dissenting.
I.
I entirely agree with Judge Cummings that rape shield laws serve a necessary and laudable purpose; on this score as well as on the other issues he has addressed, I join Judge Cummings’ persuasive dissent. As President Carter noted when he signed the legislation creating the federal rape shield law, embodied in Federal Rule of Evidence 412, such statutes are “designed to end the public degradation of rape victims and, by protecting victims from humiliation, to encourage the reporting of rape.” 14 Weekly Compilation of Presidential Documents 1902 (Oct. 30, 1978).
I, however, further agree that in the present circumstances the important (but here somewhat attenuated) interests protected by Indiana’s rape shield law must give way to the defendant’s right to put on a defense, The central problem to which rape shield *1012legislation is addressed was the common law practice in which courts “considered the victim’s character for chastity pertinent to whether or not she consented to the act that led to the charge of rape.” Vivian Berger, Man’s Trial, Woman’s Tribulation: Rape Cases in the Courtroom, 77 Colum.L.Rev. 1 (1977). The testimony that Stephens sought to offer, however, neither sought to prove the victim’s character nor was intended to address the question of consent. Rather than attempting to prove the truth of any matter about Wilburn’s character, Stephens ostensibly wanted to offer his story to show its effect on the listener. This, of course, is a common distinction in the law of evidence. See generally 6 John Henry Wigmore Evidence § 1789 (Chadbourn Rev.1976). Nor was Stephens’ testimony intended as evidence of consent, but rather it is evidence of a motive to fabricate. Stephens’ theory was that what he said so enraged Wilburn that it led her to concoct a rape charge.1
None of this suggests that there is any merit to Stephens’ claim that he is entitled to habeas relief on the grounds that the Indiana Supreme Court misapplied state law. The Indiana Supreme Court’s definitive construction of the state’s statute is, as far as the federal courts are concerned, a correct interpretation of state law. See Murdock v. Memphis, 87 U.S. (20 Wall.) 590, 22 L.Ed. 429 (1875). But whether the state court gets the state law right or wrong, a defendant is entitled to a writ of habeas corpus only if he is held in custody “in violation of the Constitution or laws ... of the United States.” 28 U.S.C. § 2254(a). But, because the scope of a defendant’s constitutional right to testify on his own behalf requires a weighing of the relevant interests, it is important to recognize that the state’s interest here is not, as the majority would have it, the very interest that undergirds all rape shield legislation. Maj. at 1003.
A criminal defendant’s right to testify on his own behalf is a part of his larger right to call witnesses and present evidence in his defense. See Rock v. Arkansas, 483 U.S. 44, 52, 107 S.Ct. 2704, 2709, 97 L.Ed.2d 37 (1987). The right to offer the testimony of witnesses is “in plain terms the right to present a defense, the right to present the defendant’s version of the facts as well as the prosecution’s to the jury so it may decide where the truth lies.... This right is a fundamental element of due process of law.” Washington v. Texas, 388 U.S. 14, 19, 87 S.Ct. 1920, 1923, 18 L.Ed.2d 1019 (1967).
But the defendant’s right to present the testimony of witnesses is not boundless. Rather, as the majority explains, it may be required to bow to other legitimate interests. Maj. at 1002. It is therefore clear that rape shield laws are constitutional, at least when applied to prevent a defendant from introducing — as proof of consent — evidence of the victim’s sexual history or reputation. See Moore v. Duckworth, 687 F.2d 1063 (7th Cir.1982); cf. Michigan v. Lucas, 500 U.S. 145, 111 S.Ct. 1743, 114 L.Ed.2d 205 (1991) (upholding a procedural requirement of Michigan’s rape shield law as against constitutional challenge).
But the balance is a delicate one, far more delicate than the majority’s somewhat curso*1013ry treatment would suggest. Moreover, our efforts at striking this balance should be sensitive to the guidance that Supreme Court precedent provides. To this end, cases addressing the right to cross-examine witnesses provide a useful benchmark, since the Court has recognized that a defendant’s dual rights to cross-examine the prosecution’s witnesses and to bring forward his own evidence serve similar interests. “Just as an accused has the right to confront the prosecution’s witnesses for the purpose of challenging their testimony, he has the right to present his own witnesses to establish a defense.” Washington, 388 U.S. at 19, 87 S.Ct. at 1923; see also Peter Westen, Confrontation and Compulsory Process: A Unified Theory of Evidence for Criminal Cases, 91 Harv.L.Rev. 567 (1978).
