with whom Justice Brennan joins, dissenting.
When the Government withholds from a defendant evidence that might impeach the prosecution’s only witnesses, that failure to disclose cannot be deemed harmless error. Because that is precisely the nature of the undisclosed evidence in this case, I would affirm the judgment of the Court of Appeals and would not remand for further proceedings.
H
The federal grand jury indicted the respondent, Hughes Anderson Bagley, on charges involving possession of fire*686arms and controlled substances with intent to distribute. Following a bench trial, Bagley was found not guilty of the firearms charges, guilty of two counts of knowingly and intentionally distributing Valium, and guilty of several counts of a lesser included offense of possession of controlled substances. He was sentenced to six months’ imprisonment and a special parole term of five years on the first count of distribution, and to three years of imprisonment, which were suspended, and five years’ probation, on the second distribution count. He received a suspended sentence and five years’ probation for the possession convictions.
The record plainly demonstrates that on the two counts for which Bagley received sentences of imprisonment, the Government’s entire case hinged on the testimony of two private security guards who aided the Bureau of Alcohol, Tobacco and Firearms (ATF) in its investigation of Bagley. In 1977 the two guards, O’Connor and Mitchell, worked for the Milwaukee Railroad; for about three years, they had been social acquaintances of Bagley, with whom they often shared coffee breaks. 7 Tr. 2-3; 8 Tr. 2a-3a. At trial, they testified that on two separate occasions they had visited Bagley at his home, where Bagley had responded to O’Connor’s complaint that he was extremely anxious by giving him Valium pills. In total, Bagley received $8 from O’Connor, representing the cost of the pills. At trial, Bagley testified that he had a prescription for the Valium because he suffered from a bad back, 14 Tr. 963-964. No testimony to the contrary was introduced. O’Connor and Mitchell each testified that they had worn concealed transmitters and body recorders at these meetings, but the tape recordings were insufficiently clear to be admitted at trial and corroborate their testimony.
Before trial, counsel for Bagley had filed a detailed discovery motion requesting, among other things, “any deals, promises or inducements made to witnesses in exchange for their testimony.” App. 17-19. In response to the discovery request, the Government had provided affidavits sworn by *687O’Connor and Mitchell that had been prepared during their investigation of Bagley. Each affidavit recounted in detail the dealings the witnesses had had with Bagley and closed with the declaration, “I made this statement freely and voluntarily without any threats or rewards, or promises of reward having been made to me in return for it.” Brief for United States 3, quoting Memorandum of Points and Authorities in Support of Pet. for Habeas Corpus, CV80-3592-RJK(M) (CD Cal.) Exhibits 1-9. Both of these agents testified at trial thereafter, and the Government did not disclose the existence of any deals, promises, or inducements. Counsel for Bagley asked O’Connor on cross-examination whether he was testifying in response to pressure or threats from the Government about his job, and O’Connor said he was not. 7 Tr. 89-90. In light of the affidavits, as well as the prosecutor’s silence as to the existence of any promises, deals, or inducements, counsel did not pursue the issue of bias of either guard.
As it turns out, however, in May 1977, seven months prior to trial, O’Connor and Mitchell each had signed an agreement providing that ATF would pay them for information they provided. The form was entitled “Contract for Purchase of Information and Payment of Lump Sum Therefor,” and provided that the Bureau would, “upon the accomplishment of the objective sought to be obtained . . . pay to said vendor a sum commensurate with services and information rendered.” App. 22-23. It further invited the Bureau’s special agent in charge of the investigation, Agent Prins, to recommend an amount to be paid after the information received had proved “worthy of compensation.” Agent Prins had personally presented these forms to O’Connor and Mitchell for their signatures. The two witnesses signed the last of their affidavits, which declared the absence of any promise of reward, the day after they signed the ATF forms. After trial, Agent Prins requested that O’Connor and Mitchell each be paid $500, but the Bureau reduced these “rewards” to $300 each. App. to *688Pet. for Cert. 14a. The District Court Judge concluded that “it appears probable to the Court that O’Connor and Mitchell did expect to receive from the United States some kind of compensation, over and above their expenses, for their assistance, though perhaps not for their testimony.” Id., at 7a.
Upon discovering these ATF forms through a Freedom of Information Act request, Bagley sought relief from his conviction. The District Court Judge denied Bagley’s motion to vacate his sentence stating that because he was the same judge who had been the original trier of fact, he was able to determine the effect the contracts would have had on his decision, more than four years earlier, to convict Bagley. The judge stated that beyond a reasonable doubt the contracts, if disclosed, would have had no effect upon the convictions:
“The Court has read in their entirety the transcripts of the testimony of James P. O’Connor and Donald E. Mitchell at the trial .... Almost all of the testimony of both of those witnesses was devoted to the firearm charges in the indictment. The Court found the defendant not guilty of those charges. With respect to the charges against the defendant of distributing controlled substances and possessing controlled substances with the intention of distributing them, the testimony of O’Connor and Mitchell was relatively very brief. With respect to the charges relating to controlled substances cross-examination of those witnesses by defendant’s counsel did not seek to discredit their testimony as to the facts of distribution but rather sought to show that the controlled substances in question came from supplies which had been prescribed for defendant’s own use. As to that aspect of their testimony, the testimony of O’Connor and Mitchell tended to be favorable to the defendant.” Id., at 8a.
