concurring in part and dissenting in part, with whom RIPPLE and ROVNER, Circuit Judges, join Part II.
This case underscores both the importance and the difficulty of the task with which courts are charged under the Antiterrorism and Effective Death Penalty Act of 1996, Pub.L. No. 104-132, 110 Stat. 1214, in applying the new substantive standards for the issuance of writs of habeas corpus in proceedings brought by state prisoners under 28 U.S.C. § 2254. Although I concur with most of the majority’s analysis of the new law, I dissent from the way it has been applied here.
*878I
With one reservation, I concur in Part I of the majority’s opinion, which holds that the amended version of § 2254 applies to Lindh’s case. My reservation relates only to the ancillary decision to deny Lindh’s motion to supplement the record in this court. See ante, at 867. Lindh wanted to demonstrate that the record in the Wisconsin court reflected Dr. Leigh Roberts’ suspension of privileges at the University of Wisconsin Hospitals by the time of the trial. This is not an effort to introduce new evidence in any sense of that term; it is only an effort to show us what was legitimately in the record before the Wisconsin courts. In my view, in light of the changes in the law that have occurred and the lack of prejudice to all concerned, we should grant that motion.
I am also in substantial agreement with Parts II.A and II.B of the majority’s opinion, which hold that the amended version of § 2254(d)(1), construed properly, does not unconstitutionally trample on the Article III powers of the federal judiciary. In particular, I agree that nothing in the language of § 2254(d) requires the federal courts to “defer” to the legal determinations of the state courts, and therefore we need not confront the serious constitutional issues that would arise if the contrary were true. I begin to part company with the majority where it foreshadows its discussion in Part II.C of its opinion and suggests that we “need not shoulder the potentially difficult task of determining when an appellate gloss on a decision of the Supreme Court has so far departed from its wellsprings as to be the ‘real’ source of law_” Ante, at 869. For better or for worse, that task will be ours in virtually every case that comes before us under the amended law, because it is rare indeed that we will see something identical in all particulars to a case already decided by the Supreme Court. However minor the factual variations, we both can and must look for guidance in our own decisions, decisions from other appellate courts (federal and state), and persuasive secondary sources.
I also disagree with the majority’s statement, ante at 870, that one ingredient of Lindh’s claim is not well established. As I explain in more detail below, the majority’s description of the Wisconsin scheme is incomplete, and is therefore potentially misleading. Whether the Wisconsin court labeled this phase “dispositional” in its opinion in State v. Koput, 142 Wis.2d 370, 418 N.W.2d 804 (1988), is not controlling as a matter of federal law. These are questions, however, that go to the application of amended § 2254(d), not to its compatibility with Article III.
Finally, in broad terms I agree with the majority’s interpretation of the key components of § 2254(d)(1) and its explanation of the difference between something “contrary to” clearly established law and something that is an “unreasonable application of’ such law. My concern is with a few statements that I fear could be misunderstood. For example, the majority states broadly that the “contrary to” language of § 2254(d) does not authorize issuance of a writ whenever a court errs. If all it means by that is that doctrines like harmless error continue to apply, then the statement is a harmless truism. If it were read to imply that prejudicial errors of law can be ignored, however, it would be wrong. The statute does authorize issuance of a writ, if the state court has committed a pure error of law. This is because the federal court has the power to decide whether the state court followed clearly established law as declared by the Supreme Court of the United States and because no deference is owed to the state court’s own views of that issue. The real problem is a familiar one (though no easier for being familiar): when is the issue purely one of law, and when is it a mixed question of law and fact. If the federal court concludes (de novo) that the state court has not made an error of law, then the question becomes whether there has been an unreasonable application of the law. This, I agree, is another way of describing the “mixed question of law and fact” situation. Unreasonableness makes sense as a standard here, if for no other reason than the deference that is owed to the state court’s determination of the facts, see § 2254(e). The majority suggests that when the constitutional question is a matter of degree, rather than of concrete entitlements, the federal court *879must honor a “reasonable” decision by the state court. Most constitutional entitlements in the area of criminal procedure have elements of both: if a defendant’s rights under Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963), have been violated with respect to part of the prosecutor’s file, it is little comfort that the prosecutor turned over other Brady material; the same goes for Confrontation Clause rights of the sort we have here. Just because one can cross-examine a witness about prior inconsistent statements, for example, does not mean that cross-examination is unnecessary to show that the same witness had a plea bargain with the government. Many of these rights, in short, will not be matters of degree when we address them at the proper level of detail: the deprivation will be clear, and the federal court must decide whether the right was violated.
