It is axiomatic that the government must turn square corners when it undertakes a criminal prosecution. This axiom applies regardless of whether the target of the prosecution is alleged to have engaged in the daintiest of white-collar crimes or the most heinous of underworld activities. It follows that courts must be scrupulous in holding the government to this high standard as to sympathetic and unsympathetic defendants alike. The case before us plays out against the backdrop of these aphorisms.

More than ten years after he pleaded guilty to racketeering and related charges, petitioner-appellee Vincent Ferrara learned that the government had failed to disclose important exculpatory evidence to him beforehand. He sought relief under 28 U.S.C. § 2255, imploring the district court to vacate the remainder of his 22-year incarcerative sentence. The district court, in the person of the able judge who originally had sentenced the petitioner, granted his petition.

In arriving at this result, the district court relied principally on the rule announced in Brady v. Maryland, 373 U.S. 83, 87, 83 S. Ct. 1194, 10 L. Ed. 2d 215 (1963). See Ferrara v. United States, 384 F. Supp. 2d 384, 432 (D.Mass.2005). Although our reasoning differs somewhat — we rely solely on the operation of the rule announced in Brady v. United States, 397 U.S. 742, 748, 755, 90 S. Ct. 1463, 25 L. Ed. 2d 747 (1970) — we affirm the judgment below.

Because the government is appealing an order granting a petition for post-trial relief pursuant to 28 U.S.C. § 2255, we begin by describing the facts upon which the district court based its decision. We supplement those facts, as necessary, with other facts contained in the record.

On March 22, 1990, a federal grand jury sitting in the District of Massachusetts returned a superseding indictment charging eight men, including the petitioner (an alleged member of the Patriarca Family of La Cosa Nostra1), with racketeering and related offenses. The petitioner was named in thirty-five of the sixty-five counts.

The centerpiece of the superseding indictment — counts 1 and 2 — charged the petitioner and his codefendants with conspiring to participate in the affairs of a racketeering enterprise (the Patriarca Family), see 18 U.S.C. § 1962(d), and with participating in the affairs of a racketeering enterprise, see id. § 1962(c). To help establish the pattern of racketeering activity necessary to support these accusations against the petitioner, the indictment alleged three predicate acts of murder: that the petitioner and others had conspired to kill and had killed Giacomo DiFronzo and Anthony Corlito (Racketeering Acts A-1 and A-2) and that the petitioner and codefendant Pasquale Barone had conspired to kill and had killed Vincent James Limoli (Racketeering Act A-3). Although counts 3 and 4 charged the petitioner and Barone with the substantive crimes of conspiring to murder and actually murdering Limoli in order to bolster their positions in the Patriarca Family in violation of 18 U.S.C. § 1959 — the government theorized that the petitioner ordered Barone to kill Limoli because Limoli had stolen drugs from a member of the Patriarca Family — the petitioner was not charged separately for the conduct described in Racketeering Acts A-1 and A-2.

Since the prosecution's evidence linking the petitioner to the DiFronzo and Corlito murders appeared "sparse and weak," Ferrara, 384 F.Supp.2d at 389 n. 1, the government had a heightened interest in hanging the Limoli murder around the petitioner's neck (after all, so long as he was found responsible for any one of the murder-related racketeering acts, he would face a mandatory sentence of life imprisonment rather than a sentence as low as 151 months). Moreover, establishing that the petitioner orchestrated Limoli's murder on behalf of the Patriarca Family would further the government's goal of securing a lengthy sentence for the petitioner's codefendant, Raymond Patriarca, Jr.

Walter Jordan (Barone's brother-in-law) was slated to be the government's key witness anent the Limoli murder. Jordan had participated in the slaying and then fled to North Carolina. When arrested there, he agreed to cooperate with the government in exchange for immunity and protection.

On July 27, 1988, Jordan told the grand jury that the petitioner, upon learning that Limoli had stolen drugs belonging to a member of the Patriarca Family, had ordered Barone to assassinate Limoli. Barone recruited Jordan to assist him. At Barone's bidding, Jordan set up an apocryphal drug deal that lured Limoli to a restaurant in Boston's North End. After leading their prey down a side street, Barone shot him.

The killers then met with the petitioner. Although the petitioner questioned them about the Limoli murder in a manner suggesting ignorance of what had happened, Jordan maintained that the petitioner had ordered the shooting. Shortly thereafter, Barone informed Jordan that they needed to leave town because the petitioner was going to kill them. Jordan fled that same night.

In July of 1991, Jordan met in Salt Lake City with Assistant United States Attorneys Jeffrey Auerhahn and Gregg Sullivan, Special Agent Michael Buckley of the Federal Bureau of Investigation, and a Boston police detective, Martin Coleman (who was serving as a member of an intergovernmental task force). The group went through a series of trial preparation sessions. After the last session, Jordan told Coleman that Barone had never obtained the petitioner's permission to eliminate Limoli. He also denied that Barone ever said that the petitioner had either ordered or blessed the murder.

Recognizing the potential impact of this recantation, Coleman relayed the information to Auerhahn after they returned to Boston. Auerhahn immediately set up a conference call, during which Jordan reiterated what he had told Coleman. Auerhahn then arranged another trial preparation session to deal with this unexpected turn of events. Jordan, Auerhahn, Coleman, and Buckley attended this conclave, which took place in Minneapolis in August of 1991. When Jordan again began to recant his grand jury testimony, Buckley became indignant. Jordan was shown the door, and Auerhahn and Buckley directed Coleman to "straighten him out." Id. at 395.

After Coleman and Jordan returned to the meeting, the prosecution team again asked Jordan about the petitioner's role in Limoli's murder. Fearing that he would otherwise face a whipsaw (losing not only the immunity previously granted but also the government's protection against retaliation by the Patriarca Family), Jordan reverted to his original story. Contrary to his normal practice, Auerhahn did not take detailed notes at the Minneapolis meeting. The district court found that this was deliberate; Auerhahn "did not want to create a record of the changes in Jordan's testimony." Id.