*689The foregoing statement, as to which the Court remands for further consideration, is seriously flawed on its face. First, the testimony that the court describes was in fact the only inculpatory testimony in the case as to the two counts for which Bagley received a sentence of imprisonment. • If, as the judge claimed, the testimony of the two information “vendors” was “very brief” and in part favorable to the defendant, that fact shows the weakness of the prosecutor’s case, not the harmlessness of the error. If the testimony that might have been impeached is weak and also cumulative, corroborative, or tangential, the failure to disclose the impeachment evidence could conceivably be held harmless. But when the testimony is the start and finish of the prosecution’s case, and is weak nonetheless, quite a different conclusion must necessarily be drawn.
Second, the court’s statement that Bagley did not attempt to discredit the witnesses’ testimony, as if to suggest that impeachment evidence would not have been used by the defense, ignores the realities of trial preparation and strategy, and is factually erroneous as well. Initially, the Government’s failure to disclose the existence of any inducements to its witnesses, coupled with its disclosure of affidavits stating that no promises had been made, would lead all but the most careless lawyer to step wide and clear of questions about promises or inducements. The combination of nondisclosure and disclosure would simply lead any reasonable attorney to believe that the witness could not be impeached on that basis. Thus, a firm avowal that no payment is being received in return for assistance and testimony, if offered at trial by a witness who is not even a Government employee, could be devastating to the defense. A wise attorney would, of necessity, seek an alternative defense strategy.
Moreover, counsel for Bagley in fact did attempt to discredit O’Connor, by asking him whether two ATF agents had pressured him or had threatened that his job might be in *690jeopardy, in order to get him to cooperate. 7 Tr. 89-90. But when O’Connor answered in the negative, ibid., counsel stopped this line of questioning. In addition, counsel for Bagley attempted to argue to the District Court, in his closing argument, that O’Connor and Mitchell had “fabricated” their accounts, 14 Tr. 1117, but the court rejected the proposition:
“Let me say this to you. I would find it hard to believe really that their testimony was fabricated. I think they might have been mistaken. You know, it is possible that they were mistaken. I really did not get the impression at all that either one or both of those men were trying at least in court here to make a case against the defendant.” Id., at 1117-1118. (Emphasis added.)
The District Court, in so saying, of course had seen no evidence to suggest that the two witnesses might have any motive for “mak[ing] a case” against Bagley. Yet, as Justice Blackmun points out, the possibility of a reward, the size of which is directly related to the Government’s success at trial, gave the two witnesses a “personal stake” in the conviction and an “incentive to testify falsely in order to secure a conviction.” Ante, at 683.
Nor is this case unique. Whenever the Government fails, in response to a request, to disclose impeachment evidence relating to the credibility of its key witnesses, the truth-finding process of trial is necessarily thrown askew. The failure to disclose evidence affecting the overall credibility of witnesses corrupts the process to some degree in all instances, see Giglio v. United States, 405 U. S. 150 (1972); Napue v. Illinois, 360 U. S. 264 (1959); United States v. Agurs, 427 U. S. 97, 121 (1976) (Marshall, J., dissenting), but when “the ‘reliability of a given witness may well be determinative of guilt or innocence,’” Giglio, supra, at 154 (quoting Napue, supra, at 269), and when “the Government’s case depend[s] almost entirely on” the testimony of a certain witness, 405 U. S.,_ at 154, evidence of that witness’ possible *691bias simply may not be said to be irrelevant, or its omission harmless. As The Chief Justice said in Giglio v. United States, in which the Court ordered a new trial in a case in which a promise to a key witness was not disclosed to the jury:
“[Without [Taliento’s testimony] there could have been no indictment and no evidence to carry the case to the jury. Taliento’s credibility as a witness was therefore an important issué in the case, and evidence of any understanding or agreement as to a future prosecution would be relevant to his credibility and the jury was entitled to know of it.
“For these reasons, the due process requirements enunciated in Napue and other cases cited earlier require a new trial.” Id., at 154-155.
Here, too, witnesses O’Connor and Mitchell were crucial to the Government’s case. Here, too, their personal credibility was potentially dispositive, particularly since the allegedly corroborating tape recordings were not audible. It simply cannot be denied that the existence of a contract signed by those witnesses, promising a reward whose size would depend “on the Government’s satisfaction with the end result,” ante, at 683, might sway the trier of fact, or cast doubt on the truth of all that the witnesses allege. In such a case, the trier of fact is absolutely entitled to know of the contract, and the defense counsel is absolutely entitled to develop his case with an awareness of it. Whatever the applicable standard of materiality, see infra, in this instance it undoubtedly is well met.