II
In order to explain why I disagree with the majority’s resolution of Lindh’s own case under the principles set forth in its opinion, it is necessary to supplement the statement of facts and the description of the procedures used in Wisconsin cases where a plea of not guilty by reason of mental disease or defect is entered.
Lindh went on his shooting rampage in the City-County Building in Madison, Wisconsin, on January 15, 1988. He himself was shot during the melee, was apprehended immediately, and was taken to the hospital under police custody. At 9:00 p.m. that evening, the district attorney’s office contacted Dr. Roberts at his home and asked him to interview Lindh at the intensive care unit. Roberts agreed and promptly went to the hospital, arriving around 9:30 p.m. Lindh spoke with Roberts for some time, telling him about the circumstances preceding the murders and his feelings. Roberts elicited statements from Lindh indicating that he was not experiencing suicidal feelings, depression, hallucinations, or the like.
Lindh was arraigned in Dane County Circuit Court on February 26,1988, and entered pleas of not guilty and not guilty by reason of mental disease. At that time, the court appointed Roberts and Dr. Frederick Fosdal, another forensic psychiatrist, to examine Lindh on behalf of the state. Shortly thereafter, Roberts learned that the University of Wisconsin Hospitals were investigating allegations that he had engaged in sexual misconduct with a female patient — allegations that he knew could lead to loss of his license and criminal penalties. In May 1988, the Medical Examining Board asked Roberts to provide it with records, revealing to Roberts that the Board, too, was concerned about these allegations.
In the meantime, Roberts continued to work on the Lindh case. On June 22, 1988, he interviewed Lindh again. The very next day, June 23,1988, attorneys for the University of Wisconsin Hospitals referred the sexual misconduct allegations to the Dane County district attorney’s office. Recognizing the potential for a conflict of interest, that office referred the case to a special prosecutor from the Milwaukee district attorney’s office. Roberts learned of that appointment and of the criminal investigation on July 8,1988. It was not until August 17, 1988, that Roberts first produced a written report on the case, which concluded that Lindh was not suffering from a mental disease or defect at the time of the shootings. On August 22,1988, he interviewed Lindh a final time. Lindh alleges, as noted above, that the state trial court record shows that Roberts was suspended from the University Hospitals some time during the summer. Whatever the status of that allegation and his unsuccessful motion to supplement the record, it is clear that the state Medical Examining Board filed a civil complaint against him in August to take away his license, based on alleged sexual exploitation of three patients. Crucially, all these events transpired before the trial in Lindh’s case.
On September 12, 1988, the state filed a motion in limine requesting that the trial court prohibit any cross-examination of Roberts concerning the allegations of misconduct pending against him or concerning any ramifications of those pending charges. Lindh opposed the motion, arguing that the circumstances raised serious questions about Roberts’ bias, motive, and interest. First, Roberts would present himself as a distinguished *880professional, and the jury might give less weight to his opinion if it knew that proceedings to revoke his license were pending. Second, defense counsel noted that University Hospitals had already suspended Roberts’ privileges, which was important for the jury to know. Finally, Lindh argued that Roberts would have an incentive to please the prosecution, regardless of the technical fact that the particular prosecuting attorney was from Milwaukee rather than Madison. The case had received state-wide notoriety, and Roberts himself was a well-known expert witness. Unmoved by Lindh’s arguments, the trial court granted the state’s motion and found that evidence relating to the investigation of Roberts’ misconduct was “totally irrelevant and immaterial.”