In particular, the Court’s decision in Davis v. Alaska, 415 U.S. 308, 94 S.Ct. 1105, 39 L.Ed.2d 347 (1974), is instructive here. The Supreme Court there balanced a criminal defendant’s right to cross-examine adverse witnesses against a state’s interest in protecting the anonymity of juvenile offenders. In Davis, the key prosecution witness was, at the time of trial, on probation by order of a juvenile court. The defendant sought to cross-examine the witness, suggesting to the jury that his probationary status might lead him to curry favor with the prosecutor’s office, and might thereby be a source of bias. The trial court — ultimately to be affirmed by the Alaska Supreme Court — refused to allow the cross-examination. According to the trial court, the introduction of such evidence was forbidden by a state law- protecting the anonymity of juvenile offenders. The state insisted that this law did not deprive the defendant of his Sixth Amendment rights, citing the state’s interest in sheltering juvenile offenders from embarrassment and injury to reputation. The Supreme Court, however, struck the balance otherwise. It held that the defendant’s interest in using this evidence — to suggest that the witness had a motive to he — outweighed the state’s interest in protecting the witness’s reputation. Since Davis, the Court has consistently reaffirmed a defendant’s right to introduce evidence suggesting a had a motive to fabricate. See Delaware v. Van Arsdall, 475 U.S. 673, 106 S.Ct. 1431, 89 L.Ed.2d 674 (1986); Olden v. Kentucky, 488 U.S. 227, 109 S.Ct. 480, 102 L.Ed.2d 513 (1988) (per cu-riam).
The analogy between Davis and the instant case is compelling. As in Davis, Stephens sought to show that the prosecution witness had a reason to he. In addressing a constitutional challenge to rape shield legislation, “Davis is clearly apropos when the rape victim’s prior sexual conduct has some relevance to estabhsh a motive for false accusation.” 23 Charles Alan Wright & Kenneth W. Graham, Jr., Federal Practice and Procedure: Evidence § 5387, at 568 (1980) (emphasis added). “Davis compels the courts to ‘carve a fairly broad exception [for bias impeachment] in statutes generally barring proof of sexual history.’ ” Edward J. Im-winkelreid, Exculpatory Evidence § 8-7, at 199 (1990) (citation omitted).
As far as I have discovered, every court that has applied Davis and its progeny to rape shield legislation has found that such laws cannot be invoked to prevent a defendant from introducing evidence that a prosecution witness had a motive to fabricate.2 See United States v. Stamper, 766 F.Supp. 1396 (W.D.N.C.1991), aff'd sub nom. without op. Re One Female Juvenile Victim, 959 F.2d 231 (4th Cir.1992) (evidence that victim previously admitted to falsely accusing her mother’s' boyfriend of sexual molestation must be admitted where victim claims that her father’s friend molested her); Lewis v. State, 591 So.2d 922 (Fla.1991) (evidence of victim’s sexual activity to be admitted where victim lied to her mother about her sexual activity and accused her stepfather of rape only days before gynecological examination, scheduled by her mother, to determine whether daughter was sexually active); State v. DeLawder, 28 Md.App. 212, 344 A.2d 446 (1975) (evidence that victim was sexually active must be admitted where victim allegedly accused defendant of rape out of fear of telling her mother that she was pregnant); *1014State v. Jalo, 27 Or.App. 845, 557 P.2d 1359 (1976) (evidence of victim’s sexual conduct to be admitted where ten-year-old victim engaged in sexual conduct with defendant’s 13 year-old son, and allegedly accused defendant of attempted rape after defendant told victim that he would tell her parents). Cf. Sandoval v. Acevedo, 996 F.2d 145 (7th Cir.1993), cert. denied, — U.S. -, 114 S.Ct. 307, 126 L.Ed.2d 255 (1993) (declining to decide — in light of its finding that any error was harmless — whether the Confrontation Clause entitles defendant to impeach victim’s own statements about her sexual history); Wood v. Alaska, 957 F.2d 1544, 1550 (9th Cir.1992), quoting Davis, 415 U.S. at 318, 94 S.Ct. at 1111 (while a defendant has a constitutional right to cross-examine witnesses to show the jury facts from which the jurors “ ‘could appropriately draw inferences relating to the reliability of the witness,’ ” the fact that here the victim “posed in the nude or acted in pornographic performances does not in any way indicate that she is a dishonest person or had a motive to lie in this case”).