Coleman rarely wrote memoranda. However, he believed that Jordan's recantation was "very important" and "should be brought to the attention of everyone involved in the investigation." As a result, he prepared a handwritten memorandum in which he memorialized the conversations that had taken place before the Minneapolis meeting (the Coleman memo). That document, admitted as a full exhibit during the evidentiary hearing on the section 2255 petition, read as follows:

On Wednesday, July 24, 1991 at about 11:30 M.T. while having a conversation with Tony Jordan a govt. witness, Mr. Jordan stated to me that Patty's Barone had fucked up and did not get permission to kill Jimmy Limoli.

On Thursday, July 25, 1991 we traveled back to Bos. and I told AUSA Mr. Auerhahn that I had to see him about Tony Jordan. On Friday July 26, 1991 I talked to AUSA Auerhahn and told him what Mr. Jordan had said to me.

On Monday July 29, 1991 at about 10:30 am Mr. Jordan returned a call to Mr. Auerhahn's office and I ask him to repeat to AUSA Auerhahn what he had said to me on July 24.

At this time Mr. Jordan stated that he knew that Pattsie Barone had not gotten permission to kill Jimmy Limoli. He found this out after Jimmy's wake when Patty's and he got into a car with Vinny Ferrara and a Joe the Jeweler.

Vinny said to Patty's "who's next me." Jordan further stated that some time later that week that Patty's got called over to Franchesco's and Vinny told him he was dead. At this time Patty's ran out of Franchesco's . . . and over to his apartment where Jordan was and told Jordan that they had to get out of town because Vinny was going to kill them. Jordan then said why is he going to kill us and Patty's said because I did not get permission to kill Jimmy.

The district court determined that the Coleman memo comprised "a reliable account of what Jordan told Auerhahn and Coleman in 1991." Id. at 407 n. 14.

At the evidentiary hearing, Coleman testified that he provided Auerhahn a copy of his memorandum and expressed his belief that Jordan's statements were exculpatory. See id. at 395. Auerhahn asked Coleman if he (Auerhahn) could "clean [the memo] up." Assuming that the prosecutor merely wanted to correct his spelling and grammar, Coleman voiced no objection. Auerhahn proceeded to prepare a significantly toned-down version of the Coleman memo for the file (the Auerhahn memo). Even though Auerhahn drafted the second memorandum, he continued to list Coleman as its author.2

The district court found that Auerhahn prepared the second memorandum to "sanitize" the Coleman memo, so that it would be "less damaging to the government . . . if it were produced to the defendants." Id. at 396. While the Coleman memo "directly refute[d] the charges that Ferrara ordered Barone to murder Limoli," id. at 396-97, the Auerhahn memo never conclusively stated that Jordan did not believe that the petitioner was involved. In all events, even the Auerhahn memo indicated that Jordan had made prior inconsistent statements and that, on at least one occasion, Barone had denied that the petitioner played any role in Limoli's slaying.

Under the local rules in effect when Jordan recanted, the prosecution was required to turn over automatically all written material constituting "exculpatory evidence within the meaning of Giles v. Maryland, 386 U.S. 66, 87 S. Ct. 793, 17 L. Ed. 2d 737 (1967), Brady v. Maryland, 373 U.S. 83, 83 S. Ct. 1194, 10 L. Ed. 2d 215 (1963) and Giglio v. United States, 405 U.S. 150, 92 S. Ct. 763, 31 L. Ed. 2d 104 (1972)." D. Mass. R. 116.1(a)(5) (1990). The rule required initial disclosure within fourteen days of arraignment; from that point forward, the government had a continuing duty to supplement its original disclosure if and when new material surfaced. See id. 116.1(c). Here, moreover, a magistrate judge, in an order dated May 15, 1990, had directed the government to produce all exculpatory evidence (including "all documents, statements and . . . written summar[ies] of all oral evidence and statements") provided by government trial witnesses, save only Jencks Act material, see 18 U.S.C. § 3500, and material that would tend to disclose the identity of an otherwise unknown witness.

In a discovery letter dated December 14, 1989, Auerhahn identified Jordan as a cooperating witness and acknowledged the government's continuing duty to disclose exculpatory evidence. Yet Auerhahn did not furnish copies of either the Coleman memo or the Auerhahn memo, nor did he disclose the substance of Jordan's recantation to the defense. Even after Jordan had told him that the petitioner did not order the murder, Auerhahn, in a pleading dated October 15, 1991, asserted that the government had fully complied with Local Rule 116.1 and indicated that it would disclose the remaining exculpatory material in its possession in its trial brief.

The district court supportably found that "Auerhahn recognized that [the Coleman memo] memorialized exculpatory information." Ferrara, 384 F.Supp.2d at 395. Notwithstanding this knowledge, the government never disclosed the Coleman memo, the Auerhahn memo, or any summary of Jordan's vacillatory statements to the petitioner. Its trial brief, filed on October 16, 1991, over Auerhahn's signature, gave the impression that Jordan would serve as a stalwart witness for the prosecution, declaring that he would "testify to Barone's statement that Limoli was killed on the orders of Vincent Ferrara." The district court warrantably found that this brief "made it clear that the government's case concerning the Limoli murder depended heavily, if not exclusively, on Jordan's testimony." Id. at 398.

The petitioner and four codefendants went to trial early in 1992. On January 21 — after jury empanelment but before opening statements — the five men entered into linked plea agreements (i.e., each plea agreement was contingent on the district court's acceptance of the other plea agreements).

The plea agreements were structured to allow the defendants to remain silent about the factual basis for their pleas and, particularly, the existence of La Cosa Nostra. The petitioner's agreement called for him to plead guilty to all thirty-five counts against him, for which he would receive a 264-month sentence, immunity from federal prosecution for as-yet-uncharged murders, and immunity from state prosecution for the Limoli murder. It did not require him to admit to any role in the DiFronzo or Corlito murders.

