I agree with most of what Justice Marshall has said in his dissenting opinion. I, too, dissent, but I base my conclusion also on evidentiary factors that the Court rejects with some emphasis. The Court holds that psychiatric testimony about a defendant’s future dangerousness is admissible, despite the fact that such testimony is wrong two times out of three. The Court reaches this result — even in a capital case — because, it is said, the testimony is subject to cross-examination and impeachment. In the present state of psychiatric knowledge, this is too much for me. One may accept this in a routine lawsuit for money damages, but when a person’s life is at stake — no matter how heinous his offense — a requirement of greater reliability should prevail. In a capital case, the specious testimony of a psychiatrist, colored in the eyes of an impressionable jury by the inevitable untouchability of a medical specialist’s words, equates with death itself.
I
To obtain a death sentence in Texas, the State is required to prove beyond a reasonable doubt that “there is a probability that the defendant would commit criminal acts of violence that would constitute a continuing threat to society.” Tex. *917Code Crim. Proc. Ann., Art. 37.071(b)(2) (Vernon 1981). As a practical matter, this prediction of future dangerousness was the only issue to be decided by Barefoot’s sentencing jury.1
At the sentencing hearing, the State established that Barefoot had two prior convictions for drug offenses and two prior convictions for unlawful possession of firearms. None of these convictions involved acts of violence. At the guilt stage of the trial, for the limited purpose of establishing that the crime was committed in order to evade police custody, see Barefoot v. State, 596 S. W. 2d 875, 886-887 (Tex. Crim. App. 1980), cert. denied, 453 U. S. 913 (1981), the State had presented evidence that Barefoot had escaped from jail in New Mexico where he was being held on charges of statutory rape and unlawful restraint of a minor child with intent to commit sexual penetration against the child’s will. The prosecution also called several character witnesses at the sentencing hearing, from towns in five States. Without mentioning particular examples of Barefoot’s conduct, these witnesses testified that Barefoot’s reputation for being a peaceable and law-abiding citizen was bad in their respective communities.
Last, the prosecution called Doctors Holbrook and Grigson, whose testimony extended over more than half the hearing. Neither had examined Barefoot or requested the opportunity to examine him. In the presence of the jury, and over defense counsel’s objection, each was qualified as an expert psychiatrist witness. Doctor Holbrook detailed at length his training and experience as a psychiatrist, which included a position as chief of psychiatric services at the Texas *918Department of Corrections. He explained that he had previously performed many “criminal evaluations,” Trial Tr. 2069, and that he subsequently took the post at the Department of Corrections to observe the subjects of these evaluations so that he could “be certain those opinions that [he] had were accurate at the time of trial and pretrial.” Id., at 2070. He then informed the jury that it was “within [his] capacity as a doctor of psychiatry to predict the future dangerousness of an individual within a reasonable medical certainty,” id., at 2072 (emphasis supplied), and that he could give “an expert medical opinion that would be within reasonable psychiatric certainty as to whether or not that individual would be dangerous to the degree that there would be a probability that that person would commit criminal acts of violence in the future that would constitute a continuing threat to society,” id., at 2073 (emphasis supplied).
Doctor Grigson also detailed his training and medical experience, which, he said, included examination of “between thirty and forty thousand individuals,” including 8,000 charged with felonies, and at least 300 charged with murder. Id., at 2109. He testified that with enough information he would be able to “give a medical opinion within reasonable psychiatric certainty as to the psychological or psychiatric makeup of an individual,” id., at 2110 (emphasis supplied), and that this skill was “particular to the field of psychiatry and not to the average layman.” Id., at 2111.
Each psychiatrist then was given an extended hypothetical question asking him to assume as true about Barefoot the four prior convictions for nonviolent offenses, the bad reputation for being law-abiding in various communities, the New Mexico escape, the events surrounding the murder for which he was on trial and, in Doctor Grigson’s case, the New Mexico arrest. On the basis of the hypothetical question, Doctor Holbrook diagnosed Barefoot “within a reasonable psychiatric] certainty,” as a “criminal sociopath.” Id., at 2097. He testified that he knew of no treatment that could change *919this condition, and that the condition would not change for the better but “may become accelerated” in the next few years. Id., at 2100. Finally, Doctor Holbrook testified that, “within reasonable psychiatric certainty,” there was “a probability that the Thomas A. Barefoot in that hypothetical will commit criminal acts of violence in the future that would constitute a continuing threat to society,” and that his opinion would not change if the “society” at issue was that within Texas prisons rather than society outside prison. Id., at 2100-2101.
Doctor Grigson then testified that, on the basis of the hypothetical question, he could diagnose Barefoot “within reasonable psychiatric certainty” as an individual with “a fairly classical, typical, sociopathic personality disorder.” Id., at 2127-2128. He placed Barefoot in the “most severe category” of sociopaths (on a scale of one to ten, Barefoot was “above ten”), and stated that there was no known cure for the condition. Id., at 2129. Finally, Doctor Grigson testified that whether Barefoot was in society at large or in a prison society there was a “one hundred percent and absolute” chance that Barefoot would commit future acts of criminal violence that would constitute a continuing threat to society. Id., at 2131 (emphasis supplied).
On cross-examination, defense counsel questioned the psychiatrists about studies demonstrating that psychiatrists’ predictions of future dangerousness are inherently unreliable. Doctor Holbrook indicated his familiarity with many of these studies but stated that he disagreed with their conclusions. Doctor Grigson stated that he was not familiar with most of these studies, and that their conclusions were accepted by only a “small minority group” of psychiatrists— “[i]t’s not the American Psychiatric Association that believes that.” Id., at 2134.