Indeed, Giglio essentially compels this result. The similarities between this case and that one are evident. In both cases, the triers of fact were left unaware of Government inducements to key witnesses. In both cases, the individual trial prosecutors acted in good faith when they failed to disclose the exculpatory evidence. See Giglio, supra, at 151— 153; App. to Pet. for Cert. 13a (Magistrate’s finding that *692Bagley prosecutor would have disclosed information had he known of it). The sole difference between the two cases lies in the fact that in Giglio, the prosecutor affirmatively stated to the trier of fact that no promises had been made. Here, silence in response to a defense request took the place of an affirmative error at trial — although the prosecutor did make an affirmative misrepresentation to the defense in the affidavits. Thus, in each case, the trier of fact was left unaware of powerful reasons to question the credibility of the witnesses. “[T]he truth-seeking process is corrupted by the withholding of evidence favorable to the defense, regardless of whether the evidence is directly contradictory to evidence offered by the prosecution.” Agurs, supra, at 120 (Marshall, J., dissenting). In this case, as in Giglio, a new trial is in order, and the Court of Appeals correctly reversed the District Court’s denial of such relief.
H — H
Instead of affirming, the Court today chooses to reverse and remand the case for application of its newly stated standard to the facts of this case. While I believe that the evidence at issue here, which remained undisclosed despite a particular request, undoubtedly was material under the Court’s standard, I also have serious doubts whether the Court’s definition of the constitutional right at issue adequately takes account of the interests this Court sought to protect in its decision in Brady v. Maryland, 373 U. S. 83 (1963).
A
I begin from the fundamental premise, which hardly bears repeating, that “[t]he purpose of a trial is as much the acquittal of an innocent person as it is the conviction of a guilty one.” Application of Kapatos, 208 F. Supp. 883, 888 (SDNY 1962); see Giles v. Maryland, 386 U. S. 66, 98 (1967) (Fortas, J., concurring in judgment) (“The State’s obligation is not to convict, but to see that, so far as possible, truth emerges”). When evidence favorable to the defendant is known to exist, *693disclosure only enhances the quest for truth; it takes no direct toll on that inquiry. Moreover, the existence of any small piece of evidence favorable to the defense may, in a particular case, create just the doubt that prevents the jury from returning a verdict of guilty. The private whys and wherefores of jury deliberations pose an impenetrable barrier to our ability to know just which piece of information might make, or might have made, a difference.
When the state does not disclose information in its possession that might reasonably be considered favorable to the defense, it precludes the trier of fact from gaining access to such information and thereby undermines the reliability of the verdict. Unlike a situation in which exculpatory evidence exists but neither the defense nor the prosecutor has uncovered it, in this situation the state already has, resting in its files, material that would be of assistance to the defendant. With a minimum of effort, the state could improve the real and apparent fairness of the trial enormously, by assuring that the defendant may place before the trier of fact favorable evidence known to the government. This proposition is not new. We have long recognized that, within the limit of the state’s ability to identify so-called exculpatory information, the state’s concern for a fair verdict precludes it from withholding from the defense evidence favorable to the defendant’s case in the prosecutor’s files. See, e. g., Pyle v. Kansas, 317 U. S. -213, 215-216 (1942) (allegation that imprisonment resulted from perjured testimony and deliberate suppression by authorities of evidence favorable to him “charge a deprivation of rights guaranteed by the Federal Constitution”).1
*694This recognition no doubt stems in part from the frequently considerable imbalance in resources between most criminal defendants and most prosecutors’ offices. Many, perhaps most, criminal defendants in the United States are represented by appointed counsel, who often are paid minimal wages and operate on shoestring budgets. In addition, unlike police, defense counsel generally is not present at the scene of the crime, or at the time of arrest, but instead comes into the case late. Moreover, unlike the government, defense counsel is not in the position to make deals with witnesses to gain evidence. Thus, an inexperienced, unskilled, or unaggressive attorney often is unable to amass the factual support necessary to a reasonable defense. When favorable evidence is in the hands of the prosecutor but not disclosed, the result may well be that the defendant is deprived of a fair chance before the trier of fact, and the trier of fact is deprived of the ingredients necessary to a fair decision. This grim reality, of course, poses a direct challenge to the traditional model of the adversary criminal process,2 and perhaps *695because this reality so directly questions the fairness of our longstanding processes, change has been cautious and halting. Thus, the Court has not gone the full road and expressly required that the state provide to the defendant access to the prosecutor’s complete files, or investigators who will assure that the defendant has an opportunity to discover every existing piece of helpful evidence. But cf. Ake v. Oklahoma, 470 U. S. 68 (1985) (access to assistance of psychiatrist constitutionally required on proper showing of need). Instead, in acknowledgment of the fact that important interests are served when potentially favorable evidence is disclosed, the Court has fashioned a compromise, requiring that the prosecution identify and disclose to the defendant favorable material that it possesses. This requirement is but a small, albeit important, step toward equality of justice.3
B
Brady v. Maryland, 373 U. S. 83 (1963), of course, established this requirement of disclosure as a fundamental element of a fair trial by holding that a defendant was denied due process if he was not given access to favorable evidence that is material either to guilt or punishment. Since Brady was decided, this Court has struggled, in a series of decisions, to define how best to effectuate the right recognized. To my mind, the Brady decision, the reasoning that underlay it, and the fundamental interest in a fair trial, combine to give the criminal defendant the right to receive from the prosecutor, and the prosecutor the affirmative duty to turn *696over to the defendant, all information known to the government that might reasonably be considered favorable to the defendant’s case. Formulation of this right, and imposition of this duty, are “the essence of due process of law. It is the State that tries a man, and it is the State that must insure that the trial is fair.” Moore v. Illinois, 408 U. S. 786, 809-810 (1972) (Marshall, J., concurring in part and dissenting in part). If that right is denied, or if that duty is shirked, however, I believe a reviewing court should not automatically reverse but instead should apply the harmless-error test the Court has developed for instances of error affecting constitutional rights. See Chapman v. California, 386 U. S. 18 (1967).
