dissenting.
Without resolving definitively whether the federal issue was presented adequately, the majority determines that the district court erred in granting habeas relief to the petitioner because the constitutional error committed in the rendition of the instructions was harmless. I agree with my colleagues that the state has waived the argument that the federal issue was not fairly presented to the state courts. Therefore, the district court appropriately addressed the merits of the petitioner’s constitutional claim. However, because I do not believe that the instructional error can be deemed harmless, I cannot join my colleagues in reversing the judgment of the district court. In my view, the conclusion that the error is harmless is based on an impermissible substitution of their judgment on a factual matter for that of the state court jury. I do not believe that it is the proper role for a federal habeas court to intrude so drastically into the prerogative of the jury.
At the outset, it must be stressed that the analysis that follows presupposes the continued vitality of Falconer v. Lane, 905 F.2d 1129 (7th Cir.1990), in this circuit despite the Supreme Court’s critique of that decision in Gilmore v. Taylor, — U.S. —, 113 S.Ct. 2112, 124 L.Ed.2d 306 (1993). Until that matter is raised and briefed in plenary fashion, Falconer remains the law of the circuit and we must proceed accordingly.
In Falconer, a panel of this court held that the Illinois pattern murder instructions, earlier invalidated on state law grounds by the Illinois Supreme Court in People v. Reddick, 123 Ill.2d 184, 122 Ill.Dec. 1, 526 N.E.2d 141 (1988), also violated the Due Process Clause of the Fourteenth Amendment. The federal infirmity identified by this court in Falconer therefore was different from the state ground relied upon by the state court in Reddick. In Reddick, the Supreme Court of Illinois had taken the view that, as a matter of state law, the instructions should have placed on the government the burden of disproving beyond a reasonable doubt a mitigating mental state. By contrast, this court acknowledged that, as a matter of federal constitutional law, the burden of proof with respect to an affirmative defense may be placed on either party.1 It held, however, that the pattern instructions were infirm because, although the murder instructions preceded the voluntary manslaughter instructions, they did not explicitly tell the jury that it could not return a murder verdict if it found that the defendant possessed a mitigating mental state. It was possible, concluded the court, for a jury to find that a defendant was guilty of murder without ever considering whether he was entitled to the voluntary manslaughter conviction. Explicit misdirection on this scale, concluded the court, violates the Due Process Clause. In reaching this conclusion, the court relied principally on the Supreme Court’s holding in Cupp v. Naughten, 414 U.S. 141, 94 S.Ct. 396, 38 L.Ed.2d 368 (1973). In that case, the Court had held that federal courts may not overturn a state conviction on the ground that the jury instructions were erroneous unless those instructions can be said to have infected the entire trial. Id. at 147, 94 S.Ct. at 400.
The issue before us today is whether the error identified in Falconer can be considered harmless. If it can be so considered, we must determine the applicable standard in making such a determination. In Brecht v. Abrahamson, — U.S. —, —, 113 S.Ct. 1710, 1717, 123 L.Ed.2d 353 (1993), the Supreme Court, through the pen of the Chief Justice, held that “trial error” ought to be evaluated on habeas review under the standard enunciated earlier in Kotteakos v. United States, 328 U.S. 750, 66 S.Ct. 1239, 90 *1394L.Ed. 1557 (1946). “Trial error,” the Chief Justice wrote, “ ‘occur[s] during the presentation of the case to the jury.’ ” Id. — U.S. at —, 113 S.Ct. at 1717 (quoting Arizona v. Fulminante, 499 U.S. 279, 307, 111 S.Ct. 1246, 1263, 113 L.Ed.2d 302 (1991)). It is susceptible to harmless error analysis because it may be quantitatively assessed in the context of the other evidence that is presented at trial. Id. As the Supreme Court set forth in Brecht, under this standard, a reviewing court must determine whether the error “‘had substantial and injurious effect or influence in determining the jury’s verdict.’ ” Brecht, — U.S. at —, 113 S.Ct. at 1722 (quoting Kotteakos, 328 U.S. at 776, 66 S.Ct. at 1253). “At the other end of the spectrum,” continued the Chief Justice, are structural defects in the trial mechanism that “infect the entire trial process,” id. — U.S. at —, 113 S.Ct. at 1717, and therefore require automatic reversal. “Trial error” usually involves the admissibility of evidence or the propriety of the argument of counsel. Here, however, we deal with another form of error that arises in the course of trial— instructional error. It is well established at this point that instructional error must be assessed quite differently from other errors that arise in the course of trial. Some are “structural” in nature and not at all subject to harmless error analysis. See Sullivan v. Louisiana, — U.S. —, —, 113 S.Ct. 2078, 2082, 124 L.Ed.2d 182 (1993) (holding that a constitutionally deficient reasonable-doubt instruction cannot be harmless error). On the other hand, other instructions that misstate the task of the jury in assessing the evidence before it are subject to harmless