After an hour of deliberation, the jury answered “yes” to the two statutory questions, and Thomas Barefoot was sentenced to death.
*920rH HH
A
The American Psychiatric Association (APA), participating in this case as amicus curiae, informs us that “[t]he unreliability of psychiatric predictions of long-term future dangerousness is by now an established fact within the profession.” Brief for American Psychiatric Association as Amicus Curiae 12 (APA Brief). The APA’s best estimate is that two out of three predictions of long-term future violence made by psychiatrists are wrong. Id., at 9, 13. The Court does not dispute this proposition, see ante, at 899-901, n. 7, and indeed it could not do so; the evidence is overwhelming. For example, the APA’s Draft Report of the Task Force on the Role of Psychiatry in the Sentencing Process (1983) (Draft Report) states that “[considerable evidence has been accumulated by now to demonstrate that long-term prediction by psychiatrists of future violence is an extremely inaccurate process.” Id., at 29. John Monahan, recognized as “the leading thinker on this issue” even by the State’s expert witness at Barefoot’s federal habeas corpus hearing, Hearing Tr. 195, concludes that “the ‘best’ clinical research currently in existence indicates that psychiatrists and psychologists are accurate in no more than one out of three predictions of violent behavior,” even among populations of individuals who are mentally ill and have committed violence in the past. J. Monahan, The Clinical Prediction of Violent Behavior 47-49 (1981) (emphasis deleted) (J. Monahan, Clinical Prediction); see also id., at 6-7, 44-50. Another study has found it impossible to identify any subclass of offenders “whose members have a greater-than-even chance of engaging again in an assaultive act.” Wenk, Robison, & Smith, Can Violence Be Predicted?, 18 Crime & Delinquency 393, 394 (1972). Yet another commentator observes: “In general, mental health professionals . . . are more likely to be wrong than right when they predict legally relevant behavior. When predicting violence, dangerousness, and suicide, they are far more *921likely to be wrong than right.” Morse, Crazy Behavior, Morals, and Science: An Analysis of Mental Health Law, 51 S. Cal. L. Rev. 527, 600 (1978) (Morse, Analysis of Mental Health Law). Neither the Court nor the State of Texas has cited a single reputable scientific source contradicting the unanimous conclusion of professionals in this field that psychiatric predictions of long-term future violence are wrong more often than they are right.2
The APA also concludes, see APA Brief 9-16, as do researchers that have studied the issue,3 that psychiatrists simply have no expertise in predicting long-term fixture dan*922gerousness. A layman with access to relevant statistics can do at least as well and possibly better; psychiatric training is not relevant to the factors that validly can be employed to make such predictions, and psychiatrists consistently err on the side of overpredicting violence.4 Thus, while Doctors Grigson and Holbrook were presented by the State and by self-proclamation as experts at predicting future dangerousness, the scientific literature makes crystal clear that they had no expertise whatever. Despite their claims that they were able to predict Barefoot’s future behavior “within reasonable psychiatric certainty,” or to a “one hundred percent and absolute” certainty, there was in fact no more than a one in three chance that they were correct.5
*923B
It is impossible to square admission of this purportedly scientific but actually baseless testimony with the Constitution’s paramount concern for reliability in capital sentencing.6 *924Death is a permissible punishment in Texas only if the jury finds beyond a reasonable doubt that there is a probability the defendant will commit future acts of criminal violence. The admission of unreliable psychiatric predictions of future violence, offered with unabashed claims of “reasonable medical certainty” or “absolute” professional reliability, creates an intolerable danger that death sentences will be imposed erroneously.
The plurality in Woodson v. North Carolina, 428 U. S. 280, 305 (1976), stated:
“Death, in its finality, differs more from life imprisonment than a 100-year prison term differs from one of only a year or two. Because of that qualitative difference, there is a corresponding difference in the need for reliability in the determination that death is the appropriate punishment in a specific case.”
The Court does not see fit to mention this principle today, yet it is as firmly established as any in our Eighth Amendment jurisprudence. Only two weeks ago, in Zant v. Stephens, 462 U. S. 862, 884 (1983), the Court described the need for reliability in the application of the death penalty as one of the *925basic “themes . . . reiterated in our opinions discussing the procedures required by the Constitution in capital sentencing determinations.” See Eddings v. Oklahoma, 455 U. S. 104, 110-112 (1982) (capital punishment must be “imposed fairly, and with reasonable consistency, or not at all”); id., at 118-119 (O’Connor, J., concurring); Beck v. Alabama, 447 U. S. 625, 637-38, and n. 13 (1980); Green v. Georgia, 442 U. S. 95, 97 (1979); Lockett v. Ohio, 438 U. S. 586, 604 (1978) (plurality opinion); Gardner v. Florida, 430 U. S. 349, 359 (1977) (plurality opinion); id., at 363-364 (White, J., concurring in judgment). State evidence rules notwithstanding, it is well established that, because the truth-seeking process may be unfairly skewed, due process may be violated even in a noncapital criminal case by the exclusion of evidence probative of innocence, see Chambers v. Mississippi, 410 U. S. 284 (1973), or by the admission of certain categories of unreliable and prejudicial evidence, see Watkins v. Sowders, 449 U. S. 341, 347 (1981) (“[i]t is the reliability of identification evidence that primarily determines its admissibility”); Foster v. California, 394 U. S. 440 (1969).7 The reliability and admissibility of evidence considered by a capital sentencing factfinder is obviously of still greater constitutional concern. Cf. Green v. Georgia, 442 U. S. 95 (1979); Gardner v. Florida, 430 U. S. 349 (1977).
