dissenting:
At the guilt-innocence phase of Drinkard’s capital murder trial, Drinkard presented evidence that he was intoxicated at the time of the murders. Pursuant to § 8.04(a) of the Texas Penal Code, the trial court instructed the jury as follows: “Voluntary intoxication does not constitute a defense to the commission of a crime.” The jury returned a guilty verdict. At the punishment phase of Drin-kard’s trial, Drinkard once again presented evidence that he was intoxicated at the time of the murders. Pursuant to § 8.04(b) of the Texas Penal Code, the trial court instructed the jury as follows: “[I]f you find that the defendant at the time of the commission of the offense for which he is on trial was temporarily insane as a result of intoxication, then you may take such condition into consideration in mitigation of penalty attached for the offense for which the defendant is being tried.” The jurors returned affirmative answers to both special issues submitted to them, and the trial court imposed a sentence of death. Today, the majority holds that there is no reasonable likelihood that Drin-kard’s jury interpreted the § 8.04(b) instruction given at the punishment phase of his trial to foreclose consideration of evidence of intoxication not rising to the level of temporary insanity. I respectfully disagree with the majority’s analysis and conclusions; accordingly, I dissent.
The majority opinion makes three distinct holdings in support of its conclusion that the § 8.04(b) instruction did not violate the Eighth Amendment. First, the majority holds that the plain language of the § 8.04(b) instruction concerns evidence of temporary insanity caused by intoxication, not evidence of intoxication in general. Second, the majority holds that even if the jury could have interpreted the § 8.04(b) instruction, standing alone, to foreclose consideration of lower-level intoxication, the jury could not have done so in light of the trial court’s general instruction to consider all the evidence. Third, the majority holds that even if the jury interpreted the § 8.04(b) instruction to foreclose consideration of lower-level intoxication with regard to the first special issue, concerning deliberateness, the jury could not have done so with regard to the second special issue, concerning future dangerousness. I will address each of these arguments in turn.
*771At the outset, however, I emphasize the legal standard that the Supreme Court has established for such cases: A challenged jury instruction is unconstitutional if “there is a reasonable likelihood that the jury has applied the challenged instruction in a way that prevents the consideration of constitutionally relevant evidence.” Boyde v. California, 494 U.S. 370, 380, 110 S.Ct. 1190, 1198, 108 L.Ed.2d 316 (1990). In order to meet this standard, “a defendant need not establish that the jury was more likely than not to have been impermissibly inhibited by the instruction.” Id.
The majority opinion correctly cites Boyde’s language, but then fails to follow its holding. According to Boyde, there is no constitutionally “correct” interpretation of a challenged instruction. Nor is there a constitutionally “erroneous” interpretation of a challenged instruction. In accordance with Boyde, the only relevant inquiry is whether there is a reasonable likelihood that the jury interpreted the challenged instruction in a constitutionally impermissible way. If so, the instruction is unconstitutional, regardless of whether other, constitutionally permissible interpretations are possible, or even more likely. In the context of the Supreme Court’s holdings in Lockett v. Ohio, 438 U.S. 586, 98 S.Ct. 2954, 57 L.Ed.2d 973 (1978), and Eddings v. Oklahoma, 455 U.S. 104, 102 S.Ct. 869, 71 L.Ed.2d 1 (1982), our inquiry must be whether there is a reasonable likelihood that the jury has applied the challenged instruction in a way that prevents the consideration of any relevant mitigating evidence, including evidence of intoxication falling short of temporary insanity. If we find such reasonable likelihood resulting from the court’s instruction, we must grant habeas relief.
I
The majority first holds that the plain language of the § 8.04(b) instruction, standing alone, does not foreclose consideration of lower-level intoxication. The challenged jury instruction reads as follows:
Evidence of temporary insanity caused by intoxication may be introduced by the defendant in mitigation of the penalty attached to the offense for which he is being tried. Intoxication means disturbance of mental or physical capacity resulting from the introduction of any substance into the body. Temporary insanity caused by intoxication means that the defendant’s mental capacity was so disturbed from the introduction of the substance into the body that the defendant did not know that his conduct was wrong. Therefore, if you find that the defendant at the time of the commission of the offense for which he is on trial was temporarily insane as a result of intoxication, then you may take such condition into consideration in mitigation of penalty attached for the offense for which the defendant is being tried.
