dissenting.
I.
The appeal of William H. Pflaumer is before us under puzzling circumstances. Pflaumer was convicted on twenty-one counts of mail fraud and one count of conspiracy to commit mail fraud. When the case was first before us on direct appeal, we held that Pflaumer was entitled to a new trial for two independent reasons: the government’s improper withholding of specifically requested exculpatory materials and the district court’s erroneous jury instruction with respect to proof of membership in a conspiracy. See United States v. Oxman, 740 F.2d 1298, 1319 (3d Cir.1984), vacated and remanded sub nom. United States v. Pflaumer, — U.S. —, 105 S.Ct. 3550, 87 L.Ed.2d 673 (1985). Following that decision the United States petitioned for certiorari but did so only with respect to the first reason, see Petitioner’s Petition for a Writ of Certiorari at i, United States v. Pflaumer, as Pflaumer pointed out in his brief opposing the government’s petition, see Respondent’s Brief in Opposition at 2, United States v. Pflaumer. The government’s reply memorandum, filed on June 14, 1985, urged that the second ground for granting a new trial related only to the conspiracy charge on which Pflaumer was convicted, and not to the twenty-one substantive mail fraud counts. See Petitioner’s Reply Memorandum at 1-2, United States v. Pflaumer. Without briefing or argument, the Court, on July 2, 1985, entered an order that provides in full:
On WRIT OF CERTIORARI to the United States Court of Appeals for the Third Circuit
THIS CAUSE having been submitted on the petition for writ of certiorari and response thereto,
ON CONSIDERATION WHEREOF, it is ordered and adjudged by this Court that the judgment of the above court in this cause is vacated, and that this cause is remanded to the United States Court of Appeals for the Third Circuit for further consideration in light of United States v. Bagley, 473 U.S. — [105 S.Ct. 3375, 87 L.Ed.2d 481] (1985).
Certified Copy of Mandate, United States v. Pflaumer (received August 5, 1985).
Although the Court refers to the petition for writ of certiorari and “response thereto,” it does not mention the government’s reply memorandum. Thus it is not clear that the Justices even considered the reply memorandum or consciously addressed Pflaumer’s argument that independent grounds supported this court’s reversal of Pflaumer’s conviction. The Court simply remanded “for further consideration in light of United States v. Bagley.” The Bagley ease deals only with the government’s duties to produce exculpatory materials. Therefore the Court did not explicitly instruct us to reconsider our ruling that the erroneous jury instruction justified a new trial. Nevertheless, the Court vacated *1235our judgment, not our opinion, as unquestionably it was authorized to do. See 28 U.S.C. § 2106 (1982). The necessary effect of vacating the judgment is to deprive it of finality and leave open for reconsideration all issues previously decided, regardless of the Justices’ conscious intent.
.Equally puzzling is the Justices’ failure to note a significant difference between the record that was before the Court in Bagley and the one before it in Pflaumer’s case. Both cases involve post-conviction motions for a new trial based upon discoveries, made after trials, that the government had failed to disclose evidence that the defendants had specifically requested from the government before the trials and that they might have used to impeach government witnesses who testified at the trials. Bagley, however, was a bench trial, and the defendant there requested a new trial from the very judge who made the findings of fact that resulted in the challenged conviction. That judge obviously was familiar with his own deliberations. Pflaumer’s case, on the other hand, was a jury trial, and Federal Rule of Evidence 606(b) prohibited the judge to whom the new-trial motion was addressed from inquiring into the jury’s deliberative processes. This distinction is important because in Bagley the Court struggled — somewhat inconclusively as I note below — to establish an appropriate test for ascertaining whether withheld requested impeachment evidence is material to the challenged conviction.
None of the opinions that address the materiality issue in Bagley explain why the test for materiality should be identical in bench and jury trials. Possibly what the Supreme Court had in mind when it summarily vacated the judgment of this court “for further consideration in light of United States v. Bagley” was that we consider whether, in light of the strictures of Rule 606(b), the test for materiality in the nonjury context would be workable in the jury trial context. Possibly the Court, on the other hand, tacitly assumed that the two contexts were indistinguishable for purposes of materiality analysis.
Finally, the Bagley opinion itself is puzzling because of the form of the Court’s judgment. Although the district court had found beyond a reasonable doubt that, had the government not withheld the requested information, the trial’s outcome would have been the same, the Court “remand[ed] the case to [the court of appeals] for a determination whether there is a reasonable probability that, had the inducement offered by the Government to [witnesses] been disclosed to the defense, the result of the trial would have been different.” Bagley, 105 S.Ct. at 3385.1 Unfortunately, the Court does not discuss the scope of review a court of appeals is to employ in making such a determination; a determination that, depending upon whether the lower court trial is a jury trial or bench trial, obviously will require the consideration of different issues.
But while I am thus triply puzzled by the less than pellucid direction we have received from the Supreme Court, one thing seems clear to me: we must reconsider both grounds on which we awarded a new trial, at least to the extent of determining whether, independent of the issue of the undisclosed exculpatory material, the error in the jury instruction warrants a new trial on all counts. If we conclude that the error in the instruction does mandate a new trial on all counts, we simply would enter the same judgment as before. Therefore, it seems logical to me to consider that issue first.