My view is based in significant part on the reality of criminal practice and on the consequently inadequate protection to the defendant that a different rule would offer. To implement Brady, courts must of course work within the confines of the criminal process. Our system of criminal justice is animated by two seemingly incompatible notions: the adversary model, and the state’s primary concern with justice, not convictions. Brady, of course, reflects the latter goal of justice, and is in some ways at odds with the competing model of a sporting event. Our goal, then, must be to integrate the Brady right into the harsh, daily reality of this apparently discordant criminal process.
At the trial level, the duty of the state to effectuate Brady devolves into the duty of the prosecutor; the dual role that the prosecutor must play poses a serious obstacle to implementing Brady. The prosecutor is by trade, if not necessity, a zealous advocate. He is a trained attorney who must aggressively seek convictions in court on behalf of a victimized public. At the same time, as a representative of the state, he must place foremost in his hierarchy of interests the determination of truth. Thus, for purposes of Brady, the prosecutor must abandon his role as an advocate and pore through his files, as objectively as possible, to identify the *697material that could undermine his case. Given this obviously unharmonious role, it is not surprising that these advocates oftentimes overlook or downplay potentially favorable evidence, often in cases in which there is no doubt that the failure to disclose was a result of absolute good faith. Indeed, one need only think of the Fourth Amendment’s requirement of a neutral intermediary, who tests the strength of the policeman-advocate’s facts, to recognize the curious status Brady imposes on a prosecutor. One telling example, offered by Judge Newman when he was a United States Attorney, suffices:
“I recently had occasion to discuss [Brady] at a PLI Conference in New York City before a large group of State prosecutors. ... I put to them this case: You are prosecuting a bank robbery. You have talked to two or three of the tellers and one or two of the customers at the time of the robbery. They have all taken a look at your defendant in a line-up, and they have said, ‘This is the man.’ In the course of your investigation you also have found another customer who was in the bank that day, who viewed the suspect, and came back and said, ‘This is not the man.’
“The question I put to these prosecutors was, do you believe you should disclose to the defense the name of the witness who, when he viewed the suspect, said ‘that is not the man’? In a room of prosecutors not quite as large as this group but almost as large, only two hands went up. There were only two prosecutors in that group who felt they should disclose or would disclose that information. Yet I was putting to them what I thought was the easiest case — the clearest case for disclosure of exculpatory information!” J. Newman, A Panel Discussion before the Judicial Conference of the Second Judicial Circuit (Sept. 8, 1967), reprinted in Discovery in Criminal Cases, 44 F. R. D. 481, 500-501 (1968) (hereafter Newman).
*698While familiarity with Brady no doubt has increased since 1967, the dual role that the prosecutor must play, and the very real pressures that role creates, have not changed.
The prosecutor surely greets the moment at which he must turn over Brady material with little enthusiasm. In perusing his files, he must make the often difficult decision as to whether evidence is favorable, and must decide on which side to err when faced with doubt. In his role as advocate, the answers are clear. In his role as representative of the state, the answers should be equally clear, and often to the contrary. Evidence that is of doubtful worth in the eyes of the prosecutor could be of inestimable value to the defense, and might make the difference to the trier of fact.
Once the prosecutor suspects that certain information might have favorable implications for the defense, either because it is potentially exculpatory or relevant to credibility, I see no reason why he should not be required to disclose it. After all, favorable evidence indisputably enhances the truth-seeking process at trial. And it is the job of the defense, not the prosecution, to decide whether and in what way to use arguably favorable evidence. In addition, to require disclosure of all evidence that might reasonably be considered favorable to the defendant would have the precautionary effect of assuring that no information of potential consequence is mistakenly overlooked. By requiring full disclosure of favorable evidence in this way, courts could begin to assure that a possibly dispositive piece of information is not withheld from the trier of fact by a prosecutor who is torn between the two roles he must play. A clear rule of this kind, coupled with a presumption in favor of disclosure, also would facilitate the prosecutor’s admittedly difficult task by removing a substantial amount of unguided discretion.