error analysis. See Carella v. California, 491 U.S. 263, 266-67, 109 S.Ct. 2419, 2421-22, 105 L.Ed.2d 218 (1989) (holding that an instruction that a rental car kept 5 days past the rental agreement “shall be presumed to have been embezzled” impermissibly shifts the burden of proof, but is subject to harmless error analysis); Rose v. Clark, 478 U.S. 570, 582, 106 S.Ct. 3101, 3108, 92 L.Ed.2d 460 (1986) (holding that an instruction that im-permissibly shifts the burden of proof on the issue of malice in a murder prosecution is subject to harmless error analysis); Sandstrom v. Montana, 442 U.S. 510, 526-27, 99 S.Ct. 2450, 2460-61, 61 L.Ed.2d 39 (1979) (holding that constitutionally erroneous instruction establishing conclusive presumption that perpetrator intends the ordinary consequences of voluntary acts is subject to harmless error analysis). These eases make clear that, as the majority holds, Falconer error is subject to harmless error analysis. However, as I shall detail in the paragraphs that follow, application of the harmless error rule in these cases poses problems, both conceptual and practical, not faced when we deal with other types of error that arise in the course of trial.
As in the case of instructions that establish mandatory presumptions or instructions that shift the burden of proof, it is indeed difficult to assess the effect of an instruction such as that at issue in Falconer and the present case that explicitly skews the jury’s decision-making process so that it might not even consider the mitigating circumstances that would result in acquittal of the principal charge and conviction only on the lesser included offense. When the traditional formulation of the harmless error test of Kotteakos is applied uncritically to instructional error of the type presented by Falconer, the contours of harmless error analysis are radically expanded. Federal habeas courts consequently are placed in the position of supplying missing factual findings of the jury and, indeed, of relying on evidence to uphold the conviction that the jury may not have considered. See Libby v. Duval, 19 F.3d 733, 741 (1st Cir.), cert. denied, — U.S. —, 115 S.Ct. 314, — L.Ed.2d — (1994) (Stahl, J., dissenting).
In the case of the usual forms of trial-type error such as erroneously admitted evidence or improper argument to the jury, the sort of quantitative assessment contemplated by Brecht is easily accomplished by the reviewing court. The court has before it the entire record and can easily determine whether the fact established by the erroneously admitted evidence was nevertheless established to an overwhelming degree by other lawfully admitted evidence; a judgment therefore can be made as to whether the erroneously admitted evidence had a substantial and erroneous influence on the jury’s verdict. Such an approach is far more difficult when the *1395appellate court is called upon not to assess the effect of information that the jury had before it but to assess the effect of the jury’s not having considered relevant information or not having made a finding which the law requires it to make.
As I have already noted, it is clear from the established precedent that the difficulty in applying the standard Kotteakos approach does not mean that these instructional errors ought not be evaluated under a harmless error analysis. Nor does it mean that the holding of Brecht ought to be inapplicable in such instances. It simply means that an analytical approach, tailored more precisely to the nature of the particular error on the fairness of the proceedings, must be found. As Judge Stahl of the First Circuit has pointed out in his dissenting opinion in Libby, Justice Scalia’s concurring opinion in Carella v. California, 491 U.S. 263, 267-73, 109 S.Ct. 2419, 2421-24, 105 L.Ed.2d 218 (1989), offers a formulation that is of considerable help in this situation.2 Because the inquiry is not whether guilt can be established from the record, but whether guilt was ever found properly by the jury, a reviewing court must determine that the instruction that could have misdirected the jury’s efforts in such a drastic way did not play a role in its verdict. Carella, 491 U.S. at 270, 109 S.Ct. at 2423 (Scalia, J., concurring). Under this approach, instructional error that so grossly misdirects the jury’s inquiry is harmless when it can be established that the facts that the jury necessarily found pursuant to other correct instructions are so closely related to the fact to be proved by the erroneous instruction that a rational jury could not have found the former facts without also finding the fact addressed by the erroneous instruction. In the case of an impermissible presumption, for example, the predicate acts established by correct instructions may so conclusively establish the requisite intent that no rational jury could conclude that the defendant committed the criminal act in question, but did not have the intent that was also the subject of the impermissible presumption. See Carella, 491 U.S. at 272, 109 S.Ct. at 2424 (discussing Rose, 478 U.S. at 579, 106 S.Ct. at 3106). In cases such as the one before us, in which the jury’s inquiry was affirmatively skewed so that the jury might find the defendant guilty of murder without even considering the lesser included offense of manslaughter, the error might also be harmless when the evidence before the court simply did not permit a finding of manslaughter.