*1236II.
For convenience the discussion of the erroneous jury instruction in the prior opinion of this court is quoted in the margin.2
We concluded that
[bjecause the trial court erred in rejecting Pflaumer’s request for a charge that the jury must find that he joined the *1237conspiracy prior to the last mailing, and because the government withheld speeifi-cally requested Brady material, his conviction must be reversed, and the case remanded for a new trial.
Oxman, 740 F.2d at 1319. Plainly our conjunctive reference to the two trial errors, coupled with our reversal of Pflau-mer’s conviction on all counts, indicates our belief that both errors afforded indepen*1238dent grounds for reversal on all counts. As indicated in Part I above, however, the necessary consequence of the Supreme Court’s judgment is that we must now reconsider our ruling with respect to the jury charge.
Insofar as the conspiracy count is concerned, its reconsideration need not detain us long. For the reasons set forth in our prior opinion, the charge violated the rule, set out in Grunewald v. United States, 353 U.S. 391, 406-15, 77 S.Ct. 963, 974-79, 1 L.Ed.2d 931 (1957), that, because a conspiracy to violate a federal statute ends when the unlawful object has been accomplished, evidence of post-event concealment of the conspiracy is inadmissible to show membership in the conspiracy. As we noted, the charge as given permitted the jury to rely on such evidence to show membership in the conspiracy, whereas it was only admissible to raise an inference of consciousness of guilt. See Oxman, 740 F.2d at 1304.
A separate question, not explicitly discussed in our prior opinion, is whether we can say with confidence that the erroneous instruction had no effect upon the substantive mail fraud counts on which Pflaumer was also convicted. In light of the district court’s instructions, I am unable to conclude that the error had no effect upon the jury’s deliberations on these counts.
The twenty-one substantive counts allege violations of 18 U.S.C. § 1341 (1982), which prohibits use of the mails to further a scheme to defraud. The government charged that Pflaumer was part of a scheme to defraud Pennsylvania, Maryland, and New Jersey of full excise and road use taxes properly due from a corporation, Wm. H.P., Inc. (WHP), that Pflau-mer owned. The theory of Pflaumer’s defense was that he knew nothing about the mail fraud scheme, but that Charles Gillan, President of WHP, was solely responsible for it.
The district court instructed the jury that it had to find the existence of a scheme to defraud, use of the mails in furtherance of that scheme, and participation by the defendant in the scheme with the requisite specific intent. See Joint Appendix at 1450. The court also explained that “where one does an act with knowledge that the use of the mails will follow in the ordinary course of business, or where such use can reasonably be foreseen, even though not actually intended, then he causes the mails to be used.” Id. at 1452. And the court stated that “[i]n a crime where two or more persons are charged with the commission of a crime, the guilt of any defendant may be established without proof that he personally did every act constituting the offense charged or that he personally participated at all in the offense charged.” Id. at 1456. The instruction continued,
In order to aid and abet another to commit a crime it is necessary that the particular person willfully associate himself in some way with the criminal venture and willfully participates [sic] in it as he should in something he wishes to bring about; that is to say, that he willfully seeks by some act of his to make the criminal venture succeed.
Id. at 1457.
The indictment, the government’s evidence, and the quoted instructions about the substantive mail fraud counts make it quite clear that what was charged was a scheme to defraud in which numerous persons participated. In the context of such a charge, we must consider whether the Grunewald error might have affected the jury’s deliberations over Pflaumer’s participation in that multi-party scheme, and we cannot divorce that consideration from the court’s conspiracy instruction, for common to both was the government’s burden of establishing Pflaumer’s participation in an ongoing enterprise involving other participants.
In its instruction on the conspiracy count, the district court told the jury,
One who willfully joins an existing conspiracy is charged with the same responsibility as if he had been one of the originators or instigators of the conspiracy. The guilt of a conspirator is not *1239governed by the extent or duration of his participation in the conspiracy or whether he had knowledge of all of its operations. Thus, one conspirator may play a major role, while others play a minor role or a minor part.
Even if one joined the conspiracy after it was formed and was engaged in it to a degree more limited than that of other co-conspirators, he is equally culpable so long as he was a co-conspirator.
In other words, it is not required that a person be a member of the conspiracy from its very start. He may join at any point during its progress, and he may be held responsible for all that has been done before he joins and all that may be done thereafter during its existence and while he remains a member. Simply stated, using the partnership analogy by becoming a partner he assumes all the liabilities of the partnership including those occurred [sic] before he became a member.
Joint Appendix at 1464 (emphasis supplied). Moreover, the court made references to overt acts alleged in the indictment, charging,
The Indictment alleges numerous overt acts in furtherance of each conspiracy. What must be shown is that after the defendant, whom you are considering, joined the conspiracy, some one of the conspirators performed an overt act in furtherance of the conspiracy.
Id. at 1467.