If a trial will thereby be more just, due process would seem to require such a rule absent a countervailing interest. I see little reason for the government to keep such information *699from the defendant. Its interest in nondisclosure at the trial stage is at best slight: the government apparently seeks to avoid the administrative hassle of disclosure, and to prevent disclosure of inculpatory evidence that might result in witness intimidation and manufactured rebuttal evidence.4 Neither of these concerns, however, counsels in favor of a rule of nondisclosure in close or ambiguous cases. To the contrary, a rule simplifying the disclosure decision by definition does not make that decision more complex. Nor does disclosure of favorable evidence inevitably lead to disclosure of inculpa-tory evidence, as might an open file policy, or to the anticipated wrongdoings of defendants and their lawyers, if indeed such fears are warranted. We have other mechanisms for disciplining unscrupulous defense counsel; hamstringing their clients need not be one of them. I simply do not find any state interest that warrants withholding from a presumptively innocent defendant, whose liberty is at stake in the proceeding, information that bears on his case and that might enable him to defend himself.
Under the foregoing analysis, the prosecutor’s duty is quite straightforward: he must divulge all evidence that reasonably appears favorable to the defendant, erring on the side of disclosure.
C
The Court, however, offers a complex alternative. It defines the right not by reference to the possible usefulness of the particular evidence in preparing and presenting the case, but retrospectively, by reference to the likely effect the evidence will have on the outcome of the trial. Thus, the Court holds that due process does not require the prosecutor to turn over evidence unless the evidence is “material,” and the *700Court states that evidence is “material” “only if there is a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different.” Ante, at 682. Although this looks like a post-trial standard of review, see, e. g., Strickland v. Washington, 466 U. S. 668 (1984) (adopting this standard of review), it is not. Instead, the Court relies on this review standard to define the contours of the defendant’s constitutional right to certain material prior to trial. By adhering to the view articulated in United States v. Agurs, 427 U. S. 97 (1976)—that there is no constitutional duty to disclose evidence unless nondisclosure would have a certain impact on the trial — the Court permits prosecutors to withhold with impunity large amounts of undeniably favorable evidence, and it imposes on prosecutors the burden to identify and disclose evidence pursuant to a pretrial standard that virtually defies definition.
The standard for disclosure that the Court articulates today enables prosecutors to avoid disclosing obviously exculpatory evidence while acting well within the bounds of their constitutional obligation. Numerous lower court cases provide examples of evidence that is undoubtedly favorable but not necessarily “material” under the Court’s definition, and that consequently would not have to be disclosed to the defendant under the Court’s view. See, e. g., United States v. Sperling, 726 F. 2d 69, 71-72 (CA2 1984) (prior statement disclosing motive of key Government witness to testify), cert. denied, 467 U. S. 1243 (1984); King v. Ponte, 717 F. 2d 635 (CA1 1983) (prior inconsistent statements of Government witness); see also United States v. Oxman, 740 F. 2d 1298, 1311 (CA3 1984) (addressing “disturbing” prosecutorial tendency to withhold information because of later opportunity to argue, with the benefit of hindsight, that information was not “material”), cert, pending sub nom. United States v. Pflaumer, No. 84-1033. The result is to veer sharply away from the basic notion that the fairness of a trial increases *701with the amount of existing favorable evidence to which the defendant has access, and to disavow the ideal of full disclosure.
The Court’s definition poses other, serious problems. Besides legitimizing the nondisclosure of clearly favorable evidence, the standard set out by the Court also asks the prosecutor to predict what effect various pieces of evidence will have on the trial. He must evaluate his case and the case of the defendant — of which he presumably knows very little — and perform the impossible task of deciding whether a certain piece of information will have a significant impact on the trial, bearing in mind that a defendant will later shoulder the heavy burden of proving how it would have affected the outcome. At best, this standard places on the prosecutor a responsibility to speculate, at times without foundation, since the prosecutor will not normally know what strategy the defense will pursue or what evidence the defense will find useful. At worst, the standard invites a prosecutor, whose interests are conflicting, to gamble, to play the odds, and to take a chance that evidence will later turn out not to have been potentially dispositive. One Court of Appeals has recently vented its frustration at these unfortunate consequences:
“It seems clear that those tests [for materiality] have a tendency to enc'ourage unilateral decision-making by prosecutors with respect to disclosure. . . . [T]he root of the problem is the prosecutor’s tendency to adopt a retrospective view of materiality. Before trial, the prosecutor cannot know whether, after trial, particular evidence will prove to have been material. .. . Following their adversarial instincts, some prosecutors have determined unilaterally that evidence will not be material and, often in good faith, have disclosed it neither to defense counsel nor to the court. If and when the evidence emerges after trial, the prosecutor can always argue, *702with the benefit of hindsight, that it was not material.” United States v. Oxman, supra, at 1310.
The Court’s standard also encourages the prosecutor to assume the role of the jury, and to decide whether certain evidence will make a difference. In our system of justice, that decision properly and wholly belongs to the jury. The prosecutor, convinced of the guilt of the defendant and of the truthfulness of his witnesses, may all too easily view as irrelevant or unpersuasive evidence that draws his own judgments into question. Accordingly he will decide the evidence need not be disclosed. But the ideally neutral trier of fact, who approaches the case from a wholly different perspective, is by the prosecutor’s decision denied the opportunity to consider the evidence. The reviewing court, faced with a verdict of guilty, evidence to support that verdict, and pressures, again understandable, to finalize criminal judgments, is in little better position to review the withheld evidence than the prosecutor.