The danger of an unreliable death sentence created by this testimony cannot be brushed aside on the ground that the “ ‘jury [must] have before it all possible relevant information about the individual defendant whose fate it must determine.”’ Ante, at 897, quoting Jurek v. Texas, 428 U. S. 262, 276 (1976) (joint opinion announcing the judgment). Although committed to allowing a “wide scope of evidence” at presentence hearings, Zant v. Stephens, 462 U. S., at 886, *926the Court has recognized that “consideration must be given to the quality, as well as the quantity, of the information on which the sentencing [authority] may rely.” Gardner v. Florida, 430 U. S., at 359. Thus, very recently, this Court reaffirmed a crucial limitation on the permissible scope of evidence: “ ‘[sjo long as the evidence introduced . . . do[es] not prejudice a defendant, it is preferable not to impose restrictions.’” Zant v. Stephens, 462 U. S., at 886, quoting Gregg v. Georgia, 428 U. S. 153, 203-204 (1976) (emphasis supplied). The Court all but admits the obviously prejudicial impact of the testimony of Doctors Grigson and Holbrook; granting that their absolute claims were more likely to be wrong than right, ante, at 899, n. 7, 901, the Court states that “[t]here is no doubt that the psychiatric testimony increased the likelihood that petitioner would be sentenced to death,” ante, at 905.
Indeed, unreliable scientific evidence is widely acknowledged to be prejudicial. The reasons for this are manifest. “The major danger of scientific evidence is its potential to mislead the jury; an aura of scientific infallibility may shroud the evidence and thus lead the jury to accept it without critical scrutiny.” Giannelli, The Admissibility of Novel Scientific Evidence: Frye v. United States, a Half-Century Later, 80 Colum. L. Rev. 1197, 1237 (1980) (Giannelli, Scientific Evidence).8 Where the public holds an exaggerated opinion of *927the accuracy of scientific testimony, the prejudice is likely to be indelible. See United States v. Baller, 519 F. 2d 463, 466 (CA4), cert. denied, 423 U. S. 1019 (1975). There is little question that psychiatrists are perceived by the public as having a special expertise to predict dangerousness, a perception based on psychiatrists’ study of mental disease. See J. Robitscher, The Powers of Psychiatry 187-188 (1980); Cocozza & Steadman, supra n. 2, 25 Soc. Probs., at 273; Morse, Analysis of Mental Health Law, 51 S. Cal. L. Rev., at 533-536. It is this perception that the State in Barefoot’s case sought to exploit. Yet mental disease is not correlated with violence, see J. Monahan, Clinical Prediction 77-82; Steadman & Cocozza, supra n. 2, 69 J. Crim. L. & Criminology, at 230, and the stark fact is that no such expertise exists. Moreover, psychiatrists, it is said, sometimes attempt to perpetuate this illusion of expertise, Cocozza & Steadman, supra n. 2, 25 Soc. Probs., at 274, and Doctors Grigson and Holbrook — who purported to be able to predict future dangerousness “within reasonable psychiatric certainty,” or absolutely — present extremely disturbing exam-*928pies of this tendency. The problem is not uncommon. See Giannelli, Scientific Evidence, 80 Colum. L. Rev., at 1238.
Furthermore, as is only reasonable, the Court’s concern in encouraging the introduction of a wide scope of evidence has been to ensure that accurate information is provided to the sentencing authority without restriction. The joint opinion announcing the judgment in Gregg explained the jury’s need for relevant evidence in these terms:
“If an experienced trial judge, who daily faces the difficult task of imposing sentences, has a vital need for accurate information ... to be able to impose a rational sentence in the typical criminal case, then accurate sentencing information is an indispensable prerequisite to a reasoned determination of whether a defendant shall live or die by a jury of people who may never before have made a sentencing decision.” 428 U. S., at 190 (emphasis supplied).
See California v. Ramos, post, at 1004 (Court holds jury instruction permissible at sentencing hearing on ground that it “gives the jury accurate information”) (emphasis supplied). So far as I am aware, the Court never has suggested that there is any interest in providing deceptive and inaccurate testimony to the jury.
Psychiatric predictions of future dangerousness are not accurate; wrong two times out of three, their probative value, and therefore any possible contribution they might make to the ascertainment of truth, is virtually nonexistent. See Cocozza & Steadman, supran. 2, 29 Rutgers L. Rev., at 1101 (psychiatric testimony not sufficiently reliable to support finding that individual will be dangerous under any standard of proof). Indeed, given a psychiatrist’s prediction that an individual will be dangerous, it is more likely than not that the defendant will not commit further violence. It is difficult to understand how the admission of such predictions can be justified as advancing the search for truth, particularly in light of their clearly prejudicial effect.
*929Thus, the Court’s remarkable observation that “[n]either petitioner nor the [APA] suggests that psychiatrists are always wrong with respect to future dangerousness, only most of the time,” ante, at 901 (emphasis supplied), misses the point completely, and its claim that this testimony was no more problematic than “other relevant evidence against any defendant in a criminal case,” ante, at 905-906, is simply incredible. Surely, this Court’s commitment to ensuring that death sentences are imposed reliably and reasonably requires that nonprobative and highly prejudicial testimony on the ultimate question of life or death be excluded from a capital sentencing hearing.