Because both the scheme to defraud and the conspiracy involved concert of action and because the conspiracy involved the same overt acts as did the scheme, I cannot rule out the likelihood that some jurors did not assume that the instruction that Pflau-mer “may be held responsible for all that has been done before he joins and all that may be done thereafter during its existence” was as applicable to the overt acts in furtherance of the scheme as to the overt acts in furtherance of the conspiracy.
Of course, co-conspirator statements made during and in furtherance of a conspiracy are admissible as proof of substantive offenses. In re Japanese Electronics, 723 F.2d 238, 263 (3d Cir.1983), cert. granted sub nom. Matsushita Electric Industrial Co. v. Zenith Radio Corp., — U.S. —, 105 S.Ct. 1863, 85 L.Ed.2d 157 (1985); United States v. Cahalane, 560 F.2d 601, 605 n. 4 (3d Cir.), cert. denied, 434 U.S. 1045, 98 S.Ct. 890, 54 L.Ed.2d 796 (1977). But the problem here is that, under Grunewald, proof of the post-conspiracy activities was inadmissible as proof of those offenses; at best, such evidence was admissible to prove Pflaumer’s consciousness of guilt. Yet, the district court gave no such limiting instruction, and thus we held that the jurors might have relied on that evidence to show membership in the common enterprise. See Oxman, 740 F.2d at 1304. Given the likelihood that that evidence, combined with the instruction that a conspirator becomes liable for all offenses committed by co-conspirators before that conspirator joined the conspiracy, tainted the jury’s finding that Pflaumer participated in the scheme, I cannot fairly exclude the possibility that the jury’s finding of liability for the scheme and the mailings in furtherance of its objects was affected by the Grunewald error.
Indeed, the charge that Pflaumer requested anticipated this problem, for it was addressed both to the conspiracy and the scheme to defraud.3 Moreover, the trial court, in denying Pflaumer’s post-trial mo*1240tion for a new trial, understood full well that the requested instruction was aimed at limiting the use to which evidence of conduct that occurred after the events charged in the indictment could be put with respect to membership in the conspiracy and participation in the multi-party scheme. See id. at 1665.
Considering the text of the requested instruction, the trial court’s treatment of it in the post-trial motion, and the absence of any analytical distinction between proof of membership in a conspiracy to violate the mail fraud statute and proof of membership in a group engaged jointly in a scheme to defraud, it is understandable that our prior opinion was not as explicit as it might have been in dealing with the effect of the Grünewald error on the mail fraud counts. And although Judge Sloviter’s majority opinion states that “our holding that Pflau-mer was entitled to a new trial on that basis applies only to count 22,” 774 F.2d at 1233, that was not the intention of the draftsman of the majority opinion. In any event, as the Supreme Court apparently intended, I have reconsidered and would still hold, this time explicitly, that the court should have given the requested instruction both with respect to Pflaumer’s membership in the conspiracy and with respect to his participation in the group carrying out the scheme to defraud.
III.
Because the judges in the majority hold that the Grünewald error could not have affected the jury’s deliberations on the mail fraud counts, they also address the significance of United States v. Bagley, — U.S. —, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985). As I have noted in Part I above, the remand in this ease for reconsideration in light of Bagley leaves for our solution several puzzles. Because I disagree with the solutions announced in Judge Sloviter’s opinion, I address this second issue, even though in my view the Grunewald error alone requires a new trial.
A.
Judge Sloviter observes that in Bagley the Supreme Court “clarified the legal standard for determining whether the government’s failure to disclose impeaching evidence warrants reversal of a' conviction.” 774 F.2d at 1226. The characterization of Bagley as a clarification of those standards is generous to a fault. In our prior opinion I set forth the tangled evolution of the rules on the prosecutor’s duty to disclose, from Pyle v. Kansas, 317 U.S. 213, 63 S.Ct. 177, 87 L.Ed. 214 (1942), through United States v. Agurs, 427 U.S. 97, 96 S.Ct. 2392, 49 L.Ed.2d 342 (1976), and discussed why that set of rules seemed designed to tempt prosecutors to withhold information and to encourage post-conviction litigation. See Oxman, 740 F.2d at 1306-11. Bagley certainly has changed those rules, but no one can say with a straight face that they have been “clarified.”
The Agurs rules, whatever their merit, were set forth in an opinion of the Court, while in Bagley the most significant issue — materiality of the undisclosed evidence — must be discerned from several opinions, none of which commanded a majority.
Unlike Agurs, Bagley and this case both deal with the prosecutor’s duty to disclose evidence that might be used only for impeachment purposes. In Agurs the defendant had not specifically requested the arguably exculpatory material — a criminal record indicating the witness’s propensity for violence — and the United States took the position that, absent a specific request, a prosecutor never had a duty to disclose material evidence favorable to the accused. See Brief for Petitioner at 24-33, United States v. Agurs.