When Roberts took the stand to offer his testimony, Lindh’s worst fears were realized. During the initial presentation of his credentials, the jury learned that Roberts was a faculty member of the University of Wisconsin Medical School, that he was “on the adjunct faculty of the San Francisco Theological Seminary,” that he had chaired conferences “in relation to religion and mental health,” that about three years before he had been “honored ... as essentially the mid-west psychiatrist of the year,” and, on top of all that, that he was a loving grandfather with eight grandchildren “up to nearly the age of Mr. Lindh.” Because of the trial court’s ruling, the jury never learned that Roberts’ affiliation with the University of Wisconsin was in peril, nor that he was about to lose the prestigious standing in the profession he had enjoyed up to that time. Worst of all, the jury never learned that Roberts might have had a motive to favor the prosecution in his testimony.
As Wisconsin law requires in cases like Lindh’s, there was a “continuous trial” in which the jury first considered the plea of not guilty, and, having concluded that Lindh was guilty, then considered his plea of not guilty by reason of mental disease or defect. See Wis. Stat. Ann. § 971.165(l)(a). The jury’s one-sided picture of Roberts was particularly troubling because Roberts was the prosecution’s star witness during the second (mental disease) stage of the proceedings. Emphasizing his superior ability to diagnose Lindh due to the contemporaneous interview he had conducted on the date of the shooting, Roberts testified that Lindh was not suffering from any psychosis at the time of the shootings and was fully capable of understanding the wrongfulness of his actions and conforming his behavior to the requirements of the law. The defense called Dr. Ezra Griffith, who testified that Lindh suffered from a mixed personality disorder and was in a brief reactive psychosis during the shootings. According to Griffith, Lindh was unable to appreciate the wrongfulness of his actions and conform his conduct to the law when he was in such a state. Finally, the other prosecution expert, Dr. Fosdal, offered an opinion somewhere in the middle: he agreed with Griffith that Lindh was suffering from a mixed personality disorder, but he disagreed that Lindh had experienced a brief reactive psychosis during the shootings. With this evidence before it, the jury found that Lindh was able to appreciate the wrongfulness of his conduct and to conform it to the law. See Wis. Stat. Ann. § 971.15(1). In accordance with Wisconsin law, the judge therefore proceeded to the sentencing phase of the proceeding, see Wis. Stat. Ann. § 971.165(3)(a), and imposed a sentence of life in prison.
Wisconsin uses a three-phase proceeding-in cases in which the mental responsibility of the defendant is at issue. A defendant is entitled to couple a plea of not guilty with a plea of not guilty by reason of mental disease or defect, pursuant to Wis. Stat. Ann. § 971.165(1). The statute establishes the order of proceedings for the two pleas as follows:
(a) There shall be a separation of the issues with a sequential order of proof in a continuous trial. The plea of not guilty shall be determined first and the plea of not guilty by reason of mental disease or defect shall be determined second.
Section 971.15(3), Wis. Stat. Ann., provides that “[mjental disease or defect excluding responsibility is an affirmative defense which the defendant must establish to a reasonable degree of certainty by the greater weight of the credible evidence.”
*881If the jury finds the defendant not guilty on the first plea, the court must enter a judgment of acquittal and discharge the defendant. § 971.165(l)(d). If, on the other hand, the jury finds the defendant guilty on the first plea, then the trial continues into phase two and the court withholds judgment pending the jury’s determination on the second plea. Id. At the end of the second phase, the jury either finds that the defendant is not guilty by reason of mental disease or defect, or that the defendant is guilty (ie. his affirmative defense fails). § 971.165(B). At that point, the court either enters a judgment of “not guilty by reason of mental disease or defect,” or it enters a judgment of conviction. Id. After the appropriate judgment is entered, the court concludes the proceedings with phase three, which is either a commitment hearing pursuant to § 971.17(2)(a) or a sentencing proceeding pursuant to §§ 972.13, 972.14, and 972.15. For those found not guilty by reason of mental disease or defect, the statute requires the court to commit the person to the department of health and social services for a specified period not exceeding two-thirds of the maximum term of imprisonment that could have been imposed, unless the maximum term of imprisonment is life, in which case the commitment period can also be life. § 971.17. As soon as six months after the person is committed, he is entitled to petition for conditional release. § 971.17(4)(a). He can continue filing these petitions once every six months. Importantly, the court “shall grant the petition unless it finds by clear and convincing evidence that the person would pose a significant risk of bodily harm to himself or herself or to others or of serious property damage if conditionally released.” § 971.17(4)(e). A person who wins conditional release can then petition the court for termination of the order of commitment. § 971.17(5).