Ill
A
Despite its recognition that the testimony at issue was probably wrong and certainly prejudicial, the Court holds this testimony admissible because the Court is “unconvinced . . . that the adversary process cannot be trusted to sort out the reliable from the unreliable evidence and opinion about future dangerousness.” Ante, at 901; see ante, at 899-901, n. 7. One can only wonder how juries are to separate valid from invalid expert opinions when the “experts” themselves are so obviously unable to do so. Indeed, the evidence suggests that juries are not effective at assessing the validity of scientific evidence. Giannelli, Scientific Evidence, 80 Colum. L. Rev., at 1239-1240, and n. 319.
There can be no question that psychiatric predictions of future violence will have an undue effect on the ultimate verdict. Even judges tend to accept psychiatrists’ recommendations about a defendant’s dangerousness with little regard for cross-examination or other testimony. Cocozza & Stead-man, supra n. 2, 25 Soc. Probs., at 271 (in making involuntary commitment decisions, psychiatric predictions of future dangerousness accepted in 86.7% of cases); see Morse, Analysis of Mental Health Law, 51 S. Cal. L. Rev., at 536, n. 16, 603. There is every reason to believe that inexperienced *930jurors will be still less capable of “separating] the wheat from the chaff,” despite the Court’s blithe assumption to the contrary, ante, at 901, n. 7. The American Bar Association has warned repeatedly that sentencing juries are particularly incapable of dealing with information relating to “the likelihood that the defendant will commit other crimes,” and similar predictive judgments. ABA Project on Standards for Criminal Justice, Sentencing Alternatives and Procedures § 1.1(b), Commentary, pp. 46-47 (App. Draft 1968); ABA Standards for Criminal Justice 18-1.1, Commentary, pp. 18-16, 18-24 to 18-25 (2d ed. 1980). Relying on the ABA’s conclusion, the joint opinion announcing the judgment in Gregg v. Georgia, 428 U. S., at 192, recognized that “[s]ince the members of a jury will have had little, if any, previous experience in sentencing, they are unlikely to be skilled in dealing with the information they are given.” But the Court in this case, in its haste to praise the jury’s ability to find the truth, apparently forgets this well-known and worrisome shortcoming.
As if to suggest that petitioner’s position that unreliable expert testimony should be excluded is unheard of in the law, the Court relies on the proposition that the rules of evidence generally “anticipate that relevant, unprivileged evidence should be admitted and its weight left to the factfinder, who would have the benefit of cross-examination and contrary evidence by the opposing party.” Ante, at 898. But the Court simply ignores hornbook law that, despite the availability of cross-examination and rebuttal witnesses, “opinion evidence is not admissible if the court believes that the state of the pertinent art or scientific knowledge does not permit a reasonable opinion to be asserted. ” E. Cleary, McCormick on Evidence § 13, p. 31 (2d ed. 1972). Because it is feared that the jury will overestimate its probative value, polygraph evidence, for example, almost invariably is excluded from trials despite the fact that, at a conservative estimate, an experienced polygraph examiner can detect truth or deception correctly about 80 to 90 percent of the time. Ennis & Litwack, *931supra n. 2, at 736.9 In no area is purportedly “expert” testimony admitted for the jury’s consideration where it cannot be demonstrated that it is correct more often than not. “It is inconceivable that a judgment could be considered an ‘expert’ judgment when it is less accurate than the flip of a coin.” Id., at 737. The risk that a jury will be incapable of separating “scientific” myth from reality is deemed unacceptably high.10
B
The Constitution’s mandate of reliability, with the stakes at life or death, precludes reliance on cross-examination and the opportunity to present rebuttal witnesses as an antidote for this distortion of the truth-finding process. Cross-examination is unlikely to reveal the fatuousness of psyehi-*932atric predictions because such predictions often rest, as was the case here, on psychiatric categories and intuitive clinical judgments not susceptible to cross-examination and rebuttal. Dix, supra n. 2, at 44. Psychiatric categories have little or no demonstrated relationship to violence, and their use often obscures the unimpressive statistical or intuitive bases for prediction. J. Monahan, Clinical Prediction 31; Cocozza & Steadman, supra n. 2, 25 Soc. Probs., at 274.11 The APA particularly condemns the use of the diagnosis employed by Doctors Grigson and Holbrook in this case, that of sociopathy:
“In this area confusion reigns. The psychiatrist who is not careful can mislead the judge or jury into believing that a person has a major mental disease simply on the basis of a description of prior criminal behavior. Or a psychiatrist can mislead the court into believing that an individual is devoid of conscience on the basis of a description of criminal acts alone. . . . The profession of psychiatry has a responsibility to avoid inflicting this confusion upon the courts and to spare the defendant the harm that may result. . . . Given our uncertainty about the implications of the finding, the diagnosis of socio-pathy . . . should not be used to justify or to support predictions of future conduct. There is no certainty in this area.” Draft Report 30.
It is extremely unlikely that the adversary process will cut through the facade of superior knowledge. The Chief Justice long ago observed:
*933“The very nature of the adversary system . . . complicates the use of scientific opinion evidence, particularly in the field of psychiatry. This system of partisan contention, of attack and counterattack, at its best is not ideally suited to developing an accurate portrait or profile of the human personality, especially in the area of abnormal behavior. Although under ideal conditions the adversary system can develop for a jury most of the necessary fact material for an adequate decision, such conditions are rarely achieved in the courtrooms in this country. These ideal conditions would include a highly skilled and experienced trial judge and highly skilled lawyers on both sides of the case, all of whom in addition to being well-trained in the law and in the techniques of advocacy would be sophisticated in matters of medicine, psychiatry, and psychology. It is far too rare that all three of the legal actors in the cast meet these standards.” Burger, Psychiatrists, Lawyers, and the Courts, 28 Fed. Prob. 3, 6 (June 1964).