The Court rejected this extreme position and announced three levels of scrutiny. First, in the absence of a specific request for the material at issue, the duty to disclose depends on the material’s character. If the undisclosed evidence “demonstrates that the prosecution’s case included perjured testimony and the prosecution knew, *1241or should have known of the perjury,” the defendant is entitled to a new trial if there is “any reasonable likelihood that the false testimony could have affected the judgment of the jury.” Agurs, 427 U.S. at 103, 96 S.Ct. at 2397. Even if no perjured testimony is involved, a new trial is nonetheless required if the undisclosed evidence “is so clearly supportive of a claim of innocence that it gives the prosecution notice of a duty to produce.” Id. at 112, 96 S.Ct. at 2401. Finally, if a specific request has been made, “if the subject matter of such a request is material, or indeed if a substantial basis for claiming materiality exists, it is reasonable to require the prosecutor to respond either by furnishing the information or by submitting the problem to the trial judge.” Id. at 106, 96 S.Ct. at 2399 (emphasis supplied).
Agurs was not a specific-request case, so the quoted language discussing the standard applicable to such cases may be characterized as dicta. However, the Court inserted that language in an opinion in a case in which the government argued for a much more radical rule, and the opinion of the Court did not question Justice Marshall’s statement that the Court had held “that the prosecutor’s constitutional duty to provide exculpatory evidence to the defense is not limited to cases in which the defense makes a request for such evidence.” Id. at 114, 96 S.Ct. at 2402 (Marshall, J., dissenting). Thus this court was justified in applying the “substantial basis for claiming materiality” test in United States v. McCrane, 547 F.2d 204, 207 (3d Cir.1976), and United States v. Higgs, 713 F.2d 39, 43 (3d Cir.1983), cert. denied, — U.S. —, 104 S.Ct. 725, 79 L.Ed.2d 185 (1984).
Our major difficulty with the Agurs formulation of the governing rules applicable in eases in which specific requests were made arose from the application of Justice Stevens’ “substantial basis for claiming materiality” standard. As I noted in our earlier opinion, the formulation inevitably tends to encourage nondisclosure, even of specifically-requested materials, by prosecutors hoping to succeed in post-trial litigation over materiality. See Oxman, 740 F.2d at 1310; see also United States v. Starusko, 729 F.2d 256, 262-63 (3d Cir.1984).
In our first opinion we attempted to give some workable content to the “substantial basis for claiming materiality” test for disclosure in the post-verdict context. The task was complex for several reasons. First, because the undisclosed information is unknown to defense counsel until after trial, the court ruling on a new trial motion inevitably has to speculate as to whether disclosure of the material would have led to the discovery of other relevant information and as to how defense counsel would have used it at trial. Second, the court also has to speculate as to how appropriate use of the material at trial would have affected the factfinder’s deliberations.
In such circumstances, as with all matters involving inference or speculation, allocation of the risk of nonpersuasion will usually be dispositive. We understood Agurs to have dealt with the allocation of this risk by establishing a threshold legal standard for materiality in specific-request cases—“substantial basis for claiming materiality”—that, when met, placed the burden upon the prosecution of satisfying the court that the error was harmless beyond a reasonable doubt. See Oxman, 740 F.2d at 1313 (citing Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 828, 17 L.Ed.2d 705 (1969), which recognized that the government carried the risk of nonpersuasion in this context). Thus we held that as a matter of law “the impeachment of an incriminating witness with significant evidence attacking the truthfulness of his testimony ‘might affect’ the jury’s assessment of reasonable doubt and thereby affect the outcome of the trial.” Id. We also held that the government failed to satisfy the Chapman harmless error test. See id. at 1317-19.
The dissent to our original opinion would have collapsed the inquiry into materiality and the Chapman burden of establishing harmless error in specific-request cases *1242into a single test of whether the evidence “might have” affected the jury’s verdict. See id. at 1321-22 (Sloviter, J., dissenting). The dissent apparently would have placed the risk of nonpersuasion on the “might have affected the verdict” issue upon the defendant. The majority observed that this approach “confuses the prospective Brady analysis which the prosecutor must apply before trial with the constitutionally harmless error determination which must be made after trial.” Id. at 1314.
B.
We must now discern what the Supreme Court has wrought in Bagley. Justice Blackmun speaks for a majority of the Court in Parts I and II of the Bagley opinion.
Part I merely recites the prior history of the case, but in the second part Justice Blackmun addresses the Court of Appeals’ decision concerning the prosecutor’s refusal to disclose to a defendant allegedly exculpatory material. The Court of Appeals for the Ninth Circuit, relying on Davis v. Alaska, 415 U.S. 308, 318, 94 S.Ct. 1105, 1111, 39 L.Ed.2d 347 (1974), had held that the prosecutor’s withholding of specifically-requested impeachment evidence could never satisfy the harmless error standard and thus required an automatic reversal. See Bagley v. Lumpkin, 719 F.2d 1462, 1464 (9th Cir.1983), rev’d and remanded sub nom. United States v. Bagley, — U.S. —, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985). That court’s theory was that the withheld information deprived the defendant of a constitutionally protected right of cross-examination. See id.