At the change-of-plea hearing held the next day, Auerhahn limned the factual basis for the guilty pleas. In conditionally accepting the pleas subject to the preparation of presentence investigation reports, the district court, by its own assessment made contemporaneously at the hearing, relied "in a meaningful measure" on the government's trial brief. The court's conclusions that a solid factual foundation undergirded the petitioner's plea and that the plea was knowing and voluntary stemmed from that reliance.

Despite pleading guilty to all thirty-five counts, the petitioner, during interviews by the probation department, refused to accept responsibility for any of the three aforementioned murders. He told the probation officer that he was particularly troubled that his plea intimated that he had a role in the Limoli murder. He went on to say that, "prior to entering the plea, he spoke with the decedent's father, and explained the untenable position he was in and why he found it necessary to acknowledge [an] involvement in James Limoli's death, even though he disavows any planning or participating in the killing." The petitioner's trial counsel, Oscar Goodman, confirmed that his client had consistently denied "order[ing] the murder of Mr. Limoli."

The district court convened the disposition hearing on April 29, 1992. Notwithstanding the petitioner's steadfast refusal to accept responsibility for Limoli's death, the court remained "satisfied there was a proper factual basis for the guilty plea" and that the petitioner was "guilty of the offenses" to which he had pleaded. Consequently, the court accepted his plea, ratified the plea agreement, and imposed the pre-agreed 264-month incarcerative term. The judge later stated that if the government had disclosed the information regarding Jordan's retraction, "there [was] a reasonable probability . . . the court would have found that there was not a proper factual basis to accept his plea." Id. at 423.

Barone's trial commenced in 1993. The government, in the person of Auerhahn qua prosecutor, "fully understood that it would be both essential and challenging for it to prove that Ferrara ordered Barone to kill Limoli." Id. at 402. Not surprisingly, Jordan proved to be the key witness in this endeavor. Pursuant to the terms of his agreement with the government, he testified, as he had before the grand jury, that Barone told him the petitioner had ordered the execution. Jordan also testified that Barone, with his help, carried out the order. Although Jordan's testimony was central to the prosecution's theory of the case, the government never disclosed to the defense the Coleman memo, the Auerhahn memo, or the substance of Jordan's recantation.3

Limoli's half-sister, Elizabeth DiNunzio, also testified against Barone. She vouchsafed that Limoli was slain after leaving to meet Barone and Jordan and that the petitioner subsequently had assured her that he would find out who murdered her brother.

Both Jordan and DiNunzio provided additional testimony touching upon the petitioner's role in the deaths of DiFronzo and Corlito. Jordan testified that Barone told him both that the petitioner had shot DiFronzo and that he (Barone), together with the petitioner and Limoli, had assassinated Corlito. DiNunzio testified that Limoli told her that he and the petitioner had killed DiFronzo and that he, Barone, and the petitioner had murdered Corlito.

After twice reporting that it was deadlocked, the jury eventually convicted Barone of racketeering and conspiring with the petitioner to kill Limoli. It remained unable to reach a verdict on whether Barone had murdered Limoli to advance his standing in the Patriarca Family.

In June of 2002, Jordan contacted a federal law enforcement official and confessed that he had committed perjury at Barone's trial. He further stated that he had told Auerhahn and other members of the prosecution team the truth during trial preparation sessions in the summer of 1991 and that "there was no doubt in his mind that . . . Auerhahn knew that Jordan lied when he testified that [the petitioner] had ordered the [Limoli] murder." An investigation ensued. By then, the petitioner and Barone already had filed separate motions under 28 U.S.C. § 2255, advancing various grounds for post-conviction relief. Recognizing the potential relevance of Jordan's allegations to the pending petitions, the government at long last disclosed the information concerning Jordan's 1991 recantation.

The district court allowed amendment of the pending section 2255 motions to encompass the newly discovered evidence. The petitioner filed his amended motion on July 15, 2003, arguing in pertinent part that the government had failed to disclose material exculpatory evidence and that its misconduct had induced his decision to plead guilty.4 The court held an evidentiary hearing at which Jordan, Coleman, Buckley, Sullivan, and Auerhahn testified.

In a post-hearing memorandum, dated April 6, 2004, the government argued, for the first time, that the district court would need to recognize a new constitutional rule granting defendants a due process right to receive exculpatory evidence before pleading guilty in order to afford the petitioner relief. Doing so, the government maintained, would transgress the nonretroactivity principle enunciated in Teague v. Lane, 489 U.S. 288, 310, 109 S. Ct. 1060, 103 L. Ed. 2d 334 (1989). The petitioner denounced this proposition as both untimely and legally flawed. The court agreed; it ruled that the government had waived its Teague defense by not timely asserting it and went on to find that, in any event, the petitioner's claim was not Teague-barred. Ferrara, 384 F.Supp.2d at 421.

The merits-based aspect of the lower court's Teague ruling was grounded on two rules of constitutional law that, the court said, were well-established by 1992 (when the petitioner's conviction became final). First, the court found that Jordan's recantation constituted exculpatory evidence that tended to negate the petitioner's guilt or, alternatively, could have been used to impeach a crucial prosecution witness. Id. at 423. Thus, it concluded that, under Brady v. Maryland and its progeny, the government's conduct violated the petitioner's due process rights. See id. at 432. Second, the court found that the government's nondisclosure constituted misconduct that deprived the petitioner of the ability to enter a knowing and voluntary plea and, therefore, transgressed the rule of Brady v. United States. See id. at 432-33.

That left the question of the appropriate remedy. On May 13, 2005, the district court amended the judgment of conviction in the underlying case to acquit the petitioner on counts 3 and 4 and resentenced him to time served. The government filed a motion for reconsideration, which the court, aside from correcting typographical errors, summarily denied. These timely appeals followed.5 In them, the government vigorously contests the granting of any section 2255 relief. Beyond that point, however, the government apparently does not quarrel with the appropriateness of the remedial order fashioned by the district court.

Although not couched as a separate argument, the government questions a good many of the district court's factual findings. In order to set the stage for our legal analysis, we deal first with this critique. In doing so, we scrutinize the disputed factual findings for clear error. See Ellis v. United States, 313 F.3d 636, 641 (1st Cir.2002).