But this is not to say that a defendant has a constitutional right to trot out a witness’ sexual history whenever he claims that this history gives rise to a motive to fabricate. The constitutional right under Davis should outweigh the state’s interest in protecting rape victims only where the proffered evidence, if true, would plausibly provide the witness with a reason to contrive a rape charge. Some courts appear to have lost sight of this fact. For example, in Chew v. Texas, 804 S.W.2d 633 (Tex.App.1991), pet. for discretionary rev. ref'd (July 3, 1991), the court concluded that under Davis, the defendant had a Sixth Amendment right to introduce evidence that the victim suffered from the illness of nymphomania, and therefore “had a motive to lie initially about her lack of consent to hide her sexual affliction from the public.” See also Winfield v. Commonwealth, 225 Va. 211, 301 S.E.2d 15, 21 (1983) (evidence that victim had extorted money from those with whom she had sexual relations somehow found to be “relevant and probative of a motive to fabricate”). Similarly, in Commonwealth v. Joyce, 382 Mass. 222, 415 N.E.2d 181, 187 (1981), the court concluded that the defendant had a constitutional right to show that the victim was a prostitute, on the theory that such activity had something to do with her motive to lie. According to the court, the victim, who was found naked in a car with the defendant, may have been “motivated falsely to accuse the defendant of rape by a desire to avoid further prosecution.” Id. For one, such an approach “broadly applied, denies sex workers who dare to complain of sexual violence the presumption of innocence.” Mary Joe Frug, A Postmodern Feminist Legal Manifesto (An Unfinished Draft), 105 Harv. L.Rev. 1045, 1057 (1992). Moreover, uncritical acceptance of a defendant’s every claim that the victim’s sexual history relates to a motive to he would expand a defendant’s right to present a defense — and by the same token undermine the purposes behind rape shield legislation — far beyond the bounds contemplated by the Supreme Court’s decisions in Davis, Van Arsdall, or Olden. A defendant has a constitutional right to introduce evidence of the victim’s sexual history only where the court finds it at least plausible that this history could in some way relate to a motive to fabricate.
To do otherwise runs the risk that the narrow exception for evidence of motive to fabricate swallow the entire rape shield law. This would be the result, of course, if a court were to conclude that the fact that a victim was sexually active somehow made her less worthy of belief. Insofar as the courts in Chew, Winfield, and Joyce move in this direction, they read Davis too broadly. But drawing a bright line between those cases in which the defendant plausibly suggests that the victim’s sexual history does and eases where it does not give rise to a motive to fabricate is no easy task. In the end, it is simply a matter that requires a trial judge to exercise prudent judgment.
Applying this principle to the case at bar presents a difficult question. On the one hand, to suggest that a woman who is shown to have been sexually active would become so angry that she would invent a rape charge may be more than a bit old-fashioned. But on the other, it is the former pervasiveness of this traditionalist view that ultimately led the Congress and forty-eight state legislatures to *1015enact rape shield legislation, protecting rape victims from having to endure trials in which they are cross-examined about their entire sexual past. It is paradoxically because being confronted with one’s sexual activity — or at least with its more lurid aspects — is thought to be the source of extreme embarrassment that Stephens has a right to present this evidence. Thus, I ultimately conclude that Stephens’ claim — that what he told Wilburn so enraged her that it led her to accuse him of rape — is some evidence of a motive to fabricate. As such, the Constitution requires that he be permitted to introduce this evidence at trial. Thus, unlike the overwhelming majority of applications of rape shield laws — applications that pose no serious constitutional difficulty — Indiana’s rape shield law here served to deprive Stephens of his constitutional right to present a defense.
II.
The majority, having found no constitutional violation, had no occasion to address the possible question of harmless error. Because I find that Stephens’ constitutional rights were violated, I next examine whether this violation may be treated as a harmless error.