Justice Blackmun rejected that theory, holding that impeachment evidence and exculpatory evidence stand on the same level, constitutionally. See Bagley, 105 S.Ct. at 3377-79. In other words Justice Blackmun concluded that nondisclosure of either is subject to a harmless error analysis.4
Having rejected the Ninth Circuit rule that impeachment evidence was entitled to a higher degree of constitutional protection than exculpatory evidence, Justice Black-mun turned in Part III of Bagley to “the standard of materiality applicable to the nondisclosed evidence at issue.” Bagley, 105 S.Ct. at 3381-82. Although only Justice O’Connor joined that part of the opinion, Justice White, with whom the Chief Justice and Justice Rehnquist joined, approved of Justice Blackmun’s formulation of the standard of materiality of specifically-requested and withheld evidence. See id. at 3385 (White, J., concurring in part and concurring in the judgment). That formula is whether there is “a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different.” Id. at 3384; id. at 3385 (White, J., concurring in part and concurring in the judgment). Justice Blackmun and Justice O’Connor opined further that under this formulation
the reviewing court may consider directly any adverse effect that the prosecutor’s failure to respond might have had on the preparation or presentation of the defendant’s case. The reviewing court should assess the possibility that such effect might have occurred in light of the totality of the circumstances and with an awareness of the difficulty of reconstructing in a post-trial proceeding the course that the defense and the trial would have taken had the defense not been misled by the prosecutor's incomplete response.
Id. at 3384. The judgment of the Court, in which Justice White, the Chief Justice, and Justice Rehnquist join, “remand[s] the case to [the Court of Appeals] for a determination whether there is a reasonable probability that, had the inducement offered by the Government to [witnesses] been disclosed to the defense, the result of the trial would have been different.” Id. at 3385.
*1243The significance of five Justices joining in this judgment lies in the fact that all five agreed that the materiality determination shall be made, as Justice Blackmun put it, by “the reviewing court,” not by the court of first instance.5 While in making this determination we may take into account the trial court’s appraisal of the record, Bagley makes clear that the ultimate responsibility lies here.
Thus what the majority of the Supreme Court appears to have done in Bagley is to merge the standard of materiality, which must of course guide prosecutors in responding to requests for information and courts in pretrial rulings on requests, into the standard of review, which under Agurs was whether, assuming materiality as a matter of law, the error was harmless beyond a reasonable doubt. See id. at 3392-93 (Marshall, J., dissenting); id. at 3400 (Stevens, J., dissenting). Moreover, the “reasonable probability” standard of review is more favorable to the government than the “beyond a resonable doubt” standard of Chapman.
But, while it is clear that the assessment of the effect of the withholding of requested material still is to be made by the reviewing court and that the test for such review is no longer whether it is “able to declare a belief that it was harmless beyond a reasonable doubt,” Chapman, 386 U.S. at 24, 87 S.Ct. at 828, the several opinions in Bagley are silent as to the appropriate allocation of the risk of nonper-suasion. This is significant, for the extant rule was clear:
Certainly error, constitutional error, in illegally admitting highly prejudicial evidence or comments, casts on someone other than the person prejudiced by it the burden to show that it was harmless. It is for that reason that the original common-law harmless error rule put the burden on the beneficiary of the error either to prove that there was no injury or to suffer a reversal of his erroneously obtained judgment.
Id.; see Oxman, 740 F.2d at 1313-19.
Allocation of the risk of nonpersuasion is critical, for the Bagley “reasonable probability” standard, as a standard for review is hardly self-explanatory. In Chapman the Court wrote,
There is little, if any difference between our statement in Fahy v. Connecticut [375 U.S. 85, 86-87, 84 S.Ct. 229, 230, 11 L.Ed.2d 171 (1963) ] about “whether there is a reasonable possibility that the evidence complained of might have contributed to the conviction” and requiring the beneficiary of a constitutional error to prove beyond a reasonable doubt that the error complained of did not contribute to the verdict obtained.
386 U.S. at 24, 87 S.Ct. at 828. If there is little if any difference for purposes of review between “reasonable possibility” and “beyond a reasonable” doubt, there is hardly less between “reasonable probability” and “beyond a reasonable” doubt. In all cases where such formulae are used as standards for review, the outcome will usually turn upon which party must bear those risks inherent in speculation as to what may have gone on in the decisional process had the script been presented to the factfinder differently.
Mindful that in specific-request cases the prosecutor, acting ex parte, makes the materiality determination in the first instance, and thus creates the problem, I cannot read from the Justices’ silence an intent to change the settled law, which required that the government satisfy the reviewing court that, had the prosecutor disclosed the evidence before trial, the trial’s outcome would not have been different. It is true that in cases concerning prosecutors’ failure to disclose possible exculpatory materials we need no longer be convinced beyond a reasonable doubt that the outcome would have been the same, but we must still be persuaded by the government that no other *1244outcome was reasonably likely. The party that created the problem still bears the risk of nonpersuasion.
C.