(emphasis added). The majority contends that there is no reasonable likelihood that the jury read this instruction to bar its consideration of lower-level intoxication as a mitigating factor. However, the prosecution urged such an exclusive interpretation at trial, and both Texas courts and this Circuit have read the instruction to be just such a bar.
Specifically, the majority holds that there is no reasonable likelihood that the jury in Drinkard’s case interpreted the words “such condition” in the phrase “you may take such condition into consideration” to refer to “intoxication.” Maj. op. at 758. Instead, the majority claims that Drinkard’s jury must have interpreted “such condition” to refer to “temporary insanity caused by intoxication.” Maj. op. at 759. As an initial matter, I find it hard to believe that Drinkard’s jury must have interpreted the referent in question to refer to a phrase that is not even present in the sentence of the instruction at issue.1 *772Even putting grammatical semantics aside, I find it entirely unclear whether the term “such condition” in the instruction refers to “temporarily insane,” to “intoxication,” or to the entire phrase “temporarily insane as a result of intoxication.”
The majority’s analysis stands at odds both with plain language interpretations of the statute from which the instruction was derived and with plain language interpretations of nearly identical instructions given in other cases. The focus of § 8.04 of the Texas Penal Code is on voluntary intoxication in general. Accordingly, the Texas Court of Criminal Appeals has stated explicitly that the statutory language of § 8.04(b) restricts the circumstances under which evidence of intoxication may be considered mitigating evidence. See Cordova v. State, 733 S.W.2d 175, 189 (Tex.Cr.App.1987), cert. denied, 487 U.S. 1240, 108 S.Ct. 2915, 101 L.Ed.2d 946 (1988) (“In Texas, voluntary intoxication is no defense to the commission of a criminal wrong. However, such may become mitigating evidence to the penalty attached to the offense for which the defendant is being tried if the intoxication caused temporary insanity”) (internal citations omitted).
It is therefore not surprising that every published opinion interpreting the plain language of an instruction given pursuant to § 8.04(b), with the exception of that propounded by the majority today, has concluded that the instruction forecloses the jury’s consideration of evidence of intoxication unless such intoxication renders the defendant temporarily insane:
While our penal code specifically precludes voluntary intoxication as a defense to the commission of crime, mitigation of punishment is possible, but only where the level of intoxication produces temporary insanity in the defendant....
Although appellant was not prevented from introducing mitigating evidence, the above instruction required the jury to find her intoxication at the time of the killings rendered her temporarily insane before they could consider her drug use in mitigation of her punishment. The charge on its face instructed the jury to consider the mitigating evidence only in this light, thereby implying that it may not have been considered for any other purpose.
Tucker v. State, 771 S.W.2d 523, 533-34 (Tex.Cr.App.1988), cert. denied, 492 U.S. 912, 109 S.Ct. 3230, 106 L.Ed.2d 578 (1989).
[T]his instruction does not even purport to empower the jury to give mitigating effect to evidence of voluntary intoxication that does not rise to the level of temporary insanity. A juror who believed a capital accused was not so intoxicated as to be incapable of appreciating the wrongfulness of his action might nevertheless find him less morally culpable than would have been a sober man committing the same crime. Here the juror would have no way to effectuate this belief either.
Ex parte Rogers, 819 S.W.2d 533, 537 (Tex.Cr.App.1991) (Clinton, J., dissenting, joined by Baird and Maloney, JJ.).2
We do not reach the merits of the argument that the instruction denied Rogers his constitutionally secured right to have the jury consider all of his relevant mitigating evidence.... The jury was allowed to consider evidence of voluntary intoxication as mitigating if it was persuaded that Rogers was so intoxicated that he did not know that what he was doing was *773wrong.... Here, the jury was allowed to give effect to intoxication evidence but only at the defined level. The instruction’s fit with Johnson and Eddings v. Oklahoma, 455 U.S. 104, 102 S.Ct. 869, 71 L.Ed.2d 1 (1982), is uncertain, and we suggest no answer to that question today.