At the evidentiary hearing, Auerhahn asserted that he had no memory of Jordan ever distancing himself from the notion that the petitioner had ordered the Limoli murder. He also testified that he had no recollection either of Coleman telling him that Jordan had retracted his grand jury testimony or of having seen the Coleman memo prior to the institution of the section 2255 proceedings. Furthermore, he denied having authored the Auerhahn memo. This testimony clashed head-on with testimony given by Jordan and Coleman. Confronted with conflicting versions of what had transpired, the district court rejected Auerhahn's account and found as a fact that Coleman had told Auerhahn about Jordan's Salt Lake City recantation directly after it occurred; that Jordan repeated his recantation to both Coleman and Auerhahn in a telephone conversation shortly thereafter; that Auerhahn received a copy of the Coleman memo in roughly the same time frame; that Auerhahn (an experienced prosecutor) recognized that the Coleman memo memorialized important exculpatory information; and that Auerhahn then drafted the Auerhahn memo in hopes that it would be less damaging than the Coleman memo should the government be forced to produce something. Ferrara, 384 F.Supp.2d at 394-96, 407 n. 14.

On appeal, the government brushes aside these findings. It reiterates Auerhahn's testimony in its statement of facts as if that testimony were gospel and at the same time belittles Coleman's contrary testimony. To the extent that the government's case rests, in whole or in part, on this facet of Auerhahn's testimony, it is doomed to failure.

Under the applicable standard of review, a party challenging a trial court's factual findings faces a steep uphill climb. See Mitchell v. United States, 141 F.3d 8, 17 (1st Cir.1998) (explaining that factual findings are not clearly erroneous unless the reviewing court is "left with the definite and firm conviction that a mistake has been made"); State Police Ass'n v. Comm'r, 125 F.3d 1, 5 (1st Cir.1997) (similar). This standard is particularly deferential where, as here, the challenged findings hinge on the trier's credibility determinations. See Anderson v. Bessemer City, 470 U.S. 564, 575, 105 S. Ct. 1504, 84 L. Ed. 2d 518 (1985); Anthony v. Sundlun, 952 F.2d 603, 606 (1st Cir.1991).

Here, the district court possessed a deep familiarity with the facts that we, working from an algid appellate record, cannot hope to replicate. After all, Judge Wolf had overseen the pretrial maneuverings in the original case, taken the petitioner's guilty plea, imposed his sentence, presided over Barone's trial, and then revisited the matter when section 2255 proceedings were instituted. He was intimately acquainted with both the cast of characters and the array of issues. Moreover, he had an opportunity to view the witnesses at first hand. Given this prolonged exposure, the district court's credibility determinations are entitled to a considerable degree of respect. See, e.g., Anthony, 952 F.2d at 606 (stating that appellate courts should be chary about undermining findings "based on witness credibility [determinations] made by a trial judge who has seen and heard the witnesses at first hand").

A careful reading of the record confirms this intuition. Although Coleman's testimony about the events that had taken place contained a few inconsistencies, that was understandable given the passage of nearly thirteen years between those events and the evidentiary hearing. The court found that Coleman, whom several witnesses described as a dedicated law enforcement officer, would not have written a false report. Ferrara, 384 F.Supp.2d at 407 n. 14. Relatedly, the court determined that the Coleman memo provided a reliable account of what Jordan told Coleman and Auerhahn in 1991. Id. These findings, along with Coleman's plausible testimony that he had shown Auerhahn his memorandum shortly after he composed it and other testimony regarding the standard office practices of the prosecution team, gave the district court more than enough ammunition to shoot down Auerhahn's self-serving version of the relevant events. Accordingly, we discern no clear error in any of the disputed factual findings.

Notwithstanding the sturdiness of the district court's factual findings, the question remains whether those findings justify its decision to vacate the petitioner's sentence. Affording de novo review to the lower court's legal conclusions, see Ellis, 313 F.3d at 641, we now turn to that question.

The government argues that the Teague doctrine precludes the outcome reached below because, in order to grant relief, the district court had to extend the constitutional rule that defendants have a due process right to exculpatory evidence, as announced in Brady v. Maryland, 373 U.S. at 87, 83 S. Ct. 1194, beyond its established parameters to a new frontier — the guilty plea context.6 In the government's view, this infirmity in the district court's decision requires reversal of the order granting section 2255 relief. Since the Teague nonretroactivity principle is central to this debate, we start there.

Under 28 U.S.C. § 2255, a prisoner in federal custody may ask a sentencing court to vacate, set aside, or correct a sentence imposed in violation of the Constitution or some other federal law. A prisoner's section 2255 rights are limited, however, by the nonretroactivity principle announced in Teague. Pursuant to that principle, a prisoner is generally not entitled to collateral relief if granting that relief would require the court to apply a "new rule" of constitutional procedure. See O'Dell v. Netherland, 521 U.S. 151, 156, 117 S. Ct. 1969, 138 L. Ed. 2d 351 (1997). A rule of constitutional procedure is considered "new" if it "was not dictated by precedent existing at the time the defendant's conviction became final." Teague, 489 U.S. at 301, 109 S. Ct. 1060; see Graham v. Collins, 506 U.S. 461, 467, 113 S. Ct. 892, 122 L. Ed. 2d 260 (1993).

In order to determine whether a particular claim is Teague-barred, an inquiring court first must ascertain the date on which the petitioner's conviction became final and then must survey the legal landscape as it existed at that time to gauge whether then-existing precedent would have compelled it to grant the petition. See O'Dell, 521 U.S. at 156, 117 S. Ct. 1969. If a court would not have been so compelled, the petition (if valid at all) depends on a new rule within the purview of Teague, and the petitioner is barred from seeking collateral relief unless one of two narrow exceptions applies.7 See id. at 156-57, 117 S. Ct. 1969.