Another commentator has noted:
“Competent cross-examination and jury instructions may be partial antidotes . . . , but they cannot be complete. Many of the cases are not truly adversarial; too few attorneys are skilled at cross-examining psychiatrists, laypersons overweigh the testimony of experts, and, in any case, unrestricted use of experts promotes the incorrect view that the questions are primarily scientific. There is, however, no antidote for the major difficulty with mental health ‘experts’ — that they simply are not experts .... In realms beyond their true expertise, the law has little special to learn from them; too often their testimony is . . . prejudicial.” Morse, Analysis of Mental Health Law, 51 S. Cal. L. Rev., at 626.
*934See id., at 535-536. See also Dix, supra n. 2, at 44-45; Ennis & Litwack, supra n. 2, at 745; Note, supra n. 8, 70 Calif. L. Rev., at 1079-1080; J. Robitscher, The Powers of Psychiatry 202-203 (1980).
Nor is the presentation of psychiatric witnesses on behalf of the defense likely to remove the prejudicial taint of misleading testimony by prosecution psychiatrists.12 No reputable expert would be able to predict with confidence that the defendant will not be violent; at best, the witness will be able to give his opinion that all predictions of dangerousness are unreliable. Consequently, the jury will not be presented with the traditional battle of experts with opposing views on the ultimate question. Given a choice between an expert who says that he can predict with certainty that the defendant, whether confined in prison or free in society, will kill again, and an expert who says merely that no such prediction can be made, members of the jury charged by law with making the prediction surely will be tempted to opt for the expert who claims he can help them in performing their duty, and who predicts dire consequences if the defendant is not put to death.13
Moreover, even at best, the presentation of defense psychiatrists will convert the death sentence hearing into a battle of *935experts, with the Eighth Amendment’s well-established requirement of individually focused sentencing a certain loser. The jury’s attention inevitably will turn from an assessment of the propriety of sentencing to death the defendant before it to resolving a scientific dispute about the capabilities of psychiatrists to predict future violence. In such an atmosphere, there is every reason to believe that the jury may be distracted from its constitutional responsibility to consider “particularized mitigating factors,” see Jurek v. Texas, 428 U. S., at 272, in passing on the defendant’s future dangerousness. See Davis, supra n. 12, at 310.
One searches the Court’s opinion in vain for a plausible justification for tolerating the State’s creation of this risk of an erroneous death verdict. As one Court of Appeals has observed:
“A courtroom is not a research laboratory. The fate of a defendant. . . should not hang on his ability to successfully rebut scientific evidence which bears an ‘aura of special reliability and trustworthiness,’ although, in reality the witness is testifying on the basis of an unproved hypothesis . . . which has yet to gain general acceptance in its field.” United States v. Brown, 557 F. 2d 541, 556 (CA6 1977).
Ultimately, when the Court knows full well that psychiatrists’ predictions of dangerousness are specious, there can be no excuse for imposing on the defendant, on pain of his *936life, the heavy burden of convincing a jury of laymen of the fraud.14
IV
The Court is simply wrong in claiming that psychiatric testimony respecting future dangerousness is necessarily admissible in light of Jurek v. Texas, 428 U. S. 262 (1976), or Estelle v. Smith, 451 U. S. 454 (1981). As the Court recognizes, Jurek involved “only lay testimony.” Ante, at 897. Thus, it is not surprising that “there was no suggestion by the Court that the testimony of doctors would be inadmissible,” ibid., and it is simply irrelevant that the Jurek Court did not “disapproval” the use of such testimony, see Estelle v. Smith, 451 U. S., at 473.
*937In Smith, the psychiatric testimony at issue was given by the same Doctor Grigson who confronts us in this case, and his conclusions were disturbingly similar to those he rendered here. See id,., at 459-460. The APA, appearing as amicus curiae, argued that all psychiatric predictions of future dangerousness should be excluded from capital sentencing proceedings. The Court did not reach this issue, because it found Smith’s death sentence invalid on narrower grounds: Doctor Grigson’s testimony had violated Smith’s Fifth and Sixth Amendment rights. Id., at 473. Contrary to the Court’s inexplicable assertion in this case, ante, at 899, Smith certainly did not reject the APA’s position. Rather, the Court made clear that “the holding in Jurek was guided by recognition that the inquiry [into dangerousness] mandated by Texas law does not require resort to medical experts.” 451 U. S., at 473 (emphasis added). If Jurek and Smith held that psychiatric predictions of future dangerousness are admissible in a capital sentencing proceeding as the Court claims, this guiding recognition would have been irrelevant.
The Court also errs in suggesting that the exclusion of psychiatrists’ predictions of future dangerousness would be contrary to the logic of Jurek. Jurek merely upheld Texas’ substantive decision to condition the death sentence upon proof of a probability that the defendant will commit criminal acts of violence in the future. Whether the evidence offered by the prosecution to prove that probability is so unreliable as to violate a capital defendant’s rights to due process is an entirely different matter, one raising only questions of fair procedure.15 Jurek’s conclusion that Texas may impose the *938death penalty on capital defendants who probably will commit criminal acts of violence in no way establishes that the prosecution may convince a jury that this is so by misleading or patently unreliable evidence.