Rogers v. Scott, 70 F.3d 340, 343-44 (5th Cir.1995), cert. denied, - U.S. -, 116 S.Ct. 1881, 135 L.Ed.2d 176 (1996) (emphasis added).3
My dissent is not based on the operation of the statutory special issues in isolation in Nethery’s case; instead, it is based on another instruction that the trial court submitted along with the special issues that, in effect, took all three of the special issues out of operation with respect to Nethery’s evidence of intoxication....
A reasonable juror could read that instruction as providing that Nethery’s evidence of intoxication could not be considered at all — including under the special issues — unless Netheiy was so intoxicated that he was rendered temporarily insane ....
... Because Nethery’s jury was entirely precluded from considering the evidence of his non-insane state of intoxication, I believe that the § 8.04 instruction given by the trial judge in Nethery’s ease was a straight-forward violation of this well-established Eighth Amendment principle.
Nethery v. Collins, 993 F.2d 1154, 1163-65 (5th Cir.1993), cert. denied, — U.S. -, 114 S.Ct. 1416, 128 L.Ed.2d 87 (1994) (King, J., dissenting).4
Perhaps most troubling about the majority’s reading of the § 8.04(b) instruction in Drinkard’s case is the fact that the State’s brief concedes that instructions given pursuant to § 8.04(b) foreclose jurors’ eonsider-ation of evidence of intoxication not rising to the level of temporary insanity:
Texas law permissibly limits the circumstances under which voluntary intoxication can be given mitigating effect to those instances in which it renders the defendant unable to determine right from wrong or incapable of conforming his conduct to the law....
By requiring that voluntary intoxication result in temporary insanity, as defined by state law, Texas properly restricts the jury’s consideration of mitigating evidence to those circumstances in which the intoxication actually results in a reduced culpability.
Respondent-Appellee’s Opposition to Application for Certificate of Probable Cause, at 24-25 (emphasis added).5
In addition to the assertions of the State in its briefs and at trial, and in addition to the weight of precedent, common sense also dictates that the § 8.04(b) instruction in Drin-kard’s case “clearly directed the sentencer to disregard evidence.” Boyde, 494 U.S. at 384, 110 S.Ct. at 1200. Although the § 8.04(b) instruction in Drinkard’s case did not explicitly prohibit jurors from considering evidence of lower-level intoxication, the Supreme Court has held that an instruction telling a jury what it “may” consider necessarily implies that it may not consider other factors. This truism is embodied in the ancient legal maxim expressio unius est exclusio alterius, the expression of one thing is to the exclusion of another.
The Supreme Court in Hitchcock v. Dug-ger endorsed exactly that inferential step, finding that “it could not be clearer” that, by instructing advisory jurors that they could consider evidence of certain statutory factors, a trial judge instructed them that they could not consider evidence of other, nonstatutory *774factors. Hitchcock, 481 U.S. 393, 398-99, 107 S.Ct. 1821, 1824-25, 95 L.Ed.2d 347 (1987), vacated on other grounds, 505 U.S. 1215, 112 S.Ct. 3020, 120 L.Ed.2d 892 (1992). The rationale of Hitchcock supports an interpretation of the § 8.04(b) instruction in Drin-kard’s case — which affirmatively stated which evidence of intoxication jurors “may” consider — as “clearly directing” jurors not to consider evidence of intoxication not resulting in insanity.
In light of the overwhelming number of § 8.04(b) interpretations — from the parties, from members of this Court, and from Texas state courts — to reach conclusions opposite that of the majority, I do not accept the majority’s unsupported conclusions regarding the plain language of the § 8.04(b) instruction in Drinkard’s case.
II
The majority next holds that the trial court’s general instruction directing the jury to “consider all the evidence” remedied any infirmity in the more specific § 8.04(b) instruction. However, the Supreme Court has held that such a contradictory, permissible instruction in a jury charge will not cure an otherwise constitutionally impermissible instruction:
Nothing in these specific sentences or in the charge as a whole makes clear to the jury that one of these contradictory instructions carries more weight than the other. Language that merely contradicts and does not explain a constitutionally infirm instruction will not suffice to absolve the infirmity. A reviewing court has no way of knowing which of the two irreconcilable instructions the jurors applied in reaching their verdict.