I simply cannot agree with the Court that the due process right to favorable evidence recognized in Brady was intended to become entangled in prosecutorial determinations of the likelihood that particular information would affect the outcome of trial. Almost a decade of lower court practice with Agurs convinces me that courts and prosecutors have come to pay “too much deference to the federal common law policy of discouraging discovery in criminal cases, and too little regard to due process of law for defendants.” United States v. Oxman, supra, at 1310-1311. Apparently anxious to assure that reversals are handed out sparingly, the Court has defined a rigorous test of materiality. Eager to apply the “materiality” standard at the pretrial stage, as the Court permits them to do, prosecutors lose sight of the basic principles underlying the doctrine. I would return to the original theory and promise of Brady and reassert the duty of the prosecutor to disclose all evidence in his files that might reasonably be considered favorable to the defendant’s case. No *703prosecutor can know prior to trial whether such evidence will be of consequence at trial; the mere fact that it might be, however, suffices to mandate disclosure.5
*704D
In so saying, I recognize that a failure to divulge favorable information should not result in reversal in all cases. It may be that a conviction should be affirmed on appeal despite the prosecutor’s failure to disclose evidence that reasonably might have been deemed potentially favorable prior to trial. The state’s interest in nondisclosure at trial is minimal, and should therefore yield to the readily apparent benefit that full disclosure would convey to the search for truth. After trial, however, the benefits of disclosure may at times be tempered by the state’s legitimate desire to avoid retrial when error has been harmless. However, in making the determination of harmlessness, I would apply our normal constitutional error test and reverse unless it is clear beyond a reasonable doubt that the withheld evidence would not have affected the outcome of the trial. See Chapman v. California, 386 U. S. 18 (1967); see also Agurs, 427 U. S., at 119-120 (Marshall, J., dissenting).6
*705Any rule other than automatic reversal, of course, dilutes the Brady right to some extent and offers the prosecutor an incentive not to turn over all information. In practical effect, it might be argued, there is little difference between the rule I propose — that a prosecutor must disclose all favorable evidence in his files, subject to harmless-error review — and the rule the Court adopts —that the prosecutor must disclose only the favorable information that might affect the outcome of the trial. According to this argument, if a constitutional right to all favorable evidence leads to reversal only when the withheld evidence might have affected the outcome of the trial, the result will be the same as with a constitutional right only to evidence that will affect the trial outcome. See Capra, Access to Exculpatory Evidence: Avoiding the Agurs Problems of Prosecutorial Discretion and Retrospective Review, 53 Ford. L. Rev. 391, 409-410, n. 117 (1984). For several reasons, however, I disagree. First, I have faith that a prosecutor would treat a rule requiring disclosure of all information of a certain kind differently from a rule requiring disclosure only of some of that information. Second, persistent or egregious failure to comply with the constitutional duty could lead to disciplinary actions by the courts. Third, the standard of harmlessness I adopt is more protective of the defendant than that chosen by the Court, placing the burden on the prosecutor, rather than the defendant, to prove the harmlessness of his actions. It would be a foolish prosecutor who gambled too glibly with that standard of review. And finally, it is unrealistic to ignore the fact that at the appellate stage the state has an interest in avoiding retrial where the error is harmless beyond a reasonable doubt. That interest counsels against requiring a new trial in every case.
*706Thus, while I believe that some review for harmlessness is in order, I disagree with the Court’s standard, even were it merely a standard for review and not a definition of “materiality.” First, I see no significant difference for truth-seeking purposes between the Giglio situation and this one; for the same reasons I believe the result must therefore be the same here as in Giglio, see supra, at 691-692, I also believe the standard for reversal should be the same. The defendant’s entitlement to a new trial ought to be no different in the two cases, and the burden he faces on appeal should also be the same. Giglio remains the law for a class of cases, and I reaffirm my belief that the same standard applies to this case as well. See Agurs, supra, at 119-120 (Marshall, J., dissenting).
Second, only a strict appellate standard, which places on the prosecutor a burden to defend his decisions, will remove the incentive to gamble on a finding of harmlessness. Any lesser standard, and especially one in which the defendant bears the burden of proof, provides the prosecutor with ample room to withhold favorable evidence, and provides a reviewing court with a simple means to affirm whenever in its view the correct result was reached. This is especially true given the speculative nature of retrospective review:
“The appellate court’s review of ‘what might have been’ is extremely difficult in the context of an adversarial system. Evidence is not introduced in a vacuum; rather, it is built upon. The absence of certain evidence may thus affect the usefulness, and hence the use, of other evidence to which defense counsel does have access. Indeed, the absence of a piece of evidence may affect the entire trial strategy of defense counsel.” Capra, supra, at 412.
As a consequence, the appellate court no less than the prosecutor must substitute its judgment for that of the trier of fact under an inherently slippery test. Given such factors as a reviewing court’s natural inclination to affirm a judgment *707that appears “correct” and that court’s obvious inability to know what a jury ever will do, only a strict and narrow test that places the burden of proof on the prosecutor will begin to prevent affirmances in cases in which the withheld evidence might have had an impact.