While the district court found no constitutional error, it had no hesitation in concluding that, had there been error, it would have been harmless. According to the district court, Stephens’ version of the story is “inconceivable” because only fifteen minutes elapsed between the time Stephens arrived at Wilburn’s house and the time he arrived at the Straits’ house, following the incident. Thus, “the jury would have been hard pressed to believe this whirlwind courtship took place in a matter of fifteen minutes.” Stephens v. Morris, 756 F.Supp. at 1143. Because the court was “persuaded that no reasonable jury would have found the proffered testimony would have raised a doubt as to the petitioner’s guilt,” depriving him of his constitutional right to offer that testimony was a harmless error. Id.
This is a common approach to harmless error analysis. Where an error is not harmless, the necessary remedy is a new trial. However, many appellate judges (as well as trial judges reviewing state convictions in habeas cases) apparently believe that, where “there is no legally sufficient evidentiary basis for a reasonable jury” to acquit the defendant, there is no point in ordering a new trial. Instead, they essentially enter summary judgment for the prosecution, declaring any constitutional error to be harmless. See Fed.R.Civ.P. 56(a)(1).
The only problem with this procedure is that it is patently unconstitutional. All criminal defendants — even the most guilty of them — have a- constitutional right to have a jury, not an appellate judge, find them guilty beyond a reasonable doubt. While I do not believe this proposition ever to have been in serious doubt, the Supreme Court, speaking through Justice Scalia, made the matter abundantly clear last Term. The harmless error question asks:
[N]ot what effect the constitutional error might generally be expected to have upon a reasonable jury, but rather what effect it had upon the guilty verdict in the ease at hand. Harmless-error review looks, we have said, to the basis on which the “jury actually rested its verdict.” The inquiry, in other words, is not whether, in a trial that occurred without the error, a guilty verdict would surely have been rendered, but whether the guilty verdict actually rendered in this trial was surely unattributable to the error. That must be so, because to hypothesize a guilty verdict that was never in fact rendered — no matter how inescapable the findings to support that verdict might be — would violate the jury-trial guarantee.
Sullivan v. Louisiana, — U.S. -, - - -, 113 S.Ct. 2078, 2081-82, 124 L.Ed.2d 182 (1993) (citations omitted) (emphasis in original).
This principle helps explain the distinction that the Court drew, in Arizona v. Fulminante, 499 U.S. 279, 111 S.Ct. 1246, 113 L.Ed.2d 302 (1991), between “trial errors”, which may be treated as harmless, and “structural errors” which may not. While most constitutional errors may be treated as harmless, there are some rights that “are so *1016basic to a fair trial that their infraction can never be treated as harmless error.” Chapman v. California, 386 U.S. 18, 23, 87 S.Ct. 824, 827-28, 17 L.Ed.2d 705 (1967). Where the error so infects the trial that it would be impossible to determine that the error did not affect the jury verdict, harmless error analysis is infeasible, and a rule of per se reversal applies. See Charles J. Ogletree, Jr., Arizona v. Fulminante: The Harm of Applying Harmless Error to Coerced Confessions, 105 Harv.L.Rev. 152, 162 (1991) (“a trial error seems to be one for which we can sometimes know for sure whether it has caused inaccuracy in a trial outcome, and a structural error seems to be one for which we can never know with any certainty”).
Most constitutional errors involve the wrongful admission of evidence that ought to be excluded. All such errors are “trial errors,” and harmless error analysis is unproblematic. The Fulminante Court therefore concluded that in “reviewing the erroneous admission of an involuntary confession, the appellate court, as it does with the admission of other forms of improperly admitted evidence, simply reviews the remainder of the evidence against the defendant to determine whether the admission of the confession was harmless beyond a reasonable doubt.” Fulminante, 499 U.S. at 310, 111 S.Ct. at 1265.3
Unlike “trial errors,” which may be treated as harmless, “structural errors” require per se reversal. It is perhaps arguable that the error here — interference with a defendant’s right to present evidence — is a structural error. For example, the Supreme Court noted in Rose v. Clark, 478 U.S. 570, 578, 106 S.Ct. 3101, 3106, 92 L.Ed.2d 460 (1986) (emphasis added), that harmless error analysis “presupposes a trial, at which the defendant, represented by counsel, may present evidence and argument before an impartial judge and jury.”