Francis v. Franklin, 471 U.S. 307, 322, 105 S.Ct. 1965, 1975, 85 L.Ed.2d 344 (1985). The majority sidesteps this issue by suggesting that the two instructions are not at odds— that there is no reasonable likelihood that Drinkard’s jury interpreted the trial court’s general instruction and the § 8.04(b) instruction to contradict each other. In light of both common sense and relevant case law, I find such an analysis untenable.
There is more than one way that Drin-kard’s jury could have interpreted the general instruction and the § 8.04(b) instruction to “contradict” each other. The jury could have, of course, interpreted the general instruction to mean “Do consider evidence of lower-level intoxication,” while interpreting the § 8.04(b) instruction to mean “Do not consider evidence of lower-level intoxication.” The jury could have interpreted the general instruction as constituting the general rule and interpreted the § 8.04(b) instruction to carve out a specific exception. Further, the jury could have squared the two instructions through textual analysis. The general instruction directs jurors that they may consider all of the evidence “in determining each of these Special Issues.” Consistent with this instruction, the jury could have considered all of Drinkard’s evidence of intoxication, but only for the purpose of determining whether such evidence rose to the level of temporary insanity. Thus, the jury could have considered such evidence in the process of determining the answers to the special issues, but still could have considered themselves foreclosed from considering evidence of lower-level intoxication in mitigation of punishment, pursuant to the § 8.04(b) instruction. This interpretation renders the instructions facially complementary, though clearly unconstitutional.
I do not proffer any of these interpretations as the “correct” interpretation of the jury charge in Drinkard’s case, nor do I claim that any one interpretation is the most likely. Such claims are not what the law requires. I present these possible interpretations in order to illustrate the uncertainty surrounding the relationship between these two instructions.
Simply put, no language in either the general instruction or the § 8.04(b) special instruction given in Drinkard’s case provides any indication of how the two instructions should relate to each other. This sense of uncertainty was explicitly recognized by another panel of our Court when describing an essentially identical jury charge:
The trial judge did not explicitly instruct the jury whether it could consider the evidence of intoxication in answering the two *775questions. It did instruct that the jury could consider all evidence submitted during both the guilt and punishment phases of the trial, and, significantly, counsel argued the weight the jury ought to accord to the intoxication evidence. Nonetheless, we cannot say with confidence how the jury put the instruction and the questions together. We are describing the uneer-tainty because it is the context in which the procedural bar was invoked.
Rogers, 70 F.3d at 344. We simply do not know how Drinkard’s jury put these instructions together. In the face of such uncertainty, I do not accept the majority opinion’s unsupported assertion that “[t]his general instruction necessarily and undeniably directed the jury to consider Drinkard’s evidence of intoxication in answering the special issues.” Maj. op. at 760.6
Ill
Third, the majority holds that the § 8.04(b) instruction by its own terms applied to only the first special issue, concerning whether the murder was committed deliberately. Therefore, the argument continues, the instruction could not have foreclosed jurors’ consideration of Drinkard’s evidence of intoxication with regard to the second special issue, involving future dangerousness. This portion of the majority’s analysis is flawed in several respects.
The majority’s analysis is exactly the type of “technical hairsplitting” that the Supreme Court has repeatedly warned us not to perform when analyzing challenged instructions under the “reasonable likelihood” standard:
In evaluating the instructions, we do not engage in a technical parsing of this language of the instructions, but instead approach the instructions in the same way that the jury would — with a “eommonsense understanding of the instructions in the light of all that has taken place at trial.”
Johnson v. Texas, 509 U.S. 350, 368, 113 S.Ct. 2658, 2669, 125 L.Ed.2d 290 (1993) (quoting Boyde, 494 U.S. at 381, 110 S.Ct. at 1198).
Jurors do not sit in solitary isolation booths parsing instructions for subtle shades of meaning in the same way that lawyers might. Differences among them in interpretation of instructions may be thrashed out in the deliberative process, with eommonsense understanding of the instructions in the light of all that has taken place at the trial likely to prevail over technical hairsplitting.
