dissenting.

This is, so far as I know, the first ease in Anglo-American jurisprudence to hold that a judge erred because he gave a jury instruction that is 100 percent correct. Whatever the merits of this rule — and I agree with Judge Thompson that there is not much— such an extraordinary departure from established law may not be applied on collateral review. See Teague v. Lane, 489 U.S. 288 , 109 S.Ct. 1060 , 103 L.Ed.2d 334 (1989).

Our rule has always been that a trial court may “respond to an inquiry from the jury by rereading relevant instructions that properly state the law applicable to the case.” United States v. Papia, 560 F.2d 827, 843 (9th Cir.1977). The majority somehow concludes— based on “shades of Eddings ” — that principles of reliability in capital sentencing required a new, more detailed supplemental instruction in this case. See maj. op. at 837. But Teague directs us to “survey the legal landscape ... and determine whether a state court considering the defendant’s claim at the time his conviction became final would have felt compelled by existing precedent to conclude that the rule he seeks was required by the Constitution.” Caspari v. Bohlen, 510 U.S. 383, 390 , 114 S.Ct. 948, 953 , 127 L.Ed.2d 236 (1994) (citations and internal quotation marks omitted). 1 McDowell’s conviction became final in 1989. See McDowell v. California, 490 U.S. 1059 , 109 S.Ct. 1972 , 104 L.Ed.2d 441 (1989) (denying petition for cer-tiorari on direct appeal). A glimpse at the legal landscape reveals that never before has any court held that a judge must give a new supplemental instruction rather than refer *844 the jury to the relevant, legally correct instructions already given. Not in 1989; not ever until today. This fact alone may not be dispositive: Every case is different, and every case requires judges to patiently peruse the relevant case law and distill the right answer. If existing precedent compels the result — even if it’s not spelled out in a single case — the state court must apply it. But the state court need not consult a crystal ball to figure out what rabbit the federal courts are next going to pull out of a hat.

The trial judge here could have set out on an eternal quest through the library stacks without discovering the grail which the majority heralds. The line of cases following Bollenbach v. United States, 326 U.S. 607, 612 , 66 S.Ct. 402, 405 , 90 L.Ed. 350 (1946), leads to an unremarkable conclusion: When a trial court responds to a jury’s question, it cannot give an answer that is misleading, nonresponsive or legally incorrect. 2 This tells us nothing about a case where the court gives an instruction that is correct The majority must therefore invoke Eddings v. Oklahoma, 455 U.S. 104 , 102 S.Ct. 869 , 71 L.Ed.2d 1 (1982), which it does: “Bollenbach is not driving this case; Eddings is.” Maj. op. at 841. The argument seems to be that because death is different, the usual assumption — that jurors will read, understand and follow legally correct instructions — is trumped by a new presumption of abiding stupidity.

Does Eddings really stand for this proposition? I can’t find it anywhere in that opinion. The trial judge in Eddings , who was also the sentencer, ruled that he could not consider the defendant’s troubled youth as a mitigating factor. Eddings, 455 U.S. at 109 , 102 S.Ct. at 873-74 . The Supreme Court reversed and held that the sentencer in a capital case must consider “any aspect of a defendant’s character or record and any of the circumstances of the offense that the defendant proffers as a basis for a sentence less than death.” Id. at 110 , 102 S.Ct. at 874 , quoting Lockett v. Ohio, 438 U.S. 586, 604 , 98 S.Ct. 2954, 2964-65 , 57 L.Ed.2d 973 (1978). The Court noted that the judge’s misunderstanding of the law was tantamount to a legally incorrect instruction: “In this instance, it was as if the trial judge had instructed a jury to disregard the mitigating evidence Eddings proffered on his behalf.” Id. at 114, 102 S.Ct. at 877 . Eddings is about prohibiting consideration of mitigating evidence, not curing juror confusion. 3

When McDowell’s jury returned a note expressing confusion about mitigating factors, it may have been in the same position as the trial judge in Eddings . A fair reading of the note suggests the jury may not have *845 been considering the mitigating evidence the defendant had proffered. Did the trial court make the same legal mistake as the judge in Eddings? Did it effectively “instruct [the] jury to disregard the mitigating evidence [the defendant] proffered on his behalf?” Just the opposite. The trial court assiduously applied Eddings by referring the jury to the relevant instructions about mitigating evidence. The majority agrees that these instructions were technically flawless: The instructions directed the jury to consider “[a]ny other circumstance which extenuates the gravity of the crime even though it is not a legal excuse for the crime and any other aspect of the defendant’s character or personal history that the defendant offers as a basis for a sentence less than death.” Compare Eddings, 455 U.S. at 110 , 102 S.Ct. at 874 . Not only were the instructions absolutely correct, they incorporated the precise holding of Eddings .