But this may overstate the case, as the Court held in Delaware v. Van Arsdall, 475 U.S. 673, 106 S.Ct. 1431, 89 L.Ed.2d 674 (1986), that a denial of a defendant’s right to cross-examine adverse witnesses (which is closely analogous to the right to present evidence) may be treated as harmless. But the mechanics of the analysis are instructive here. In Van Arsdall, the Court addressed the argument that all errors involving the erroneous exclusion of evidence require per se reversal. “Because it is impossible to know how wrongfully excluded evidence would have affected the jury, the argument runs, reversal is mandated.” Id. at 683, 106 S.Ct. at 1437.
The Court rejected this contention. While Confrontation Clause errors are amenable to harmless-error analysis, the “correct inquiry is whether, assuming that the damaging potential of the cross-examination were fully realized, a reviewing court might nonetheless say that the error was harmless beyond a reasonable doubt.” Id. at 684, 106 S.Ct. at 1438 (emphasis added).
Fairly read, Van Arsdall requires that, where a trial court deprives a defendant of his constitutional right to attack the credibility of a prosecution witness, a reviewing court engaging in harmless error analysis must assume that the defendant would have fully impeached the witness’ credibility. In effect, the reviewing court should assume that the impeachment would have been so devastating that the error, in effect, was in admitting the testimony of the prosecution witness in the first instance. The court should then engage in the conventional harmless error analysis, setting aside the evidence that was “erroneously admitted,” and deciding the effect that this error had on the verdict. If the court is convinced that the verdict did not rely on the testimony of the prosecution witness, the error may be treated as harmless and the conviction affirmed. Cf. Sandoval, 996 F.2d at 149-50 (limiting defendant’s ability to impeach victim’s statements rendered harmless *1017by trial judge’s instructing the jury to disregard the statement).
In this case, Stephens sought to introduce testimony suggesting that Wilburn was fabricating the rape charge. Depriving Stephens of his opportunity to testify to this effect was constitutional error. This error was harmless only if the court can say that Wilburn’s testimony did not have a “substantial and injurious effect or influence in determining the jury’s verdict.” Because Wilburn was the key prosecution witness, we obviously cannot make such an assertion.
Because Stephens'was denied his constitutional right to testify on his own behalf, and because we cannot conclude that the error was harmless, the district court’s denial of habeas corpus relief should be reversed. Stephens is entitled to a new trial, one in which he is permitted to tell his story to the jury. If the story is as implausible as the district court believes, Stephens will again be convicted in short order. But because Stephens’ first trial was constitutionally deficient, I respectfully dissent.
. There is some force to the suggestion that this evidence was not introduced solely to show a motive to fabricate but also to explain why Wilburn withdrew her consent and threw Stephens out of the trailer. In fact, Stephens’ counsel also suggested that in addition to being upset by Stephens' remarks, Wilburn fabricated the rape charge in order to appease her landlady who was upset by the disturbance. But the Indiana Supreme Court, in reviewing the state court proceeding, found that Stephens sought to introduce the evidence to show both the reason why he was thrown out of the trailer and as a motive to fabricate. “[According to Stephens] she became so angry she made him stop and leave. He claimed her anger caused her to bring this criminal action against him.” Stephens v. State, 544 N.E.2d 137, 138-39 (Ind.1989). The district court on habeas review did not take issue with the state court's description of its proceedings, adding that Stephens further argued that because he was not allowed to introduce the evidence, “he was precluded from presenting his argument to the jury with regard to how the evidence was important.” Stephens v. Morris, 756 F.Supp. 1137, 1141-42 (N.D.Ind.1991). While the record does not make clear how prominent a role the excluded evidence would have played in Stephens' defense, I am inclined neither to reject the description of Stephens’ defense offered by each of the courts previously to review the proceeding nor to hold Stephens to a higher standard of clarity.
. The majority, of course, avoids this question by simply overlooking Davis and the particularized constitutional right to present evidence that attacks the credibility of prosecution witnesses.
. The Court's decision in Brecht v. Abrahamson, - U.S. -, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993), abolished the use of the "harmless beyond a reasonable doubt” standard on habeas review. Instead, the Court found that on collateral review a court is to treat an error as harmless unless it "had substantial and injurious ef-feet or influence in determining the jury’s verdict." Id. -U.S. at -, 113 S.Ct. at 1718, quoting Kotteakos v. United States, 328 U.S. 750, 776, 66 S.Ct. 1239, 1253, 90 L.Ed. 1557 (1946). But Brecht affects only the standard of review, not the mechanics of the analysis.