The Wisconsin Supreme Court discussed the predecessor statute to § 971.165 extensively in State v. Koput, supra, in which the principal issue was whether the jury verdict in phase two must be unanimous. It distinguished the responsibility phase of the bifurcated trial used in Wisconsin from the guilt/innocence phase, labeling it “a special proceeding in the criminal process in which the defendant has the burden of proof to establish his lack of responsibility to a reasonable certainty by the greater weight of the credible evidence.” 418 N.W.2d at 805. It rejected the public defender’s arguments that the usual protections for a criminal trial, such as the need for a unanimous verdict by a jury of twelve, and the requirement of proof beyond a reasonable doubt, applied in the phase two proceeding. The affirmative defense tried during phase two related only to “responsibility,” and had the effect of relieving the person of the sanctions for criminal conduct. The finding of guilt during phase one, however, is unaffected by phase two. For that reason, the Wisconsin Court concluded that phase two was “dispositional” in nature and that the jury rules for phase one were inapplicable. Id. at 812.
In its opinion in Lindh’s ease, the Wisconsin Supreme Court indicated that the Confrontation Clause of the Sixth Amendment, as well as its analog in Article I, section 7 of the Wisconsin Constitution, applies to the phase two proceeding. See State v. Lindh, 161 Wis.2d 324, 468 N.W.2d 168, 175 (1991). On one level, that is almost enough for us, because it tells us the significance Wisconsin attaches to this stage of the proceedings. But as an independent matter of federal law, the Wisconsin Supreme Court’s conclusion is plainly correct. Clearly established law from the Supreme Court shows that nothing about the phase two proceeding would justify an exemption from the Confrontation Clause.
First, Wisconsin’s practice of determining criminal responsibility in the second stage of a bifurcated trial, where the defendant has the burden of proof, is no different from many other state rules that the Supreme Court has approved that similarly shift the burden of proof. See, e.g., Leland v. Oregon, 343 U.S. 790, 72 S.Ct. 1002, 96 L.Ed. 1302 (1952) (state may require accused to prove his insanity beyond a reasonable doubt); Patterson v. New York, 432 U.S. 197, 97 S.Ct. 2319, 53 L.Ed.2d 281 (1977) (constitutional to require defendant to prove affirmative defense of extreme emotional disturbance in order to qualify for mitigating factor that reduced a charge to manslaughter); McMil*882lan v. Pennsylvania, 477 U.S. 79, 106 S.Ct. 2411, 91 L.Ed.2d 67 (1986) (constitutional to use “preponderance of the evidence” standard for the State’s burden of proof at sentencing for minimum sentencing law); Martin v. Ohio, 480 U.S. 228, 107 S.Ct. 1098, 94 L.Ed.2d 267 (1987) (constitutional for Ohio to require defendant to prove affirmative defense of self-defense by a preponderance, as long as instructions required prosecution to prove all elements of the offense of aggravated murder beyond a reasonable doubt). Thus, the Wisconsin Supreme Court was on firm ground in Koput when it found nothing objectionable about the use of a non-unanimous jury and a shift in the burden of proof for Wisconsin’s criminal responsibility determination.