While the course of this circuit’s approach to harmless error in the Falconer situation has perhaps not been a seamless garment, our cases, read as a whole, do recognize these principles. As the majority suggests, our cases do contain language that, taken alone, suggests that Falconer error can never be harmless. Notably, however, each of •these cases did explore the possibility that the evidence of record might not reasonably raise the lesser included offense of manslaughter. See Taylor v. Gilmore, 954 F.2d 441, 454 (7th Cir.1992), rev’d on other grounds, — U.S. —, 113 S.Ct. 2112, 124 L.Ed.2d 306 (1993) (noting that the error was “inherently prejudicial,” but also examining the record to determine that the lesser included offense was raised by the evidence); Fleming v. Huch, 924 F.2d 679, 683 (7th Cir.1991) (same); Rose v. Lane, 910 F.2d 400, 403 (7th Cir.), cert. denied, 498 U.S. 983, 111 S.Ct. 515, 112 L.Ed.2d 526 (1990) (same). A later case makes no such reference to the impossibility of harmless error but, notably, approaches the harmless error analysis by asking whether the lesser included offense was reasonably raised by the evidence. See Flowers v. Illinois Dep’t of Corrections, 962 F.2d 703, 706 (7th Cir.1992), vacated on other grounds, — U.S. —, 113 S.Ct. 2954, 125 L.Ed.2d 656 (1993). Our present Chief Judge followed a similar analysis when, writing for the court in United States v. Kerley, 838 F.2d 932 (7th Cir.1988), he held that the district court’s error in not instructing the jury with respect to one element of the offense was harmless because the element was “not contestable.” Id. at 939.
*1396The foregoing approach may well result in a determination of harmless error in substantially fewer instances than in the usual “trial error” situation. However, this difference in result is due to the difference in the problem presented. The Supreme Court has acknowledged that all errors cannot be neatly classified as either “structural” or “trial” errors. In Brecht, the Chief Justice, referring to Justice White’s earlier observation in Fulminante,3 noted that “structural” and “trial” errors were at opposite ends of the “spectrum” of constitutional errors. Brecht, — U.S. at —, 113 S.Ct. at 1717 (emphasis added). Explicit misdirection to the jury on its responsibility to consider the evidence tending to support acquittal on the principal charge and conviction on the lesser included offense is an error very different from the admission of tainted evidence. That it must be treated differently by a reviewing court ought not be surprising.
Justice Scalia’s approach to error of this sort was formulated prior to Brecht. It is clear, however, that his analysis is not dependent upon a particular formulation of the standard of review. The Justice’s opinion is an explanation of the particular dangers presented by instructions that deprive the jury of its fact-finding role — an explanation that makes clear that such an alteration in the jury’s function cannot easily be neutralized because it is far closer to a “structural” error than the typical trial-type error. Certainly, allowing the approach urged by Justice Sca-lia in Carella to survive Brecht is compatible with the principles of judicial restraint and federalism re-emphasized in that opinion. As Judge Stahl points out, fact-finding by federal judges on habeas review is hardly evidence of judicial restraint. Nor is it required by a healthy concept of federalism. Federal courts are to respect the factual findings of the state courts,4 not supplement them.
As the majority quite frankly admits, reb-anee on the harmless error doctrine in this case requires the judges of this court to perform a task that the jury may never have addressed because of the erroneous jury instructions. It requires that the panel resolve matters of credibility and weigh the evidence on the primary issue of guilt or innocence. Mr. Carter has a right to have his guilt or innocence determined by a jury, not by federal appellate judges. Accordingly, I respectfully dissent.
. See Patterson v. New York, 432 U.S. 197, 97 S.Ct. 2319, 53 L.Ed.2d 281 (1977).
. Justice Scalia was addressing in Carella an erroneous instruction that created a conclusive presumption. He noted, however, that his analysis is applicable to other situations where the jury has been deprived of its fact-finding role. Carella, 491 U.S. at 270, 109 S.Ct. at 2423.
.Arizona v. Fulminante, 499 U.S. 279, 290-91, 111 S.Ct. 1246, 1254-55, 113 L.Ed.2d 302 (1991) (White, J., dissenting in part).
. See Sumner v. Mata, 455 U.S. 591, 598, 102 S.Ct. 1303, 1307, 71 L.Ed.2d 480 (1982) (per curiam) (holding that federal courts owe deference to the findings of fact of state courts on habeas review).