Exercising the independent judgment that Bagley still requires, I am not satisfied that there is no reasonable probability that, had disclosure been made in this case, the outcome would have been the same. As I noted in our prior opinion, the evidence of Pflaumer’s involvement was less than overwhelming. See Oxman, 740 F.2d at 1315-16; 1317-18. Judge Sloviter notes that the jury was presented with substantial evidence of his guilt. See 774 F.2d at 1231. The fact remains, however, that Pflaumer also presented a defense. The gist of that defense was that the scheme to defraud was that of Gillan, the President of WHP, and that Frank Jock, his principal accuser, falsely implicated Pflaumer to secure Jock’s release from prison. To bolster Jock’s perhaps doubtful credibility, Pflau-mer argued, the government then bargained for additional supporting testimony. And in light of the revelation that Luciano, a witness supporting Jock’s implication of Pflaumer, also received lenient treatment in exchange for testimony, Pflaumer’s theory was clearly plausible. Had it been revealed that Ralph Wille, another witness who testified against Pflaumer, also received immunity, the theory would have been still more credible. The jury would then have noticed that every witness inculpating Pflaumer in any significant way had purchased lenient treatment.
Pflaumer’s vigorous impeachment of Jock, moreover, invited the jury to seek corroboration of Jock’s testimony from other witnesses. Wille, who testified that Pflaumer exerted day-to-day control over the affairs of WHP, provided the most significant corroboration, as his testimony afforded the jury a basis for inferring that Pflaumer must have known of the mail fraud scheme. Impeachment of Wille could well have undermined the jurors’ confidence in such an inference.
Judge Sloviter relies upon the fact that Wille received only a grant of use immunity, rather than the direct benefits received by Jock and Luciano. See 774 F.2d at 1231. Mindful of Justice Blackmun’s admonition that the reviewing court assess the possibility that prejudice to the presentation of the defendant’s case “might have occurred in light of the totality of the circumstances and with an awareness of the difficulty of reconstructing in a post-trial proceeding the course that the defense and the trial would have taken had the defense not been misled by the prosecutor’s incomplete response,” Bagley, 105 S.Ct. at 3384, I attach little significance to that distinction. Clearly, defense counsel would have argued to the jury that the greatest benefit the government had at its disposal was immunity from prosecution.
Thus I conclude that the difference between the Chapman harmless beyond a reasonable doubt standard and the Bagley reasonable probability of changing the outcome standard does not in this instance require a different result. The government has not persuaded me that, had it not withheld the information impeaching its only significant unimpeached witness, the outcome would with reasonable probability have been the same.
III.
Since it is likely that the Supreme Court will be given another opportunity to consider this case, it seems appropriate to suggest that from the perspective of a judge obliged to exercise the review function announced in Bagley the majority has created a monstrosity.
In the first place elimination of any distinction between the standard of materiality and the standard of review will have the inevitable effect of encouraging the prose-cutorial misconduct of which this court has so long complained. See, e.g., United States v. Starusko, 729 F.2d 256, 262-63 (3d Cir.1984); United States v. Kaplan, 554 F.2d 577, 581 (3d Cir.1977). The highest court has now instructed prosecutors to decide for themselves whether nondisclo*1245sure of requested information will produce a reasonable probability of a different result. It has, moreover, at the review stage, lessened the burden that they formerly bore of satisfying us beyond a reasonable doubt that such nondisclosure did not alter the outcome of the trial. This has reduced the risk of nondisclosure in three ways.
First, as we indicated in Starusko, such self-help by government attorneys is inconsistent with the Model Rules of Professional Conduct and should be referred to the attention of the appropriate disciplinary authorities. See Starusko, 729 F.2d at 264. Given the standard of materiality announced in Bagley, our disciplinary tiger is obviously toothless, for all prosecutors can plausibly urge in a disciplinary proceeding that they acted in the belief that the requested information would not produce a reasonable probability of a different result.
Further, the watered-down standard of materiality, by encouraging prosecutors to make their own determinations on the disclosure of information, will inevitably move the judicial resolution of disputes over discovery in criminal cases from the pretrial stage, where they properly should be resolved, to the post-trial stage, where their resolution will always be far more difficult.
And finally, by presenting the hard-pressed judges of the courts of appeal with still another standard of review without disclosing to us in some comprehensible manner how the new standard really differs from the old, the Court has compounded our difficulties. The differences between the majority and dissent as to the application of Bagley in this case are illustrative, but they are only the beginning. We had almost two decades of experience in applying the Chapman standard, and only rarely, in my experience, did it produce differences in opinion among us as to which errors required a new trial. The Bagley standard seems designed to produce such differences and thereby to increase the burden of opinion writing.
In short, whatever other concerns motivated the Court’s majority in Bagley, problems of judicial administration of the criminal justice system as a whole seem to have been entirely ignored.
IV.
Because the jury instruction error affected both the conspiracy and the multi-party scheme to defraud aspects of the case, a new trial on all counts is required. Because the impeaching information about Wille, withheld despite a specific request by the defendant, was on this record information that if disclosed would have presented a reasonable probability that the result would have been different, a new trial is required. I would order a new trial on all counts.
. Compare United States v. Agurs, 427 U.S. 97, 114, 96 S.Ct. 2392, 2402, 49 L.Ed.2d 342 (1976), a case in which the Supreme Court, while reviewing a jury conviction, relied on a district court’s appraisal of the record in ruling that nondisclosure of impeachment evidence not specifically requested did not warrant a new trial.