Boyde, 494 U.S. at 380-81, 110 S.Ct. at 1198. The majority quotes the language “at the time of the commission of the offense” in one clause of the § 8.04(b) instruction, uses that language to impose a temporal restriction on the whole instruction, draws a distinction between the “backward-looking” first special issue and the “forward-looking” second special issue, and concludes that the jurors must have fenced off the second special issue as a safe haven, a sort of limitation-free zone, for the consideration of evidence of voluntary intoxication. This portion of the majority opinion provides a perfect illustration of a court “parsing instructions for subtle shades of meaning in the same way that lawyers might.”
Even parsing the instructions, I still do not reach the majority’s conclusions. Breaking down the language and grammar of the § 8.04(b) instruction given in Drinkard’s case provides no support for the majority’s conclusion that the instruction affects only the first special issue. The relevant portion of the instruction is a conditional sentence, following an “ifithen” structure:
[I]f you find that the defendant at the time of the commission of the offense for which he is on trial was temporarily insane as a result of intoxication, then you may take such condition into consideration in mitigation of penalty attached for the offense for which the defendant is being tried.
*776The word “if’ signals the condition of the sentence; the word “then” signals the contingency. Both parts of the instruction have temporal components. The condition (“If you find ... ”) is a future condition; it will be realized, if at all, in the jury room. However, this future condition is restricted in time, because the direct object of the future verb “find” is a dependent clause with a past tense verb (“was [temporarily insane]”). Likewise, the contingency (“then you may ...”) is a future contingency; it -will occur, if at all, in the jury room. However, the contingency of the instruction contains no language that restricts its scope to “at the time of the commission of the offense,” or any other past framework. Restating the instruction using symbols, the jury was thus instructed “If you find (in the future) that x occurred (in the past), then you may do y (in the future).” Any restriction on the application of the instruction would have to appear in the contingency (“then you may ... ”), which directs the jury how to apply certain evidence, not in the condition (“If you find ... ”), which only identifies the circumstances under which the contingency will be realized.
When reduced to its basic elements, the majority’s analysis states that language in the “If you find ...” part of the instruction imposes a temporal restriction on the “then you may ...” contingency. Such a thesis is contrary to common sense and unsupported in the language of the instruction. I do not find the words “only with regard to the first special issue” implicit in the language “you may take such condition into consideration in mitigation of penalty attached for the offense for which the defendant is on trial.”
The majority’s position is also directly contrary to the arguments of the State’s attorneys. For if the jury’s findings as to “backward-looking” events were relevant only to the “backward-looking” special issue, jurors could not use past events to predict future behavior. As Mr. Millin argued for the State:
The second issue involves whether or not you find that there’s a probability that Mr. Drinkard will commit future acts of violence, criminal acts of violence, such that they would be or he would be a continuing threat to society. In this regard, as I’m sure you discussed on the voir dire process, that basically the best way — the only way that a person can predict another’s future conduct is based on his past conduct. We have to prove beyond a reasonable doubt that there’s such a probability that this person will act in the future as he’s acted in the past because we would never be able to prove to a 100 percent certainty.
Trial transcript, vol. 36, at 25-26. Therefore, both in terms of grammar, technically parsed in the most legalistic sense, and in terms of common sense, no language in the challenged instruction directs jurors to cabin the effect of the instruction within the first special issue.
The majority is quite right to point out that challenged instructions should be analyzed in the context in which they are made. Cupp v. Naughten, 414 U.S. 141, 146-47, 94 S.Ct. 396, 400, 38 L.Ed.2d 368 (1973). The majority is also correct to consider the arguments of Drinkard’s attorneys as part of that context. Boyde, 494 U.S. at 384-85, 110 S.Ct. at 1200. Both of Drinkard’s attorneys did argue, quite forcefully, that the jury should consider the fact that Drinkard was intoxicated at the time of the murders when deciding both of the special issues. However, the majority’s analysis in this regard is remiss in two respects.
First, while the arguments of counsel are relevant a jury’s interpretation of challenged jury instructions, the court’s instructions themselves carry substantially more weight. Boyde, 494 U.S. at 384-85, 110 S.Ct. at 1200. Therefore, an attorney’s arguments to the jury are simply insufficient to cure an otherwise unconstitutional instruction given by the court. Taylor v. Kentucky, 436 U.S. 478, 488-89, 98 S.Ct. 1930, 1936, 56 L.Ed.2d 468 (1978).