Second, no matter how for purposes of state law Wisconsin wishes to characterize phase two, it is clear that it is the definitive step that yields either a finding of criminal responsibility for the offense, and hence incarceration, or a finding of no criminal responsibility, and hence hospitalization for as little as six months or as much as a lifetime. The majority disregards the provisions of Wis. Stat. Ann. § 971.17 when it asserts, ante at 875, that the only consequence of a decision that Lindh was not criminally responsible would be to “change the place of his confinement.” The majority dismisses the possibility of release under § 971.17 as trivial, equating it to the chance of release under an indeterminate sentencing scheme. This position, however, overlooks both the critical features of the second phase of the Wisconsin proceeding and the defining characteristics of any sentencing proceeding.
It is undisputed that the second phase trial begins with a plea of “not guilty by reason of mental disease” and ends with a jury verdict of guilty or not guilty. Like all jury verdicts of guilty or not guilty, that is “dispositive” in the sense that it governs what kind of further proceedings will be required, but it is not “dispositive” in the same way a sentencing proceeding is. At the risk of stating the obvious, the purpose of a sentencing hearing is to pronounce the sentence, and the jury’s phase two verdict does not even begin to address that subject. A sentence might be a fine or incarceration, or some combination of the two. It might include a period of probation or supervised release; construing the term broadly, it might even include commitment to a mental institution for a period of years. These issues are simply not before the court or the jury during the phase two trial in Wisconsin. Its sole purpose is instead to decide whether the person is criminally responsible, and hence will receive a criminal sentence, or if the person is not to be held responsible and instead is to receive involuntary commitment to a mental institution. (The latter, of course, occurs every day in contexts far removed from the criminal justice system.) The fact that the verdict of “guilty” from phase one stands does not distinguish this from many other situations— such as self-defense, heat of passion, or justification — where it is clear that the defendant committed the criminal act, but he will not be punished for it. Furthermore, the analogy to parole or release under an indeterminate sentencing scheme does not hold up. No such system of which I am aware confers an entitlement on the defendant to win release unless the court finds by clear and convincing evidence that he must remain incarcerated. Yet this is exactly how the commitment rules in Wisconsin operate for individuals found not to be criminally responsible for their acts. See Wis. Stat. Ann. § 971.17(4)(d), (5).
The Supreme Court has always required application of the Confrontation Clause to proceedings that begin with a plea of not guilty and conclude with a finding of guilt or acquittal. Cruz v. New York, 481 U.S. 186, 190, 107 S.Ct. 1714, 1717, 95 L.Ed.2d 162 (1987) (for purposes of the Confrontation Clause, a witness against a defendant is anyone whose testimony “is part of the body of evidence that the jury may consider in assessing his guilt.”); Pointer v. Texas, 380 U.S. 400, 403, 85 S.Ct. 1065, 1067-68, 13 L.Ed.2d 923 (1965) (holding the right of confrontation applicable to the states under the Fourteenth Amendment). More than that, it has applied the Confrontation Clause to capital sentencing proceedings, where it is true that the only difference is in the nature of the sentence imposed (albeit an exceptionally important difference), explicitly recognizing *883the need for adversarial debate in such a context. See generally Gardner v. Florida, 430 U.S. 349, 360, 97 S.Ct. 1197, 1206, 51 L.Ed.2d 393 (1977) (“Our belief that debate between adversaries is often essential to the truth-seeking function of trials requires us also to recognize the importance of giving counsel an opportunity to comment on facts which may influence the sentencing decision in capital cases”). The plain fact here is that Lindh could, in the words of the Wisconsin Supreme Court, have been “relieve[d] ... of the sanctions for criminal conduct” at the end of the phase two proceeding. See Koput, 418 N.W.2d at 812. It is no expansion of clearly established Supreme Court law to find here that stakes of this magnitude in the proceeding bring the Confrontation Clause into play, as the Wisconsin Supreme Court itself recognized.