. Pflaumer contends that the trial court erred in refusing to give a proffered instruction as follows:
The indictment charges that the last tax return mailed as part of the alleged scheme and conspiracy was dated June 30, 1979. It was at that time, therefore, that the alleged conspiracy or scheme terminated. You may not find any defendant guilty merely on the basis of anything he did or said after June, 1979. This is because a conspirator cannot join a conspiracy or scheme after its attempts to achieve illicit objectives [have] ended____ In short, if you find that the government has not proved beyond a reasonable doubt that Mr. Pflaumer knowingly and willfully became a member of the alleged conspiracy or scheme to defraud Pennsylvania, New Jersey and Maryland of diesel fuel taxes prior to June, 1979, you must acquit Mr. Pflaumer of all charges.
App. at 1498. To understand this request one must appreciate what was charged in the indictment.
The first twenty-one counts alleged twenty-one separate mailings, each a separate substantive violation of 18 U.S.C. § 1341 (1982). The first mailing took place on April 3, 1978, the last on June 30, 1979. These mailings were alleged to violate section 1341 because they were for the purpose of executing a scheme or artifice, which arguably extended over a longer period, to defraud the three states. The scheme itself does not, however, violate federal law; only the use of the mailings to effectuate the scheme does so. See Kann v. United States, 323 U.S. 88, 94-95, 65 S.Ct. 148, 150-51, 89 L.Ed. 88 (1944); United States v. Tarnopol, 561 F.2d 466, 471 (3d Cir.1977).
The twenty-second count charges a conspiracy in violation of 18 U.S.C. § 371 (1982), which proscribes conspiracies “to commit any offense against the United States." The conspiracy charged in the indictment alleges:
a scheme and artifice to defraud the Commonwealth of Pennsylvania, the State of New Jersey, and the State of Maryland by means of false and fraudulent pretenses and representations, and for the purpose of executing such scheme and artifice to defraud and attempting so to do, to place in and cause to be placed in an authorized depository for mail matter, certain matter as described in Counts One through Twenty One, in violation of Title 18 United States Code, Section 1341____
App. at 17. As with the substantive section 1341 counts, the fraudulent scheme itself does not violate the federal conspiracy statute. The federal conspiracy must be to violate a federal substantive law — in this instance, the substantive prohibition against the use of the 21 mailings alleged in the first 21 counts.
Pflaumer’s point in making the quoted request for charge was straightforward. A conspiracy to violate a substantive prohibition in a federal statute ends when the unlawful object has been accomplished. Grunewald v. United States, 353 U.S. 391, 406-15, 77 S.Ct. 963, 974-79, 1 L.Ed.2d 931 (1957). The object of the conspiracy charged in the indictment was the use of the mails to effectuate a scheme to defraud three states of tax revenues. Although the scheme may have continued until 1981 — an issue on which we express no opinion — the use of the mails to effectuate the scheme ceased in 1979. Thus, the conspiracy charged in the indictment ended in June of 1979, and in order to convict the jury must have found that Pflaumer joined the conspiracy before that date.
From the viewpoint of Pflaumer’s defense, this issue was quite significant. There is evidence from which the jury could have found that when state agencies conducted audits of WHP after June of 1979, Pflaumer’s actions tended to conceal the scheme. This evidence was admitted at trial not to show Pflaumer’s membership in the conspiracy — that was prohibited by Grunewald—but to raise an inference of consciousness of guilt. See United States v. Mastropieri, 685 F.2d 776, 790-91 (2d Cir.), cert. denied, 459 U.S. 945, 103 S.Ct. 260, 74 L.Ed.2d 203 (1982); App. at 1665-66. The jury, however — not finely versed in these subtle distinctions — might have mistakenly applied evidence of concealment to show membership in the conspiracy. Hence Pflaumer requested an instruction expressly charging that the jury must, in order to convict, establish membership in the conspiracy based on evidence of acts before June of 1979.
The court refused the requested instruction. The government defends that refusal on several grounds. One is that post-June 30, 1979 efforts at concealment “constitute a ratification of the co-conspirator’s antecedent conduct including any mailings made in furtherance of the scheme.” Appellee's Br. at 50. That contention must be rejected because it conflicts with Grunewald v. United States, supra, which holds that acts of concealment after the principal object of the conspiracy (here use of the mails to defraud three states of tax revenues) has been completed do not comprise part of the conspiracy. See 353 U.S. at 403-06, 77 S.Ct. at 973-75.
Alternatively, the government contends that the charge given was substantially in accordance with the request. The court charged in part:
In order to find any defendant guilty on Count 22 [conspiracy] you must find all the following elements beyond a reasonable doubt:
(1) That the conspiracy described in the Indictment was existing at or about the time alleged;
*1237(2) That the particular defendant willfully became a member of the conspiracy with knowledge of its illegal objectives;
(3) That one of the conspirators thereafter knowingly committed at least one of the overt acts alleged in the indictment, at or about the time charged and during the period of the conspiracy; and
(4) That the overt act was knowingly done in furtherance of some object or purpose of the conspiracy as charged.