Moreover, Jurek’s holding that the Texas death statute is not impermissibly vague does not lead ineluctably to the conclusion that psychiatric testimony is admissible. It makes sense to exclude psychiatric predictions of future violence while admitting lay testimony, see ante, at 896-897, because psychiatric predictions appear to come from trained mental health professionals, who purport to have special expertise. In view of the total scientific groundlessness of these predictions, psychiatric testimony is fatally misleading. See White v. Estelle, 554 F. Supp., at 858. Lay testimony, frankly based on statistical factors with demonstrated correlations to violent behavior, would not raise this substantial threat of unreliable and capricious sentencing decisions, inimical to the constitutional standards established in our cases; and such predictions are as accurate as any a psychiatrist could make. Indeed, the very basis of Jurek, as I understood it, was that such judgments can be made by laymen on the basis of lay testimony.
Our constitutional duty is to ensure that the State proves future dangerousness, if at all, in a reliable manner, one that ensures that “any decision to impose the death sentence be, and appear to be, based on reason rather than caprice or emotion.” Gardner v. Florida, 480 U. S., at 358. Texas’ choice of substantive factors does not justify loading the factfinding process against the defendant through the presentation of what is, at bottom, false testimony.
V
I would vacate petitioner’s death sentence, and remand for further proceedings consistent with these views.
It appears that every person convicted of capital murder in Texas will satisfy the other requirement relevant to Barefoot’s sentence, that “the conduct of the defendant that caused the death of the deceased was committed deliberately and with the reasonable expectation that the death of the deceased or another would result,” Tex. Code Crim. Proc. Ann., Art. 37.071(b)(1) (Vernon 1981), because a capital murder conviction requires a finding that the defendant “intentionally or knowingly cause[d] the death of an individual,” see Tex. Penal Code Ann. § 19.02(a)(1) (Vernon 1974); see also § 19.03(a).
Among the many other studies reaching this conclusion are APA Task Force Report, Clinical Aspects of the Violent Individual 28 (1974) (90% error rate “[u]nfortunately ... is the state of the art”) (APA, Clinical Aspects); Steadman & Morrissey, The Statistical Prediction of Violent Behavior, 5 Law & Human Behavior 263, 271-273 (1981); Dix, Expert Prediction Testimony in Capital Sentencing: Evidentiary and Constitutional Considerations, 19 Am. Crim. L. Rev. 1, 16 (1981); Schwitzgebel, Prediction of Dangerousness and Its Implications for Treatment, in W. Curran, A. McGarry, & C. Petty, Modern Legal Medicine, Psychiatry, and Forensic Science 783, 784-786 (1980); Cocozza & Steadman, Prediction in Psychiatry: An Example of Misplaced Confidence in Experts, 25 Soc. Probs. 265, 272-273 (1978); Report of the (American Psychological Association’s) Task Force on the Role of Psychology in the Criminal Justice System, 33 Am. Psychologist 1099, 1110 (1978); Steadman & Cocozza, Psychiatry, Dangerousness and the Repetitively Violent Offender, 69 J. Crim. L. & Criminology 226; 227, 230 (1978); Cocozza & Steadman, The Failure of Psychiatric Predictions of Dangerousness: Clear and Convincing Evidence, 29 Rutgers L. Rev. 1084, 1101 (1976); Diamond, The Psychiatric Prediction of Dangerousness, 123 U. Pa. L. Rev. 439, 451-452 (1974); Ennis & Litwack, Psychiatry and the Presumption of Expertise: Flipping Coins in the Courtroom, 62 Calif. L. Rev. 693, 711-716 (1974). A relatively early study making this point is Rome, Identification of the Dangerous Offender, 42 F. R. D. 185 (1968).
See, e. g., APA, Clinical Aspects 28; 1 J. Ziskin, Coping with Psychiatric and Psychological Testimony 11, 19 (3d ed. 1981); Steadman & Morrissey, supra n. 2, at 264; Morse, Analysis of Mental Health Law, 51 S. Cal. L. Rev., at 599-600, 619-622; Cocozza & Steadman, supra n. 2, 25 Soc. Probs., at 274-275; Cocozza & Steadman, supra n. 2, 29 Rutgers L. Rev., at 1099-1100.
See APA Brief 14-16; APA, Clinical Aspects 25; J. Monahan, Clinical Prediction 86; Morse, Analysis of Mental Health Law, 51 S. Cal. L. Rev., at 598-600; Steadman & Cocozza, supra n. 2, 69 J. Crim. L. & Criminology, at 229-230; Diamond, supra n. 2, at 447.
That psychiatrists actually may be less accurate predictors of future violence than laymen, Ennis & Litwack, supra n. 2, at 734-735, may be due to personal biases in favor of predicting violence arising from the fear of being responsible for the erroneous release of a violent individual, see J. Monahan, Clinical Prediction 13, 22-25, 86; Morse, Analysis of Mental Health Law, 51 S. Cal. L. Rev., at 598-600. It also may be due to a tendency to generalize from experiences with past offenders on bases that have no empirical relationship to future violence, see Shah, Dangerousness: A Paradigm for Exploring Some Issues in Law and Psychology, American Psychologist 224, 229-230 (Mar. 1978), a tendency that may be present in Grigson’s and Holbrook’s testimony. Statistical prediction is clearly more reliable than clinical prediction, J. Monahan, Clinical Prediction 82; Steadman & Morrissey, supra n. 2, at 272 — and prediction based on statistics alone may be done by anyone, Morse, Analysis of Mental Health Law, 51 S. Cal. L. Rev., at 599-600; APA Brief 15-16.