Here, however, we take note of a threshold matter: the district court's finding of waiver. See Ferrara, 384 F.Supp.2d at 412. While a district court must consider a properly raised Teague defense, see Horn v. Banks, 536 U.S. 266, 267, 122 S. Ct. 2147, 153 L. Ed. 2d 301 (2002), such a defense is not jurisdictional. Accordingly, the government can waive a Teague defense by failing to raise it in a timeous manner. See Goeke v. Branch, 514 U.S. 115, 117, 115 S. Ct. 1275, 131 L. Ed. 2d 152 (1995); Schiro v. Farley, 510 U.S. 222, 229, 114 S. Ct. 783, 127 L. Ed. 2d 47 (1994).

In the case at hand, the district court's finding of waiver rests on the government's admitted failure to raise a Teague defense until after the completion of the evidentiary hearing and the submission of an initial round of post-hearing briefing. The government assails this finding. What counts, it says, is that it raised the Teague defense at a point in the proceedings that not only allowed the petitioner ample time to respond but also allowed the court ample opportunity to consider the issue. Because we conclude that the government's Teague defense is without merit as applied to the Brady v. United States holding (which we view as supplying the dispositive rule of law in this case), we need not resolve the waiver issue. We explain briefly.

While the district court based its judgment on two constitutional rules — one emanating from Brady v. Maryland and the other from Brady v. United States — we conclude that the latter rule, by itself, resolves these appeals. This rule, which provides that a defendant's guilty plea must be voluntary in order to serve as a constitutionally valid basis for a conviction, see Brady v. United States, 397 U.S. at 748, 90 S. Ct. 1463, was firmly in place when the petitioner's conviction became final.

It is common ground that a defendant surrenders a number of constitutional rights, not the least of which are his privilege against self-incrimination and his right to a trial by jury, when he pleads guilty. For that reason, it has long been established that if a defendant's plea is not entered knowingly and voluntarily, it has been procured in violation of the Due Process Clause and, therefore, a conviction based thereon is invalid. See McCarthy v. United States, 394 U.S. 459, 466, 89 S. Ct. 1166, 22 L. Ed. 2d 418 (1969); Machibroda v. United States, 368 U.S. 487, 493, 82 S. Ct. 510, 7 L. Ed. 2d 473 (1962). By the same token, a prisoner can collaterally attack his sentence on the ground that his guilty plea was not knowing or voluntary if his claim is based on evidence not available to him at the time of the plea. See Machibroda, 368 U.S. at 493, 82 S. Ct. 510; Waley v. Johnston, 316 U.S. 101, 104, 62 S. Ct. 964, 86 L. Ed. 1302 (1942).

These rules were well-settled when the petitioner's conviction became final in 1992. In 1970, for example, the Supreme Court confirmed that a guilty plea must be both voluntary and intelligent if it is to represent a constitutionally valid predicate for a conviction. See Brady v. United States, 397 U.S. at 748, 90 S. Ct. 1463. The Brady Court also defined the standard for determining whether a defendant's plea qualifies as voluntary:

[A] plea of guilty entered by one fully aware of the direct consequences, including the actual value of any commitments made to him by the court, prosecutor, or his own counsel, must stand unless induced by threats (or promises to discontinue improper harassment), misrepresentation (including unfulfilled or unfulfillable promises), or perhaps by promises that are by their nature improper as having no proper relationship to the prosecutor's business (e.g.bribes).

Id. at 755 (citation and internal quotation marks omitted). By 1992, this standard was deeply entrenched in federal law. See, e.g., Mabry v. Johnson, 467 U.S. 504, 509, 104 S. Ct. 2543, 81 L. Ed. 2d 437 (1984); United States v. Bouthot, 878 F.2d 1506, 1511 (1st Cir.1989); Correale v. United States, 479 F.2d 944, 947 (1st Cir.1973). Consequently, the rule is not a new rule, as Teague defines that term, with respect to the petitioner's case.

In this instance (as we shall explain below), we conclude that the petitioner's plea does not qualify as voluntary under that standard. Since this is true whether or not the petitioner had a constitutional right to receive the exculpatory evidence under the rule announced in Brady v. Maryland, we need not address the much closer question of whether a court, in 1992, would have been compelled to rule that the petitioner had a Brady v. Maryland right to receive the information before pleading guilty.

A defendant who was warned of the usual consequences of pleading guilty and the range of potential punishment for the offense before entering a guilty plea must make two showings in order to set that plea aside as involuntary. First, he must show that some egregiously impermissible conduct (say, threats, blatant misrepresentations, or untoward blandishments by government agents) antedated the entry of his plea. See Brady v. United States, 397 U.S. at 755, 90 S. Ct. 1463; see also Correale, 479 F.2d at 947 & n. 3 (discussing types of prosecutorial misconduct that might render a plea involuntary). Second, he must show that the misconduct influenced his decision to plead guilty or, put another way, that it was material to that choice. See Brady v. United States, 397 U.S. at 755, 90 S. Ct. 1463 (explaining that the misconduct must "induce" the decision); Cepulonis v. Ponte, 699 F.2d 573, 577 (1st Cir.1983) (similar). In mounting an inquiry into these elements, a court must consider the totality of the circumstances surrounding the plea. Brady v. United States, 397 U.S. at 749, 90 S. Ct. 1463; United States v. Webb, 433 F.2d 400, 404 (1st Cir.1970).

In this instance, we have the benefit of extensive factual findings elaborating the circumstances attendant to the petitioner's decision to plead guilty. See Ferrara, 384 F.Supp.2d at 390-400. Assisted by those findings, we undertake the requisite two-part inquiry.

1. Misconduct. The initial question is whether the government's failure to reveal what it knew about Jordan's recantation prior to the change-of-plea hearing qualifies as the sort of misconduct that might render a plea involuntary.

It cannot be gainsaid that a defendant's decision to enter a guilty plea is sometimes influenced by his assessment of the prosecution's case. See Brady v. United States, 397 U.S. at 756, 90 S. Ct. 1463. But a plea is not rendered infirm "merely because [the defendant] discovers long after the plea has been accepted that his calculus misapprehended the quality of the [government's] case." Id. at 757, 90 S. Ct. 1463. Even if a defendant's misapprehension of the strength of the government's case induces him to throw in the towel, that misapprehension, standing alone, cannot form the basis for a finding of involuntariness. See Bousley v. United States, 523 U.S. 614, 619, 118 S. Ct. 1604, 140 L. Ed. 2d 828 (1998). It is only when the misapprehension results from some particularly pernicious form of impermissible conduct that due process concerns are implicated. See Brady v. United States, 397 U.S. at 757, 90 S. Ct. 1463; Bouthot, 878 F.2d at 1512.

