structural] errors such as judicial bias or denial of counsel, the error[s] . . . did not affect the composition of the record." Id., at 579, n. 7. Finally, neither error removed an element of the offense from the jury's consideration, id., at 580, n. 8, or prevented the jury from considering certain evidence. (In this regard, a trial in which a deficient reasonable-doubt instruction is given seems to me to be quite different from one in which no reasonable-doubt instruction is given at all.) Thus, in many respects, the Cage violation committed at Sullivan's trial bears the hallmark of an error that is amenable to harmless-error analysis.

One may question whether, even in the case of Sandstrom error, the ability to conduct harmless-error review is dependent on the existence of "beyond a reasonable doubt" jury findings. In the typical case, of course, a jury does not make explicit factual findings; rather, it simply renders a general verdict on the question of guilt or innocence. Thus, although it may be possible to conclude from the jury's verdict that it has found a predicate fact (or facts), the reviewing court is usually left only with the record developed at trial to determine whether it is possible to say beyond a reasonable doubt that the error did not contribute to the jury's verdict. Moreover, any time an appellate court conducts harmless-error review it necessarily engages in some speculation as to the jury's decisionmaking process; for in the end no judge can know for certain what factors led to the jury's verdict. Cf. Pope v. Illinois, 481 U.S. 497, 503, n. 6 (1987). Yet harmless-error review has become an integral component of our criminal justice system. See Delaware v. Van Arsdall, 475 U.S. 673, 681 (1986); Chapman v. California, 386 U.S. 18, 22 (1967).

Despite these lingering doubts, I accept the Court's conclusion that a constitutionally deficient reasonable-doubt instruction is a breed apart from the many other instructional errors that we have held are amenable to harmless-error analysis. See, e. g., Carella v. California, 491 U.S. 263 (1989) (per curiam) (instruction containing erroneous conclusive presumption); Pope v. Illinois, supra (instruction misstating an element of the offense); Rose v. Clark, supra (instruction containing erroneous burden-shifting presumption). A constitutionally deficient reasonable-doubt instruction will always result in the absence of "beyond a reasonable doubt" jury findings. That being the case, I agree that harmless-error analysis cannot be applied in the case of a defective reasonable-doubt instruction consistent with the Sixth Amendment's jury-trial guarantee. I join the Court's opinion.