Like the District Court, App. 13, and the Court of Appeals, id,., at 20, the Court seeks to justify the admission of psychiatric testimony on the ground that “ ‘[t]he majority of psychiatric experts agree that where there is a pattern of repetitive assaultive and violent conduct, the accuracy of psychiatric predictions of future dangerousness dramatically rises.’” Ante, at 902, quoting App. 13. The District Court correctly found that there *923is empirical evidence supporting the common-sense correlation between repetitive past violence and future violence; the APA states that “[t]he most that can be said about any individual is that a history of past violence increases the probability that future violence will occur.” Draft Report 29 (emphasis supplied). But psychiatrists have no special insights to add to this actuarial fact, and a single violent crime cannot provide a basis for a reliable prediction of future violence. APA, Clinical Aspects 23-24; see J. Monahan, Clinical Prediction 71-72; Steadman & Cocozza, supra n. 2, 69 J. Crim. L. & Criminology, at 229-230.
The lower courts and this Court have sought solace in this statistical correlation without acknowledging its obvious irrelevance to the facts of this case. The District Court did not find that the State demonstrated any pattern of repetitive assault and violent conduct by Barefoot. Recognizing the importance of giving some credibility to its experts’ specious prognostications, the State now claims that the “reputation” testimony adduced at the sentencing hearing “can only evince repeated, widespread acts of criminal violence.” Brief for Respondent 47. This is simply absurd. There was no testimony worthy of credence that Barefoot had committed acts of violence apart from the crime for which he was being tried; there was testimony only of a bad reputation for peaceable and law-abiding conduct. In light of the fact that each of Barefoot’s prior convictions was for a nonviolent offense, such testimony obviously could have been based on antisocial but nonviolent behavior. Neither psychiatrist informed the jury that he considered this reputation testimony to show a history of repeated acts of violence. Moreover, if the psychiatrists or the jury were to rely on such vague hearsay testimony in order to show a “pattern of repetitive assault and violent conduct,” Barefoot’s death sentence would rest on information that might “bear no closer relation to fact than the average rumor or item of gossip,” Gardner v. Florida, 430 U. S. 349, 359 (1977), and should be invalid for that reason alone. A death sentence cannot rest on highly dubious predictions secretly based on a factual foundation of hearsay and pure conjecture. See ibid.
Although I believe that the misleading nature of any psychiatric prediction of future violence violates due process when introduced in a capital sentencing hearing, admitting the predictions in this case — which were made without even examining the defendant — was particularly indefensi*924ble. In the APA’s words, if prediction following even an in-depth examination is inherently unreliable,
“there is all the more reason to shun the practice of testifying without having examined the defendant at all. . . . Needless to say, responding to hypotheticals is just as fraught with the possibility of error as testifying in any other way about an individual whom one has not personally examined. Although the courts have not yet rejected the practice, psychiatrists should.” Draft Report 32-33.
Such testimony is offensive not only to legal standards; the APA has declared that “[i]t is unethical for a psychiatrist to offer a professional opinion unless he/she has conducted an examination.” The Principles of Medical Ethics, With Annotations Especially Applicable to Psychiatry § 7(3), p. 9 (1981); see Opinions of the Ethics Committee on the Principles of Medical Ethics, With Annotations Especially Applicable to Psychiatry, p. 27 (1983). The Court today sanctions admission in a capital sentencing hearing of “expert” medical testimony so unreliable and unprofessional that it violates the canons of medical ethics.
Cf. Stein v. New York, 346 U. S. 156, 192 (1953) (prior to application of Fifth Amendment to the States, “reliance on a coerced confession vitiate[d] a [state] conviction because such a confession combines the persuasiveness of apparent conclusiveness with what judicial experience shows to be illusory and deceptive evidence”).
There can be no dispute about this obvious proposition:
“Scientific evidence impresses lay jurors. They tend to assume it is more accurate and objective than lay testimony. A juror who thinks of scientific evidence visualizes instruments capable of amazingly precise measurement, of findings arrived at by dispassionate scientific tests. In short, in the mind of the typical lay juror, a scientific witness has a special aura of credibility.” Imwinkelried, Evidence Law and Tactics for the Proponents of Scientific Evidence, in Scientific and Expert Evidence 33, 37 (E. Imwin-kelried ed. 1981).
See 22 C. Wright & K. Graham, Federal Practice and Procedure § 5217, p. 295 (1978) (“Scientific... evidence has great potential for misleading the jury. The low probative worth can often be concealed in the jargon of *927some expert. . .”). This danger created by use of scientific evidence frequently has been recognized by the courts. Speaking specifically of psychiatric predictions of future dangerousness similar to those at issue, one District Court has observed that when such a prediction “is proffered by a witness bearing the title of ‘Doctor,’ its impact on the jury is much greater than if it were not masquerading as something it is not.” White v. Estelle, 554 F. Supp. 851, 858 (SD Tex. 1982). See Note — People v. Murtishaw: Applying the Frye Test to Psychiatric Predictions of Dangerousness in Capital Cases, 70 Calif. L. Rev. 1069, 1076-1077 (1982). In United States v. Addison, 162 U. S. App. D. C. 199, 202, 498 F. 2d 741, 744 (1974), the court observed that scientific evidence may “assume a posture of mystic infallibility in the eyes of a jury of laymen.” Another court has noted that scientific evidence “is likely to be shrouded with an aura of near infallibility, akin to the ancient oracle of Delphi.” United States v. Alexander, 526 F. 2d 161, 168 (CA8 1975). See United States v. Amaral, 488 F. 2d 1148, 1152 (CA9 1973); United States v. Wilson, 361 F. Supp. 510, 513 (Md. 1973); People v. King, 266 Cal. App. 2d 437, 461, 72 Cal. Rptr. 478, 493 (1968).
