with whom Justice Blackmun joins as to Part B, concurring in part and concurring in the judgment.
I agree with the Court’s carefully constructed methodolgy for determining harmless error with respect to unlawful presumptions, but I disagree concerning its application to the facts of the present case. Unlike the Court, I find the “deadly weapon” presumption harmless; I find the “unlawful act” presumption not harmless, but for reasons other than the Court assigns. I therefore concur in the judgment of *412reversal and join all except footnote 6 and Part III of the Court’s opinion.
A
In my view the “deadly weapon” presumption was harmless for the simple reason that it had no application to the facts of the case. It disappeared (“burst”) “‘when the circumstances surrounding the use of [the] deadly weapon [were] put in evidence and testified to.’” Ante, at 397 (quoting App. 96).
The Court apparently does not disagree with that, if the jury can be presumed to have taken the “presumption is removed” portion of the instruction seriously. The Court believes, however, that “a [reasonable] juror would have felt obliged to give the presumption some application” because the instructions creating and qualifying it were “inherently contradictory.” If they were taken literally, the Court reasons, the very evidence establishing the presumption would cause it to vanish. Ante, at 401, n. 6. I find no such contradiction. It seems to me quite possible to prove that a deadly weapon was used without proving the circumstances surrounding that use. The victim, for example, is found dead of a gunshot wound and the defendant is shown to have been the only person with access to the victim, and to have been in possession of the gun that fired the fatal shot. Or even more simply (and as was the case here), both sides concede that a deadly weapon was used. To be sure, a jury would often confront practical difficulty in applying the presumption (as opposed to theoretical difficulty in understanding it, because of its “inherent contradiction”), in that it would frequently be a nice question whether a particular factual showing is only enough to establish use or also enough to establish “circumstances” as well. But I hardly think that is a problem here. Any reasonable juror must have thought that “circumstances surrounding the use” were placed in evidence when the multiple details described in Part I of the Court’s opinion were introduced, including the fact that Davis stabbed Mrs. Wood *413while engaged in a struggle with her and her son, during which “‘all three . . . stumbled around the counter, out in the aisle.’” Ante, at 395 (quoting App. 19) (emphasis added). If we take the assumption that juries follow their instructions seriously, Richardson v. Marsh, 481 U. S. 200, 211 (1987), I think we must conclude that this presumption disappeared and was therefore harmless beyond a reasonable doubt.
B
The “unlawful act” presumption is a different matter. That did not utterly disappear upon the introduction of certain evidence, but was merely, in the words of the instruction, “not conclusive” and was “rebuttable by the rest of the evidence.” App. 96. The Court concludes that this was not harmless only after looking to the entire record and determining that it “simply is not clear on Davis’ intent to kill the victim,” ante, at 411. I agree with the Court’s conclusion that this presumption was not harmless; but I think that conclusion should have followed no matter what the record contained.
The Court feels empowered to decide this case on the basis of an examination of the record because the jury was “free to look beyond the unlawful act presumption and to consider all the evidence on malice.” Ante, at 408. I agree that they were free to do so. Indeed, I believe that they had to do so. (Surely the instruction that something is “rebuttable” conveys to the reasonable jury that they not merely may but must determine whether it has been rebutted.) But what is the problem — what makes it in my view utterly impossible to say beyond a reasonable doubt, from an examination of the record, that the jury in fact found guilt on a proper basis —is that the jury would have been examining the evidence -with the wrong question in mind. Not whether it established malice beyond a reasonable doubt, but whether it was sufficient to overcome (rebut) the improper presumption. Or, to put the point differently, even if a reviewing court can prop*414erly assume that the jury made the ultimate factual determination, it cannot assume that it did so using the appropriate burden of proof. See Carella v. California, 491 U. S. 263, 273 (1989) (Scalia, J., concurring in judgment).
Given the nature of the instruction here, then, to determine from the “entire record” that the error is “harmless” would be to answer a purely hypothetical question, viz., whether, if the jury had been instructed correctly, it would have found that the State proved the existence of malice beyond a reasonable doubt. Such a hypothetical inquiry is inconsistent with the harmless-error standard announced in Chapman v. California, 386 U. S. 18, 24 (1967), and reiterated by the Court today. “[T]he issue under Chapman is whether the jury actually rested its verdict on evidence establishing the presumed fact beyond a reasonable doubt, independently of the presumption.” Ante, at 404 (emphasis added). See also Bollenbach v. United States, 326 U. S. 607, 614 (1946) (“[T]he question is not whether guilt may be spelt out of a record, but whether guilt has been found by a jury according to the procedure and standards appropriate for criminal trials”). While such a hypothetical inquiry ensures that the State has, in fact, proved malice beyond a reasonable doubt, it does not ensure that .if has proved that element beyond a reasonable doubt to the satisfaction of a jury.
* * *
For the foregoing reasons, I join all except footnote 6 and Part III of the Court’s opinion and concur in the judgment of the Court.