Let us be perfectly clear: due process does not normally require the prosecution either to turn over the whole of its file or to disclose every shred of information that might conceivably affect the defendant's decisionmaking. See Bouthot, 878 F.2d at 1512. Even though a defendant obviously would be interested in knowing all the strengths and weaknesses of the government's proof before deciding whether to plead guilty or risk a trial, the government's refusal to render the whole of its case transparent before a defendant makes that election does not, in the ordinary course, constitute the kind of severe misconduct that is needed to render a plea involuntary.

Under limited circumstances, however — everything depends on context — the prosecution's failure to disclose evidence may be sufficiently outrageous to constitute the sort of impermissible conduct that is needed to ground a challenge to the validity of a guilty plea. See Bouthot, 878 F.2d at 1511 (stating that a defendant could attack his plea under Brady v. United States by showing that the prosecution's failure to provide information constituted a "material omission tantamount to a misrepresentation"); see also Matthew v. Johnson, 201 F.3d 353, 364 n. 15 (5th Cir.2000) (suggesting that, "[e]ven if the nondisclosure is not a Brady [v. Maryland] violation," there may be situations in which the prosecution's failure to disclose evidence makes it "impossible for [a defendant] to enter a knowing and intelligent plea").

The government attempts to wedge its refusal to provide the petitioner the evidence of Jordan's recantation within the broader category of nondisclosures that are insufficient to render a plea involuntary. It argues that it had no duty to turn over its entire case file; that the petitioner took a calculated gamble in pleading guilty rather than going to trial; and that labeling his plea involuntary, years after the fact, reflects no more than a tacit admission that he overestimated the strength of the government's case.

This argument lacks force. Here, we are dealing with more than simple neglect to turn over exculpatory evidence; the government manipulated the witness (Jordan) into reverting back to his original version of events, then effectively represented to the court and the defense that the witness was going to confirm the story (now known by the prosecution to be a manipulated tale) that the petitioner was responsible for killing Limoli. These egregious circumstances make this one of those rare instances in which the government's failure to turn over evidence constitutes sufficiently parlous behavior to satisfy the misconduct prong of the involuntariness test. We explain briefly.

The government's obligation to disclose the evidence can hardly be doubted. When the grand jury handed up the original indictment on November 16, 1989, the local rules required automatic disclosure of all evidence within the government's ken that tended to negate a defendant's guilt. See D. Mass. R. 42(a)(5) (1986). Shortly after the filing of the superseding indictment, the local rules were amended to impose an additional requirement: that the government, on its own initiative, provide the defendant with any and all exculpatory evidence that might be used to impeach witnesses whom it intended to call at trial. See D. Mass. R. 116.1(a)(5) (1990); see also Giglio v. United States, 405 U.S. 150, 154, 92 S. Ct. 763, 31 L. Ed. 2d 104 (1972). And, moreover, the district court, by a magistrate judge's order dated May 15, 1990, specifically directed the government (with exceptions not relevant here8) to disclose all exculpatory evidence pertaining to its trial witnesses. The government, then, was obligated three times over to disclose to the petitioner the information concerning Jordan's recantation.

The evidence in question — Jordan's recantation, as memorialized in the Coleman memo — was plainly exculpatory. As the government concedes, Jordan would have been an integral prosecution witness had the petitioner's case proceeded to trial. Had Jordan testified along the lines delineated in the government's trial brief, the petitioner could have used the recantation evidence to mount a ferocious attack on his credibility. Since the jury's verdict may well have hinged on its evaluation of Jordan's credibility (at least with respect to the charges that centered on Limoli's murder), the evidence was of enormous significance for impeachment purposes. See Giglio, 405 U.S. at 154, 92 S. Ct. 763.

In addition, the evidence tended to negate the petitioner's guilt. After all, the Coleman memo memorialized both Barone's admission that he did not receive a green light from the petitioner to proceed with Limoli's murder and Jordan's admission that he knew the petitioner had not ordered the execution.9 Since these admissions, if accepted as true, would have precluded a jury from holding the petitioner liable for the Limoli murder, the suppressed evidence was suggestive of the petitioner's factual innocence.

The government's failure to disclose this exculpatory information cannot be palmed off as mere inadvertence (or even as slipshod performance). The district court supportably found that the lead prosecutor, Auerhahn, knew of both the Coleman memo and the substance of Jordan's recantation. See Ferrara, 384 F.Supp.2d at 394-95, 407 n. 14. This knowledge triggered the government's obligation to disclose the evidence under either version of the local rule as well as under the court's case-specific order. The court found that, instead of fulfilling that obligation, the prosecution team manipulated the witness and deliberately tried to cover up the evidence. See id. at 395-96, 397 n. 10. These findings were not clearly erroneous.

Given the presumption that prosecutors can be relied on to perform their official duties properly, see Ramírez v. Sánchez Ramos, 438 F.3d 92, 99 (1st Cir.2006), the petitioner should have been able to trust the government to turn square corners and fulfill its discovery obligations. In the absence of a positive indication that it would not comply — and there was none here — we think it is fair to say that the government, at the very least, impliedly promised that it would provide the petitioner with all exculpatory evidence. The government broke that promise egregiously when it failed either to reveal Jordan's recantation or to put the petitioner on notice that it was withholding the evidence.