Decisions of the Supreme Court of the United States also establish that Lindh’s right to confront Roberts was impermissibly restricted, through the trial court’s grant of the motion in limine. The two key cases, the majority agrees, are Davis v. Alaska, 415 U.S. 308, 94 S.Ct. 1105, 39 L.Ed.2d 347 (1974), and Delaware v. Van Arsdall, 475 U.S. 673, 106 S.Ct. 1431, 89 L.Ed.2d 674 (1986). In the 1974 Davis decision, the Supreme Court began with the observation that “[cjonfrontation means more than being allowed to confront the witness physically.” 415 U.S. at 315, 94 S.Ct. at 1110. The main and essential purpose of confrontation is “to secure for the opponent the opportunity of cross-examination.” Id. at 315-16, 94 S.Ct. at 1110 (emphasis in original). Cross-examination involves both the testing of the witness’s story and impeachment or discrediting of the witness. As the Davis Court said, “[a] more particular attack on the witness’ credibility is effected by means of cross-examination directed toward revealing possible biases, prejudices, or ulterior motives of the witness as they may relate directly to issues or personalities in the case at hand.” Id. at 316, 94 S.Ct. at 1110. The Court held that counsel should have had the opportunity not only to ask the witness in question whether he was biased, but also to explore why he might have been biased or otherwise lacked impartiality.
The Court reiterated all these points in Van Arsdall. Notably for Lindh’s ease, it drew a distinction between trial court control of cross-examination to limit harassment, confusion, repetition, or the marginally relevant, and the prohibition of all inquiry into the possibility of a witness’s bias. 475 U.S. at 679, 106 S.Ct. at 1435. It held that:
a criminal defendant states a violation of the Confrontation Clause by showing that he was prohibited from engaging in otherwise appropriate cross-examination designed to show a prototypical form of bias on the part of the witness, and thereby “to expose to the jury the facts from which jurors ... could appropriately draw inferences relating to the reliability of the witness.” 475 U.S. at 680, 106 S.Ct. at 1436, quoting from Davis v. Alaska
The particular risk of bias in Van Arsdall was quite similar to the type Lindh wished to put before the jury. The trial court in Van Arsdall prohibited inquiry into the terms of an agreement the witness had made with the prosecutor to drop a drunk driving charge in exchange for his testimony about the murder. There was nothing unusual or unprecedented in Lindh’s effort to show that Roberts, too, might have been shading his testimony with the ulterior motive of securing better treatment from the state. Thus, Supreme Court decisions establish both that the Confrontation Clause apples to the kind of proceeding we have here, and that the Confrontation Clause is violated when the defendant is precluded from exposing “possible biases, prejudices, or ulterior motives.” Davis, 415 U.S. at 316, 94 S.Ct. at 1110.
The only question remaining is whether the Confrontation Clause was violated here, when the trial court precluded all inquiry into the proceedings at which Roberts risked loss of his professional license, obloquy from his peers, and criminal conviction. Both the Wisconsin Supreme Court and the majority here believe that the administrative transfer of the case from the Dane County prosecutor’s office to the Milwaukee special prosecutor erased all possibility of concern from Roberts’ mind. In my view, this asked the wrong question (i.e. was it sufficient when *884steps were taken to assure formal insulation of the Milwaukee prosecutor from the Dane County office), and thus produced the wrong answer. The focus must instead be on Roberts’ motivations to shade his testimony, just as it was in Van Arsdall. It is worth recalling what had already happened to Roberts, before deciding whether the trial court erred in its total exclusion of the impeachment testimony. The two critical dates are July 8, 1988, and August 17, 1988. No later than July 8, 1988, Roberts knew that a criminal investigation of his activities was underway, with all that implied for his career. Yet it was not until August 17, 1988, as far as this record shows, that he first reduced his conclusions about Lindh to writing. The majority comments that there is no reason to believe Roberts changed his mind between the initial interview on the date of the shooting and August 17, but there is obviously no way to know if the trial court forbade inquiry and cross-examination on the matter. (The majority suggests that Roberts was open to “unfettered cross-examination” on this issue, but that cannot be true given the fact that Lindh’s counsel was forbidden to question Roberts on the one fact that might have led Roberts to change his mind: his legal problems, and his hope of winning favorable treatment from the prosecutors.) By the time the trial started, the state Medical Examining Board had also filed a civil complaint against Roberts to take away his license. Again, the jury knew nothing of this; instead, it was informed that he was “mid-west psychiatrist of the year.”