If you should find beyond a reasonable doubt from the evidence that the existence of the conspiracy charged in the Indictment has been proved and that during the existence of that conspiracy one of the overt acts alleged was knowingly done by one of the persons found by you to have been a conspirator, in furtherance of some object or purpose of the conspiracy, that proof of the conspiracy is complete; and it is complete as to every person found by you to have been willfully a member of that conspiracy at the time the overt act was committed, regardless of which of the conspirators did the overt act.
App. at 1458-59. Had the court stopped at this point, we could have accepted the government’s alternative argument. This much of the charge, while less specific in advancing Pflaumer’s theory than the requested instruction, is not inconsistent with the indictment and complies with Grunewald. But the court continued:
The Indictment alleges numerous overt acts in furtherance of each conspiracy. What must be shown is that after the defendant, whom you are considering, joined the conspiracy, some one of the conspirators performed an overt act in furtherance of the conspiracy.
Keep in mind that the Government must establish beyond a reasonable doubt that at least one of the overt acts as alleged in the indictment occurred while the conspiracy was still in existence. That is, the overt act must have occurred between approximately November, 1976, and January, 1981. You’re not to consider any acts that occurred after the conspiracy was terminated.
App. at 1467-68 (emphasis added). Thus, the court charged that the jury may find that the defendants joined "each conspiracy” as late as January, 1981. Plainly the court here rejected Pflaumer’s theory that a section 371 conspiracy to violate the mail fraud statute ends with the last mailing, here in June of 1979. The reference to plural conspiracies, and to overt acts after June 30, 1979, suggests that the court focused on a conspiracy to defraud the states as itself a violation of section 371. Such a conspiracy would not be a violation of section 371, nor does the indictment so charge. In any event, because the charge permitted the jury to find that the defendants joined the conspiracy after its principal object had terminated in June of 1979, the charge as given is inconsistent both with the indictment and with Grunewald v. United States, supra.
Finally, the government notes that, as the penultimate paragraph of the charge above (denoted [1]) indicates, the court charged that “[wjhat must be shown is that after the defendant ... joined the conspiracy, some one of the conspirators performed an overt act in furtherance of the conspiracy.” App. at 1467. The last overt act charged in the indictment occurred in March of 1978. The jury had a copy of the indictment. Therefore, the government maintains, although the charge permitted the jury to find an overt act as late as January, 1981, the jury must have found an overt act no later than March of 1978. Thus, the government concludes, the jury must have found that Pflaumer joined the conspiracy before that date. Appellee’s Br. at 47.
We cannot agree. We see no basis for believing that the jury would ignore the court’s instruction that it may find that the last overt act occurred in 1981. And we view with skepticism the suggestion that the jury would have understood which of the court’s instructions to obey and which to ignore. Moreover, the jury heard a significant amount of evidence suggesting possible concealment by Pflaumer after 1979, inviting it to do exactly what Pflaumer feared — find that he joined the conspiracy after the last mailing. Thus, far from suggesting that the jury may have ignored the court’s instructions, the evidence suggests that it may well have done precisely what the jury is expected to do: abide by them.
We must also determine whether the error is harmless. Fed.R.Crim.P. 52(a). If the evidence connecting Pflaumer to the conspiracy prior to June 30, 1979 were overwhelming, we could accept such an analysis. But as we observe in Part IV infra, that is not the case. Although some evidence tended to connect Pflaumer to the conspiracy before 1979, it is not sufficient to negate the possibility that the jury established his membership based in significant part on the post-1979 evidence. Because there is a reasonable likelihood that the jury may have considered acts of concealment as evidence of membership in the conspiracy alleged, we cannot find the error harmless.
Oxman, 740 F.2d at 1303-06.
. In relevant part the requested instruction read,
You have heard testimony of state auditors about their dealings with Mr. Pflaumer’s companies after June, 1979. If you find that Mr. Pflaumer only became aware of the alleged conspiracy or scheme to defraud Pennsylvania, New Jersey and Maryland of diesel fuel taxes at the time of these audits, you must acquit Mr. Pflaumer of all charges. In short, if you find that the Government has not proved beyond a reasonable doubt that Mr. Pflaumer knowingly and willfully became a member of the alleged conspiracy or scheme to defraud Pennsylvania, New Jersey and Maryland of diesel fuel taxes prior to June, 1979, you must acquit Mr. Pflaumer of all charges.
Joint Appendix at 1498 (emphasis supplied).
. Since this court has never indicated otherwise, and indeed in our prior opinion said expressly that nondisclosure may be excused if the government carries the burden of showing beyond a reasonable doubt that the error was harmless, nothing in Part II of the Bagley opinion affects this case or extant Third Circuit caselaw on the prosecutor’s duty to disclose.
. The dissenting judges, a fortiori, support the proposition that the determination must be made by the reviewing court. That allocation of decisional responsibility is consistent with Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 828, 17 L.Ed.2d 705 (1967).