Moreover, the government made affirmative misrepresentations to the petitioner. In its December 1989 discovery letter, Auerhahn explicitly represented that the prosecution would satisfy its continuing duty to disclose all exculpatory evidence in a timely manner. That promise, insofar as it pertained to Jordan's recantation and the Coleman memo, went unfulfilled. In the same vein, Auerhahn (in an October 15, 1991 filing concerning the requested discovery of grand jury testimony) explicitly represented that the government had fully complied with the applicable local rule, thereby warranting that it had either disclosed all exculpatory information within its ken10 or notified the defendants (including the petitioner) of the exculpatory evidence it was refusing to disclose. Because this filing was served more than two months after Auerhahn learned of Jordan's recantation, Auerhahn's claim of compliance plainly and inexcusably misrepresented the true state of affairs.

To sum up, the government's actions in this case do not depict some garden-variety bevue but, rather, paint a grim picture of blatant misconduct.11 The record virtually compels the conclusion that this feckless course of conduct — the government's manipulative behavior, its failure to disclose the Jordan recantation and/or the Coleman memo, and its affirmative misrepresentations (not anchored to any rational and permissible litigation strategy) — constituted a deliberate and serious breach of its promise to provide exculpatory evidence. In the circumstances of this case, then, the government's nondisclosure was so outrageous that it constituted impermissible prosecutorial misconduct sufficient to ground the petitioner's claim that his guilty plea was involuntary. See Brady v. United States, 397 U.S. at 755, 90 S. Ct. 1463; Correale, 479 F.2d at 747.

2. Prejudice. A finding of impermissible conduct is a necessary but not a sufficient condition for the success of an involuntariness argument. The petitioner also must show "a reasonable probability that, but for [the misconduct], he would not have pleaded guilty and would have insisted on going to trial." Hill v. Lockhart, 474 U.S. 52, 59, 106 S. Ct. 366, 88 L. Ed. 2d 203 (1985).12 For purposes of this standard, a reasonable probability is a probability sufficient to undermine confidence in a belief that the petitioner would have entered a plea. See Miller v. Angliker, 848 F.2d 1312, 1320 (2d Cir.1988).

A court charged with determining the existence of a reasonable probability that a defendant would have insisted on a trial in the absence of government misconduct must take an objective approach. See United States v. Walters, 269 F.3d 1207, 1215 (10th Cir.2001); United States v. Avellino, 136 F.3d 249, 256 (2d Cir.1998). The elementary question is whether a reasonable defendant standing in the petitioner's shoes would likely have altered his decision to plead guilty had the prosecution made a clean breast of the evidence in its possession. See Miller, 848 F.2d at 1322. Because a multiplicity of factors may influence a defendant's decision to enter a guilty plea, a court attempting to answer this question must use a wide-angled lens. See Brady v. United States 397 U.S. at 749, 90 S. Ct. 1463. Relevant factors include, but are not limited to, (i) whether the sequestered evidence would have detracted from the factual basis used to support the plea, see Matthew, 201 F.3d at 365; (ii) whether the sequestered evidence could have been used to impeach a witness whose credibility may have been outcome-determinative, see Conley v. United States, 415 F.3d 183, 189 (1st Cir.2005); (iii) whether the sequestered evidence was cumulative of other evidence already in the defendant's possession, see id. at 192; (iv) whether the sequestered evidence would have influenced counsel's recommendation as to the desirability of accepting a particular plea bargain, see Hill, 474 U.S. at 59, 106 S. Ct. 366; and (v) whether the value of the sequestered evidence was outweighed by the benefits of entering into the plea agreement, see White v. United States, 858 F.2d 416, 424 (8th Cir.1988).

While this checklist is useful, experience teaches that each defendant's decision as to whether or not to enter a guilty plea is personal and, thus, unique. Consequently, the compendium of relevant factors and the comparative weight given to each will vary from case to case. The ultimate aim, common to every case, is to ascertain whether the totality of the circumstances discloses a reasonable probability that the defendant would not have pleaded guilty absent the misconduct.

In this instance, the district court determined that the government's failure to produce the exculpatory evidence concerning Jordan's recantation was material to the petitioner's decision to plead guilty.13 See Ferrara, 384 F.Supp.2d at 430 (finding a reasonable probability that the petitioner would have proceeded to trial if the evidence had not been suppressed). The government assigns error to this determination. In evaluating the government's argument, we review the district court's ultimate determination de novo but accept its subsidiary factual findings so long as they are not clearly erroneous. See Ouimette v. Moran, 942 F.2d 1, 4 (1st Cir. 1991).

The government's first line of attack is procedural in nature. It argues that the petitioner has forfeited the point because he never submitted a sworn declaration attesting that he would not have pleaded guilty had he known of the exculpatory evidence. This argument lacks force: while the applicable procedural rule originally required a habeas petitioner to file a sworn statement in support of his section 2255 application, see Rule 2(b), Rules Governing Section 2255 Proceedings (1976), that rule was amended in 1982 — long before the petitioner undertook his quest for post-conviction relief. The amended rule allowed an inmate to sign his application under penalty of perjury in lieu of submitting sworn statements. See Rule 2(b), Rules Governing Section 2255 Proceedings (1982).

Although the petitioner did not follow this format — his attorney signed the application on his behalf14 — the case law makes manifest that the absence of a habeas petitioner's signature does not preclude the district court, in its discretion, from exercising jurisdiction over the petitioner's claims. See, e.g., Hendricks v. Vasquez, 908 F.2d 490, 491 (9th Cir.1990); Cresta v. Eisenstadt, 302 F. Supp. 399, 400 (D.Mass. 1969); cf. Rule 2(d), Rules Governing Section 2255 Proceedings (1982) (noting that a noncompliant motion "may be returned to the movant, if a judge of the court so directs" (emphasis supplied)). See generally Jamison v. United States, 244 F.3d 44, 45 (1st Cir.2001) (referencing unreported order in which this court disregarded defects of form in a section 2255 motion). In other words, the district court could have declined to consider the petition until this procedural defect was corrected, but it was free to proceed despite the defect.

The government's substantive attack is equally unconvincing. In essence, it contends that a reasonable defendant in the petitioner's situation would not have rejected the benefits obtained from pleading guilty even if he had known of the exculpatory evidence. To the government's way of thinking, these benefits included a guaranteed sentence, immunity for uncharged offenses, the value of securing a package deal to benefit his codefendants, and the opportunity to spare his family the ordeal of a protracted trial. This is a one-dimensional approach, which overlooks that other factors also bear on a defendant's decision as to whether to accept a plea agreement. Thus, before addressing the benefits of the bargain, we catalogue other factors of obvious relevance here.