Second, if the majority opinion is to rely on a contextual analysis, it must look at the challenged instruction in the context of “all that has taken place at trial,” Boyde, 494 U.S. at 381, 110 S.Ct. at 1198, not just those parts of the proceedings that support the majority’s conclusions. A review of the trial court record reveals that Drinkard’s intoxi*777cation evidence was a central issue. At the guilt-innocence phase of the trial, Drinkard presented evidence that he was intoxicated at the time of the murders. However, at the close of the guilt-innocence phase, the trial court specially instructed the jury that voluntary intoxication does not constitute a defense to the commission of a crime under Texas law. The message of the § 8.04(a) instruction was clear: Intoxication evidence is simply not relevant.
At the penalty phase of the trial, Drinkard once again presented evidence that he was intoxicated at the time of the murders. At the close of evidence the State waived its right to open closing arguments. Drinkard’s attorneys then argued that the jury could answer both special issues “no” based on the intoxication evidence. They argued, with regard to special issue number one, that Drin-kard did not act deliberately because at the time of the murders he had been intoxicated to the point of temporary insanity; he did not know right from wrong. Then they argued, with regard to special issue number two, that Drinkard would not be dangerous in the future because he was dangerous only when he was drunk, and he would not be able to drink while incarcerated. As support for this argument, Drinkard’s attorneys pointed to evidence concerning his intoxication during violent episodes in his past, including the murders for which Drinkard was on trial. As the majority opinion details, however, these arguments with regard to the second special issue did not focus on intoxication to the point of temporary insanity, but instead focused on intoxication generally, necessarily including evidence of lower-level intoxication.
In the State’s closing argument, Mr. Millin made two direct references to the trial court’s § 8.04(b) instruction. Neither reference limits itself to the first special issue. Indeed, in the portion of Mr. Millin’s argument quoted by the majority in a footnote, the State suggests explicitly that temporary insanity is a prerequisite to the consideration of intoxication evidence under both special issues:
The Defense talks to you about this issue of temporary insanity due to intoxication, and I suppose that comes in mostly — they connected up somehow with both special issues, but to consider that at all — and I suggest after you look at the evidence you won’t consider that at all. To consider that at all you have to decide, one, that at the time of the deaths Mr. Drinkard was intoxicated.... and, two, that by reason of this voluntary intoxication he didn’t basically know right from wrong, he didn’t know what he was doing when he killed these three people was wrong.
Trial transcript, vol. 36, at 22-23 (emphasis added); see also Trial transcript, vol. 36, at 25 (“He wasn’t intoxicated to such an extent he didn’t know right from wrong. That’s what you have to find to give him any kind of break on the intoxication.”) (emphasis added). That is the context in which Drin-kard’s jury heard the trial court’s jury charge. That is the context in which Drin-kard’s jury heard a general instruction to “consider all the evidence submitted to you,” and a special instruction, which concluded:
[I]f you find that the defendant at the time of the commission of the offense for which he is on trial was temporarily insane as a result of intoxication, then you may take such condition into consideration in mitigation of penalty attached for the offense for which the defendant is being tried.
In my opinion, the message of the § 8.04(b) instruction — especially in light of the § 8.04(a) instruction given earlier — is clear: Intoxication evidence is relevant only under the defined circumstances.
The majority today holds that there is no reasonable likelihood that Drinkard’s jury felt precluded by the instructions of the court from considering Drinkard’s proffered evidence of intoxication not rising to the level of temporary insanity. In so doing, the majority concludes that there is no reasonable likelihood that Drinkard’s jury interpreted the § 8.04(b) special instruction as the State’s attorneys interpret it, as Texas courts have interpreted it, and as several members of our Court have previously interpreted it. In the full context of this trial, I find that such misinterpretation was reasonably likely.