Even under the most protective standard of review, however, courts must be careful to focus on the nature of the evidence that was not made available to the defendant and not simply on the quantity of the evidence against the defendant separate from the withheld evidence. Otherwise, as the Court today acknowledges, the reviewing court risks overlooking the fact that a failure to disclose has a direct effect on the entire course of trial.
Without doubt, defense counsel develops his trial strategy based on the available evidence. A missing piece of information may well preclude the attorney from pursuing a strategy that potentially would be effective. His client might consequently be convicted even though nondisclosed information might have offered an additional or alternative defense, if not pure exculpation. Under such circumstances, a reviewing court must be sure not to focus on the amount of evidence supporting the verdict to determine whether the trier of fact reasonably would reach the same conclusion. Instead, the court must decide whether the prosecution has shown beyond a reasonable doubt that the new evidence, if disclosed and developed by reasonably competent counsel, would not have affected the outcome of trial.7
*708In this case, it is readily apparent that the undisclosed information would have had an impact on the defense presented at trial, and perhaps on the judgment. Counsel for Bagley argued to the trial judge that the Government’s two key witnesses had fabricated their accounts of the drug distributions, but the trial judge rejected the argument for lack of any evidence of motive. See swpra, at 690. These key witnesses, it turned out, were each to receive monetary rewards whose size was contingent on the usefulness of their assistance. These rewards “served only to strengthen any incentive to testify falsely in order to secure a conviction.” Ante, at 683. To my mind, no more need be said; this non*709disclosure could not have been harmless. I would affirm the judgment of the Court of Appeals.
As early as 1807, this Court made clear that prior to trial a defendant must have access to impeachment evidence in the Government’s possession. Addressing defendant Aaron Burr’s claim that he should have access to the letter of General Wilkinson, a key witness against Burr in his trial for treason, Chief Justice Marshall wrote:
“The application of that letter to the case is shown by the terms in which the communication was made. It is a statement of the conduct of the *694accused made by the person who is declared to be the essential witness against him. The order for producing this letter is opposed:
“First, because it is not material to the defense. It is a principle, universally acknowledged, that a party has a right to oppose to the testimony of any witness against him, the declarations which that witness has made at other times on the same subject. If he possesses this right, he must bring forward proof of those declarations. This proof must be obtained before he knows positively what the witness will say; for if he waits until the witness has been heard at the trial, it is too late to meet him with his former declarations. Those former declarations, therefore, constitute a mass of testimony, which a party has a right to obtain by way of precaution, and the positive necessity of which can only be decided at the trial.” United States v. Burr, 25 F. Cas. 30, 36 (No. 14,692d) (CC Va. 1807).
See Fortas, The Fifth Amendment: Nemo Tenetur Prodere Seipsum, 26 Clev. B. A. J. 91, 98 (1964) (“The state and [the defendant] could meet, as the law contemplates, in adversary trial, as equals — strength against strength, resource against resource, argument against argument”); see also Babcock, Fair Play: Evidence Favorable to an Accused and Effective Assistance of Counsel, 34 Stan. L. Rev. 1133, 1142-1145 (1982) (discussing challenge Brady poses to traditional adversary model).
Indeed, this Court’s recent decision stating a stringent standard for demonstrating ineffective assistance of counsel makes an effective Brady right even more crucial. Without a real guarantee of effective counsel, the relative abilities of the state and the defendant become even more skewed, and the need for a minimal guarantee of access to potentially favorable information becomes significantly greater. See Strickland v. Washington, 466 U. S. 668 (1984); id., at 712-715 (MARSHALL, J., dissenting); Babcock, supra, at 1163-1174 (discussing the interplay between the right to Brady material and the right to effective assistance of counsel).
See Newman, 44 F. R. D., at 499 (describing the “serious” problem of witness intimidation that arises from prosecutor’s disclosure of witnesses). But see Brennan, The Criminal Prosecution: Sporting Event or Quest for Truth?, 1963 Wash. U. L. Q. 279, 289-290 (disputing a similar argument).
Brady not only stated the rule that suppression by the prosecution of evidence favorable to the defendant “violates due process where the evidence is material either to guilt or to punishment,” 373 U. S., at 87, but also observed that two decisions of the Court of Appeals for the Third Circuit “state the correct constitutional rule.” Id,., at 86. Neither of those decisions limited the right only to evidence that is “material” within the meaning that the Court today articulates. Instead, they provide strong evidence that Brady might have used the word in its evidentiary sense, to mean, essentially, germane to the points at issue.