Other purportedly scientific proof has met a similar fate. See, e. g., United States v. Kilgus, 571 F. 2d 508, 510 (CA9 1978) (expert testimony identifying aircraft through “forward looking infrared system” inadmissible because unreliable and not generally accepted in scientific field to which it belongs); United States v. Brown, 557 F. 2d 541, 558-559 (CA6 1977) (expert identification based on “ion microprobic analysis of human hair” not admissible because insufficiently reliable and accurate, and not accepted in its field); United States v. Addison, 162 U. S. App. D. C., at 203, 498 F. 2d, at 745 (expert identification based on voice spectrogram inadmissible because not shown reliable); United States v. Hearst, 412 F. Supp. 893, 895 (ND Cal. 1976) (identification testimony of expert in “psycholinguistics” inadmissible because not demonstrably reliable), aff’d on other grounds, 563 F. 2d 1331 (CA9 1977).
The Court observes that this well-established rule is a matter of evidence law, not constitutional law. Ante, at 899, n. 6. But the principle requiring that capital sentencing procedures ensure reliable verdicts, see swpra, at 923-926, which the Court ignores, and the principle that due process is violated by the introduction of certain types of seemingly conclusive, but actually unreliable, evidence, see supra, at 925, and n. 7, which the Court also ignores, are constitutional doctrines of long standing. The teaching of the evidence doctrine is that unreliable scientific testimony creates a serious and unjustifiable risk of an erroneous verdict, and that the adversary process at its best does not remove this risk. We should not dismiss this lesson merely by labeling the doctrine nonconstitutional; its relevance to the constitutional question before the Court could not be more certain.
In one study, for example, the only factor statistically related to whether psychiatrists predicted that a subject would be violent in the future was the type of crime with which the subject was charged. Yet the defendant’s charge was mentioned by the psychiatrists to justify their predictions in only one third of the cases. The criterion most frequently cited was “delusional or impaired thinking.” Cocozza & Steadman, supra n. 2, 29 Rutgers L. Rev., at 1096.
For one thing, although most members of the mental health professions believe that such predictions cannot be made, defense lawyers may experience significant difficulties in locating effective rebuttal witnesses. Davis, Texas Capital Sentencing Procedures: The Role of the Jury and the Restraining Hand of the Expert, 69 J. Crim. L. & Criminology 300, 302 (1978). I presume that the Court’s reasoning suggests that, were a defendant to show that he was unable, for financial or other reasons, to obtain an adequate rebuttal expert, a constitutional violation might be found.
“Although jurors may treat mitigating psychiatric evidence with skepticism, they may credit psychiatric evidence demonstrating aggravation. Especially when jurors’ sensibilities are offended by a crime, they may seize upon evidence of dangerousness to justify an enhanced sentence.” Dix, supra n. 2, at 43, n. 215. Thus, the danger of jury deference to expert opinions is particularly acute in death penalty cases. Expert testimony of this sort may permit juries to avoid the difficult and emotionally *935draining personal decisions concerning rational and just punishment. Id., at 46. Doctor Grigson himself has noted both the superfluousness and the misleading effect of his testimony:
“ T think you could do away with the psychiatrist in these cases. Just take any man off the street, show him what the guy’s done, and most of these things are so clearcut he would say the same things I do. But I think the jurors feel a little better when a psychiatrist says it — somebody that’s supposed to know more than they know.’ ” Bloom, Killers and Shrinks, Texas Monthly 64, 68 (July 1978) (quoting Doctor Grigson).
The Court is far wide of the mark in asserting that excluding psychiatric predictions of future dangerousness from capital sentencing proceedings “would immediately call into question those other contexts in which predictions of future behavior are constantly made.” Ante, at 898. Short-term predictions of future violence, for the purpose of emergency commitment or treatment, are considerably more accurate than long-term predictions. See APA Brief 12, n. 7; Monahan, Prediction Research and the Emergency Commitment of Dangerous Mentally Ill Persons: A Reconsideration, 135 Am. J. Psychiatry 198 (1978); J. Monahan, Clinical Prediction 59-60; Schwitzgebel, supra n. 2, at 786. In other contexts where psychiatric predictions of future dangerousness are made, moreover, the subject will not be criminally convicted, much less put to death, as a result of predictive error. The risk of error therefore may be shifted to the defendant to some extent. See Addington v. Texas, 441 U. S. 418, 423-430 (1979). The APA, discussing civil commitment proceedings based on determinations of dangerousness, states that in light of the unreliability of psychiatric predictions, “[cjlose monitoring, frequent follow-up, and a willingness to change one’s mind about treatment recommendations and dispositions for violent persons, whether within the legal system or without, is the only acceptable practice if the psychiatrist is to play a helpful role in these assessments of dangerousness.” APA, Clinical Aspects 30 (emphasis supplied). In a capital case there will be no chance for “follow-up” or “monitoring.” A subsequent change of mind brings not justice delayed, but the despair of irreversible error. See Bonnie & Slobogin, The Role of Mental Health Professionals in the Criminal Process: The Case for Informed Speculation, 66 Va. L. Rev. 427, 442-446 (1980).
The Court’s focus in the death penalty cases has been primarily on ensuring a fair procedure:
“In ensuring that the death penalty is not meted out arbitrarily or capriciously, the Court’s principal concern has been more with the 'procedure by which the State imposes the death sentence than with the substantive factors the State lays before the jury as a basis for imposing death, once it has been determined that the defendant falls within the category of persons eli*938gible for the death penalty.” California v. Ramos, post, at 999 (emphasis in original).