*778IV
Finally, because the majority would decide this case on the alternative ground that the recently passed Antiterrorism and Effective Death Penalty Act (“AEDPA”) would deny habeas relief, I briefly address this issue as well. During the pendency of this appeal, the President signed into law the AEDPA which (among other things) amends federal habeas corpus law. This new law narrows the circumstances under which federal courts may grant writs of habeas corpus on behalf of people held under judgment of state courts. The state court’s temporary insanity instruction and subsequent decision so clearly denied Drinkard the constitutional guarantees of Lockett and Eddings, however, that habeas relief is justified even under the AEDPA.
The relevant section of the habeas corpus statute, 28 U.S.C. § 2254(d)(1), as amended by AEDPA § 104(3)(d), states:
(d) An application for a writ of habeas corpus on behalf of a person in custody pursuant to the judgment of a State court shall not be granted with respect to any claim that was adjudicated on the merits in State court proceedings unless the adjudication of the claim—
(1) resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States....
Because Congress included neither an effective date for this amended provision nor a clear statement regarding its retroactive application to cases pending on appeal, it is not apparent whether we should apply the AED-PA in this ease. As an initial matter, I agree with the majority’s careful analysis and conclusions that the statute is a procedural change in the standard of review, and that as such it should have retroactive effect under Landgraf v. USI Film Products, 511 U.S. 244, -, 114 S.Ct. 1483, 1499-1505, 128 L.Ed.2d 229 (1994), and United States v. Mejia, 844 F.2d 209, 211 (5th Cir.1988). I also agree with the majority’s determination that the state court decided Drinkard’s claims on the merits. Maj. op. at 765-67, 767-69. However, as to the majority’s substantive application of the AEDPA and its ultimate decision on the merits of Drinkard’s habeas petition, I respectfully disagree.
A
The majority reviews the state court’s determinations of law separately from mixed questions of law and fact. First, it holds that, as a matter of law, the trial court’s correct identification of the applicable constitutional standard guarantees that the state court’s decision was not contrary to clearly established federal law. Athough the state court apparently recognized that a sentencing judge may not bar a jury from considering any relevant evidence, § 2254(d)(1) directs us to consider a different issue. Under the AEDPA we must consider whether the state court’s adjudication “resulted in a decision contrary ... to clearly established Federal law_” (emphasis added). It is plain that identification of the proper standard is not enough; the state court’s decision must accord with the Supreme Court’s interpretation of the Constitution. For the reasons I have stated above, I think it is clear that the effect of the trial court’s § 8.04(b) instruction was to bar the jury’s consideration of mitigating evidence. Thus the trial court’s decision was contrary to the Supreme Court’s interpretation of the Eighth Amendment in Lockett and Eddings.
B
The majority also holds that, as a mixed question of law and fact, the state court did not unreasonably apply federal law in determining that its limiting instruction did not violate the Eighth Amendment. Specifically, the majority bases its reasoning on the principle that the AEDPA’s “ ‘unreasonable application’ standard of review of a state court decision must mean more than that a federal court may grant habeas relief based on its simple disagreement with the state court decision; this would amount to nothing more than a de novo review.” Maj. op. at 769.