In United States ex rel. Almeida v. Baldi, 195 F. 2d 815 (CA3 1952), cert. denied, 345 U. S. 904 (1953), the appeals court granted a petition for habeas corpus in a ease in which the State had withheld from the defendant evidence that might have mitigated his punishment. After describing the withheld evidence as “relevant” and “pertinent,” 195 F. 2d, at 819, the court concluded: “We think that the conduct of the Commonwealth as outlined in the instant case is in conflict with our fundamental principles of liberty and justice. The suppression of evidence favorable to Almeida was a denial of due process.” Id., at 820. Similarly, in United States ex rel. Thompson v. Dye, 221 F. 2d 763, 765 (CA3), cert. denied, 350 U. S. 875 (1955), the District Court had denied a petition for habeas corpus after finding that certain evidence of defendant’s drunkenness at the time of the offense in question was not “vital” to the defense and did not require disclosure. 123 F. Supp. 759, 762 (WD Pa. 1954). The Court of Appeals reversed, observing that whether or not the jury ultimately would credit the evidence at issue, the evidence was substantial and the State’s failure to disclose it cannot “be held as a matter of law to be unimportant to the defense here.” 221 F. 2d, at 767.
It is clear that the term “material” has an evidentiary meaning quite distinct from that which the Court attributes to it. Judge Weinstein, for example, defines as synonymous the words “ultimate fact,” “operative fact,” “material fact,” and “consequential fact,” each of which, he states, means “a ‘fact that is of consequence to the determination of the action.’” 1 J. Weinstein & M. Berger, Weinstein’s Evidence ¶401[03], n. 1 (1982) (quoting Fed. Rule Evid. 401). Similarly, another treatise on evidence explains that there are two components to relevance — materiality and probative value. “Materiality looks to the relation between the propositions for which the evidence is offered and the issues in the case. If the evidence is offered to help prove a proposition which is not a matter in issue, *704the evidence is immaterial.” E. Cleary, McCormick on Evidence § 185 (3d ed. 1984). “Probative value” addresses the tendency of the evidence to establish a “material” proposition. Ibid. See also 1 J. Wigmore, Evidence § 2 (P. Tillers rev. 1982). There is nothing in Brady to suggest that the Court intended anything other than a rule that favorable evidence need only relate to a proposition at issue in the case in order to merit disclosure.
Even if the Court did not use the term “material” simply to refer to favorable evidence that might be relevant, however, I still believe that due process requires that prosecutors have the duty to disclose all such evidence. The inherent difficulty in applying, prior to trial, a definition that relates to the outcome of the trial, and that is based on speculation and not knowledge, means that a considerable amount of potentially consequential material might slip through the Court’s standard. Given the experience of the past decade with Agurs, and the practical problem that inevitably exists because the evidence must be disclosed prior to trial to be of any use, I can only conclude that all potentially favorable evidence must be disclosed. Of course, I agree with courts that have allowed exceptions to this rule on a showing of exigent circumstances based on security and law enforcement needs.
In a case of deliberate prosecutorial misconduct, automatic reversal might well be proper. Certain kinds of constitutional error so infect the *705system of justice as to require reversal in all eases, such as discrimination in jury selection. See, e. g., Peters v. Kiff, 407 U. S. 493 (1972). A deliberate effort of the prosecutor to undermine the search for truth clearly is in the category of offenses antithetical to our most basic vision of the role of the state in the criminal process.
For example, in United States ex rel. Butler v. Maroney, 319 F. 2d 622 (CA3 1963), the defendant was convicted of first-degree murder. Trial counsel based his defense on temporary insanity at the time of the murder. During trial, testimony suggested that the shooting might have been the accidental result of a struggle, but defense counsel did not develop that defense. It later turned out that an eyewitness to the shooting had given police a statement that the victim and Butler had struggled prior to the murder. If defense counsel had known before trial what the eyewitness had seen, he might have relied on an additional defense, and he might have emphasized the struggle. See Note, The Prosecutor’s Constitutional *708Duty to Reveal Evidence to the Defendant, 74 Yale L. J. 136, 145 (1964). Unless the same information already was known to counsel before trial, the failure to disclose evidence of that kind simply cannot be harmless because reasonably competent counsel might have utilized it to yield a different outcome. No matter how overwhelming the evidence that Butler committed the murder, he had a right to go before a trier of fact and present his best available defense.
Similarly, in Ashley v. Texas, 319 F. 2d 80 (CA5), cert. denied, 375 U. S. 931 (1963), the defendant was sentenced to death for murder. The prosecutor disclosed to the defense a psychiatrist’s report indicating that the defendant was sane, but he failed to disclose the reports of a psychiatrist and a psychologist indicating that the defendant was insane. The non-disclosed information did not relate to the trial defense of self-defense. But the failure to disclose the evidence clearly prevented defense counsel from developing the possibly dispositive defense that he might have developed through further psychiatric examinations and presentation at trial. The nondiselosed evidence obviously threw off the entire course of trial preparation, and a new trial was in order. In such a case, there simply is no need to consider — in light of the evidence that actually was presented and the quantity of evidence to support the verdict returned — the possible effect of the information on the particular jury that heard the case. Indeed, to make such an evaluation would be to substitute the reviewing court’s judgment of the facts, including the previously undisclosed evidence, for that of the jury, and to do so without the benefit of competent counsel’s development of the information.
See also Field, Assessing the Harmlessness of Federal Constitutional Error — A Process in Need of a Rationale, 125 U. Pa. L. Rev. 15 (1976) (discussing application of harmless-error test).