Eddings said nothing at all about whether legally correct instructions can cure juror confusion about mitigating evidence. My colleagues in the majority implicitly acknowledge this when they summon their new rule from “shades of Eddings.” 4 But Teague works in black and white, not shades, penumbras, nuances or emanations. As the Supreme Court explained in Sawyer: “Even were we to agree with petitioner’s assertion that our decisions in Lockett and Eddings inform, or even control or govern, the analysis of his claim, it does not follow that they compel the rule that petitioner seeks.” Sawyer v. Smith, 497 U.S. 227, 236 , 110 S.Ct. 2822, 2828 , 111 L.Ed.2d 193 (1990) (internal quotation marks omitted). Here, as in Sawyer , no state court “considering the defendant’s claim at the time his conviction became final would have felt compelled by existing precedent to conclude that the rule he seeks was required by the Constitution.” Caspari v. Bohlen, 510 U.S. 383, 390 , 114 S.Ct. 948, 953-54 , 127 L.Ed.2d 236 (1994) (emphasis added) (citation omitted). Because the trial judge did all that Eddings required him to do, the rule the majority announces is new and may not be applied on habeas.

. Neither McDowell nor the majority argues that the rule here falls within either of the two exceptions lo Teague .

. The key passage in Bollenbach, 326 U.S. at 612-13 , 66 S.Ct. at 405 , began by stating the obvious: "When a jury makes explicit its difficulties a trial judge should clear them away with concrete accuracy.” Bollenbach's holding, however, was not so abstract. The Court held that it was error to give a new instruction that was legally incorrect: "In any event, therefore, the trial judge had no business to he 'quite cursory’ in the circumstances in which the jury here asked for supplemental instruction. But he was not even 'cursorily’ accurate. He was simply wrong.” Bollenbach, 326 U.S. at 613 , 66 S.Ct. at 405 .

The instructions in United States v. Gordon, 844 F.2d 1397, 1401 (9th Cir.1988), United States v. Walker, 575 F.2d 209, 213-14 (9th Cir.1978), United States v. Bolden, 514 F.2d 1301, 1308-09 (D.C.Cir.1975), United States v. Petersen, 513 F.2d 1133, 1135-36 (9th Cir.1975) and Powell v. United States, 347 F.2d 156, 157-58 (9th Cir.1965) were all either misleading or nonrespon-sive. See Gordon, 844 F.2d at 1401 (holding that supplemental instruction must respond to jurors’ question); Walker, 575 F.2d at 214 ("The problem is simply that the first supplemental instruction, when viewed in light of the jury’s first question, is confusing!.]”); Bolden, 514 F.2d at 1308 (holding that rereading instruction was nonresponsive because "the problems raised by question one were not resolved by the court’s further charge”); Petersen, 513 F.2d at 1135 (holding that supplemental instruction was "erroneous, misleading and contradictory”); Powell, 347 F.2d at 157-58 (holding that where question has two possible meanings, supplemental instruction which is responsive to one meaning but misleading as to the other is error). The last case on which the majority relies, United States v. Warren, 984 F.2d 325, 330 (9th Cir.1993), was decided after McDowell’s conviction became final and cannot enter the Teague analysis.

. The Supreme Court's later treatment of the case confirms that my colleagues are overreaching. For Teague purposes, the Court has explained, the rule of Eddings is simple: A state may not "impose! ] an absolute prohibition against consideration of certain mitigating evidence by the sentencer.” Sawyer v. Smith, 497 U.S. 227, 236 , 110 S.Ct. 2822, 2828 , 111 L.Ed.2d 193 (1990) (applying Teague).

. The majority notes that Eddings applied and confirmed principles of reliability in capital sentencing and holds that these principles dictate the result McDowell seeks. See maj. op. at 837-838. But the Supreme Court has squarely rejected this approach of framing the Teague inquiry at such a high level of abstraction: “In petitioner’s view, [the rule sought] was dictated by the principle of reliability in capital sentencing. But the test would be meaningless if applied at this level of generality.” Sawyer v. Smith, 497 U.S. 227, 236 , 110 S.Ct. 2822, 2828 , 111 L.Ed.2d 193 (1990).