I think the majority has the standard of review exactly wrong. The Supreme Court has consistently held that application of constitutional law to facts in habeas cases re*779quires an independent, de novo determination by federal courts. Wright v. West, 505 U.S. 277, 301-03, 112 S.Ct. 2482, 2495-96, 120 L.Ed.2d 225 (1992) (O’Connor, J., concurring) (the Supreme Court has consistently applied a de novo standard of review in mixed questions of constitutional law and fact in habeas corpus cases); see also Brown v. Allen, 344 U.S. 443, 507, 73 S.Ct. 437, 446, 97 L.Ed. 469 (1953), overruled on other grounds by Townsend v. Sain, 372 U.S. 293, 312, 83 S.Ct. 745, 756-57, 9 L.Ed.2d 770 (1963) (“Thus, so-called mixed questions or the application of constitutional principles to the facts as found leave the duty of adjudication with the federal judge.”); Irvin v. Dowd, 366 U.S. 717, 723-28, 81 S.Ct. 1639, 1643-45, 6 L.Ed.2d 751 (1961) (reviewing de novo state court determinations of mixed questions of law and fact in federal habeas case); Brewer v. Williams, 430 U.S. 387, 403, 97 S.Ct. 1232, 1242, 51 L.Ed.2d 424 (1977) (same); Cuyler v. Sullivan, 446 U.S. 335, 342, 100 S.Ct. 1708, 1715, 64 L.Ed.2d 333 (1980) (same); Miller v. Fenton, 474 U.S. 104, 112, 106 S.Ct. 445, 450, 88 L.Ed.2d 405 (1985) (same). The Supreme Court has made clear that federal courts must undertake independent, de novo review of state court habeas decisions on appeal. I am unwilling to depart from this unbroken line of Supreme Court precedent, especially since the language of § 2254(d)(1), as amended, does not demand it.7
The majority continues by stating that an application of law to facts is unreasonable only where “reasonable jurists would be of one view that the state court ruling was incorrect.” This cannot be the standard of review. Where a federal court of appeals determines that a state criminal decision is contrary to federal law, § 2254(d)(1) does not require the unanimous consent of the federal bench for habeas relief. Indeed, it does not even require unanimity among a panel of judges considering the case. The determination of reasonableness must consider only the propriety and correctness of the state court’s actions in the context of federal guarantees established by the Supreme Court. If a federal court “disagrees” with the state court’s application of federal law — if it finds that the state court unreasonably applied the law of the land — that federal court must grant habeas relief under § 2254(d)(1). It is well established that where state and federal courts disagree about the meaning of federal law, the interpretation of the federal courts must prevail. Brown, 344 U.S. at 507, 73 S.Ct. at 446.
As I have catalogued in this dissent, I think it clear that the state court’s temporary insanity instruction denied Drinkard the constitutional guarantees of Lockett and Ed-dings. The' misapplication of the Eighth Amendment to the facts of this case justify relief under § 2254(d)(1), whether or not we apply the AEDPA. Thus I respectfully disagree with the majority’s conclusions, and, accordingly, I dissent.
. The majority in Ex parte Rogers did not undertake an interpretation of the § 8.04(b) instruction. Unlike Drinkard, the petitioner in Ex parte Rogers did not object to the trial court's jury charge, and the per curiam opinion disposed of all of the petitioner’s claims by finding "that the findings and conclusions entered by the trial court are supported by the record and upon such basis the relief sought is denied." Ex parte Rogers, 819 S.W.2d at 534.
. Unlike Drinkard, the petitioner in Rogers did not object to the § 8.04(b) instruction given at the punishment phase of his trial. We thus held his Eighth Amendment claim procedurally barred. 70 F.3d at 343-44.
. The majority in Nethery expressly declined to address the merits of the constitutionality of the § 8.04(b) instruction, holding the claim procedurally barred. Nethery, 993 F.2d at 1161 n. 26.
.Instead of arguing that the instruction does not foreclose consideration of evidence of intoxication not rising to the level of temporary insanity, the State's brief argues that such evidence is not constitutionally relevant. As the majority acknowledges, evidence of intoxication at the time of the murders is clearly constitutionally relevant. Maj. op. at 758 n. 10.
. More importantly, the Supreme Court disagrees with the majority’s assertion. As I have previously noted, Hitchcock v. Dugger supplies the inferential step that the majority ignores — "it could not be clearer” that, by instructing advisory jurors that they could consider evidence of certain statutory factors, a trial judge instructed them that they could not consider evidence of other, nonstatutory factors. See, supra at 773-74 (citing Hitchcock, 481 U.S. at 398-99, 107 S.Ct. at 1824-25).
. For this unprecedented deferential standard of review, the majority cites only the word "unreasonable" in § 2254(d)(1) and one piece of legislative history, indicating that the AEDPA "requires deference to the determinations of state courts that are neither ‘contrary to,’ nor an ‘unreasonable application of,' clearly established federal law.” H.R.Conf.Rep. No. 518, 104th Cong., 2d Sess. 111 (1996), reprinted in 1996 U.S.C.C.A.N. 944, 944. This statement in the conference report does not change the standard of review, it merely restates the standard of the AEDPA, dictating that we should not upset state court decisions that do not offend federal constitutional law. I would not overturn established Supreme Court precedent with so thin a lever.