To begin, the district court supportably found that the withheld evidence would have detracted from the factual basis on which the petitioner's guilty plea was predicated. See Ferrara, 384 F.Supp.2d at 423. During the change-of-plea colloquy, the judge, in determining that a factual basis existed for accepting a plea to Racketeering Act A-3 and counts 3 and 4, relied heavily on the summary of Jordan's anticipated testimony contained in the government's trial brief. In light of the petitioner's consistent assertions that he had not ordered Limoli's murder, it is fair to assume that the information in the Coleman memo would have given the judge great pause. At a bare minimum, that evidence would have substantially detracted from the factual basis that underpinned the petitioner's plea to the charges in question. With that in mind, we deem fully supportable the judge's finding that, had the exculpatory evidence been disclosed, he probably "would have found that there was not a proper factual basis to accept [the petitioner's] plea to [the Limoli murder] charges." Id.

Next, the government's trial brief and the course of the Barone trial both indicate that, had the petitioner gone to trial, the jury's assessment of Jordan's veracity would have been hugely important. Even though the petitioner had other means at his disposal for attacking Jordan's credibility (e.g., that Jordan's deal required him to testify against the petitioner), the withheld evidence was not in any sense cumulative. See Conley, 415 F.3d at 192 (explaining that impeachment evidence is not considered cumulative so long as it provides the defendant with a new method for impeaching an already impeachable witness). Thus, if Jordan had delivered the testimony promised by the government, the withheld evidence would have provided valuable ammunition for perforating his credibility. This is important because the nondisclosure of powerful impeachment evidence is apt to skew the decisionmaking of a defendant who is pondering whether to accept a plea agreement. See, e.g., id. at 189. That sort of prejudice is especially likely to transpire where, as here, the witness's testimony is both uncorroborated and vital to the prosecution's case. See United States v. Martínez-Medina, 279 F.3d 105, 126 (1st Cir.2002).

Last — but far from least — the petitioner's original counsel, Goodman, stated that, had he known of Jordan's recantation, he would have advised his client that "the allegations concerning him ordering the murder of Mr. Limoli were highly defensible." The new evidence would, at the very least, have led Goodman to seek a substantial reduction in the plea-bargained sentence as a condition of forgoing a trial.

The district court credited this declaration and found a reasonable probability that Goodman, had he been apprised of the exculpatory evidence, would have advised the petitioner not to accept the tendered plea agreement. Ferrara, 384 F.Supp.2d at 423. That finding is not clearly erroneous. It is also relevant: although the prejudice prong of the involuntariness inquiry requires the defendant to show a reasonable probability that he would have proceeded to trial as opposed to a reasonable probability that he would have received a more favorable plea bargain, evidence that defense counsel would not have recommended acceptance of the proffered plea agreement tends to support that conclusion (regardless of the lawyer's advice as to how to proceed thereafter). See Hill, 474 U.S. at 59, 106 S. Ct. 366.

Silhouetted against these realities, the government's ululations about the benefits that inured to the petitioner by reason of his plea ring hollow. In the first place, the benefits of which the government boasts are to some extent more apparent than real. For example, although the petitioner received a sentence guarantee, the length of the sentence — 264 months — hardly seems alluring.

The government's suggestion that the petitioner, even after Jordan's recantation, still would have faced a risk of being found responsible for at least one of the murders (and, thus, of receiving a life sentence) elevates hope over reason. On this record, that possibility appears remote. After all, Jordan was billed as the government's star witness with respect to the charges involving the Limoli murder. This is critically important because the petitioner was not separately charged with the DiFronzo or Corlito killings, and the government's trial brief was devoid of any information suggesting that it could prove that the petitioner played a role in either of these two homicides.15 Although the withheld evidence would not have completely erased any chance that the petitioner would be found responsible for one of the murders, we agree with the district court that a reasonable defendant in the petitioner's shoes would have viewed the odds as greatly improved. See Ferrara, 384 F.Supp.2d at 430.

The other benefices mentioned by the government — immunity, the chance to help codefendants, and the opportunity to shield his family — are all elements that a reasonable defendant would value (some more than others) when deciding whether to accept a proffered plea agreement. That is not to say, however, that the downside of forfeiting those benefits would outweigh the upside of accruing other benefits by proceeding to trial armed with the Jordan recantation. For example, if the government failed to hold the petitioner responsible for any of the murders, conviction on all the other charges would likely have yielded a guideline sentencing range (quite probably 151-188 months) well below the 264-month sentence that the plea bargain provided. In light of the greatly reduced risk of being found liable for any of the murders, there obviously would be some likelihood that a reasonable defendant in the petitioner's predicament would think that the certainty of a 264-month sentence, even when coupled with the additional benefits of entering the plea, was less desirable than going to trial.

Given this mise-en-scène, we are not at all confident that the petitioner would have chosen to plead guilty had the government disclosed Jordan's recantation to him in a timely manner. A reasonable defendant in his position might well have insisted on going to trial.

We need go no further. The sad fact is that the government promised the petitioner that it would carry out fully its obligation to produce exculpatory evidence but instead manipulated a key witness, deliberately chose not to reveal to the petitioner the stunning evidence concerning Jordan's recantation, yet represented falsely to the petitioner that it had kept its promise. This was impermissible conduct. The district court supportably found that, absent this misconduct, there was a reasonable probability that the petitioner would not have pleaded guilty but, rather, would have rejected the proffered plea agreement and opted for a trial. Ferrara, 384 F.Supp.2d at 430. Given the totality of the circumstances and the district court's credibility calls, we are constrained to conclude that the petitioner's plea was constitutionally infirm under the rule announced in Brady v. United States (a rule established substantially before the petitioner's conviction became final). Consequently, the petitioner was entitled, as the district court concluded, to collateral relief.

Affirmed.