Cross-examination on Roberts’ legal difficulties was required for at least two reasons. First, when expert witnesses are proffered the jury is entitled to learn about their qualifications. See Daubert v. Merrell Dow Pharmaceuticals, 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993). Particularly in a field like psychiatry, where methodologies are not readily subject to the kind of objective scientific verification the Supreme Court called for in Daubert, the jury is entitled to evidence that will suggest what weight it should give to the expert’s opinions. It is fanciful to assume that a jury would give the same weight to the “mid-west psychiatrist of the year” as it would give to someone who was facing both criminal charges and civil proceedings to revoke his license. Second, as Justice Abrahamson of the Wisconsin Supreme Court so ably demonstrated in her dissenting opinion, Roberts subjectively may have believed or hoped that a “good performance” on his part for the state might cause the prosecutor to treat him leniently. The fact that the prosecutor was called a “special prosecutor” for Dane County, rather than a member of the usual staff, cannot have made any difference from Roberts’ subjective viewpoint. The courts regularly see witnesses in criminal trials who face charges in multiple jurisdictions (federal and state, or several states or parts of states), and the witnesses are aware that their performance is being watched by all concerned. If a witness were testifying in a case before the Circuit Court of DuPage County, Illinois, and that witness were facing charges in Cook County, the State’s Attorney would have no ground for keeping out cross-examination designed to show hope of favorable treatment at the hands of the Cook County authorities. Roberts’ case is no different.
Indeed, the ease is even stronger for finding that Roberts may have entertained such a hope. Not all criminal prosecutions — not even all murder prosecutions — receive statewide publicity, but the Lindh case did. Roberts, who had testified by his own account all over the state, knew that he was known in Milwaukee, just as he was known in other parts of Wisconsin. He had every reason to begin his efforts for leniency on the witness stand in Lindh’s case. (Although it is not relevant to the trial court’s ruling on the motion in limine here, the record of the actual criminal sentencing that took place after Roberts was convicted on the charges shows that his expectations were well founded. Both the prosecutor and the judge expressed intense regret at having to sentence him to anything at all, given his outstanding service to the state over many years.)
Van Arsdall and Davis make clear that the trial court did not face an “all or nothing” option with respect to the proffered impeachment testimony. It was entitled to place reasonable limits on it, to avoid having the jury distracted with Roberts’ problems. The *885trial court could, for example, have permitted cross-examination about the existence of the pending charges and forbidden discussion of their basis, instructing the jury that the nature of the professional misconduct charges was not relevant. This, or other possible limitations, would have eliminated or minimized any risk of a salacious exploration of Roberts’ sexual proclivities. The problem here, as in Van Arsdatt, lay in the total exclusion of the evidence. It is impossible, given the absence of scientific validation of this kind of psychiatric diagnosis, to say that this error was harmless. The jury literally had to decide whose word it would accept: Roberts’, Fosdal’s, or Griffith’s. We must recall that only Roberts testified both that Lindh had no personality disorder and that he was responsible at the time of the offense. The jury would have seen an entirely different picture if Roberts had been discredited, and it was left only with Fosdal, who agreed that Lindh had a personality disorder but who thought he was responsible at the time, and Griffith, who thought Lindh had the disorder and was not responsible for what he did. Lindh, recall, had only to persuade the jury by a preponderance of the evidence that he was suffering from a mental disease or defect and could not conform his actions to the law.
In my view, the adjudication of Lindh’s claim in the Wisconsin courts resulted, in the words of § 2254(d)(1), in a decision that was contrary to clearly established federal law as determined by the Supreme Court of the United States. To the extent it might better be characterized as a mixed question of law and fact, the Wisconsin decision was an unreasonable application of clearly established federal law. For the reasons I have explained, I would grant the writ of habeas corpus.