United States v. Jonathan Jay Pollard

959 F.2d 1011 | D.C. Cir. | 1992

Lead Opinion

Opinion for the Court filed by Circuit Judge SILBERMAN.

Dissenting opinion filed by Circuit Judge WILLIAMS.

SILBERMAN, Circuit Judge:

Pursuant to an agreement with the government, Jonathan Pollard pleaded guilty on June 4, 1986 to one count of conspiracy to deliver national defense information to a foreign government. See 18 U.S.C. § 794(c). Chief Judge Aubrey E. Robinson, Jr. of the district court sentenced him to life imprisonment. Pollard did not appeal his conviction. Pollard later made an unsuccessful motion under Fed. R.Crim.P. 35 to have his sentence reduced and, again, did not appeal.

Three years after sentencing, having served the intervening time in prison, Pollard sought to attack his sentence collaterally by filing a motion under 28 U.S.C. § 2255¹ seeking the district court’s approval to withdraw his guilty plea. Pollard claimed that the government obtained his guilty plea improperly — even unconstitutionally — by linking or “wiring” his wife’s plea to his own. Anne Henderson Pollard had also been arrested in connection with Pollard’s espionage, but the government refused to enter into a plea agreement with her unless he pleaded guilty as well. Pollard also asserted that the government breached the plea agreement by the nature of its arguments (allocution) to the district court at sentencing and, furthermore, that Chief Judge Robinson based Pollard’s sentence on ex parte communications from the government. Appellant sought a hearing on this rather dramatic charge and asked Chief Judge Robinson to recuse himself. Pollard’s new attorneys in the § 2255 proceeding also sought access to certain highly classified materials the government had submitted at sentencing.

The district judge declined to recuse, denied access to the classified sentencing materials, and, without holding a hearing, refused to permit Pollard to withdraw his guilty plea. See United States v. Pollard, 747 F.Supp. 797 (D.D.C.1990). This appeal followed, and we now affirm.

I.

For a period of approximately eighteen months, from June 1984 through November 1985, Jonathan Pollard, an Intelligence Research Specialist with the United States Navy, removed large amounts of highly classified U.S. intelligence information from his office, copied it, and delivered it to agents of the Israeli government. Initially, Pollard was not paid, but during the last twelve months of this period, he met regularly with his Israeli handlers in order to receive specific tasking information, and he received between $1,500 and $2,500 per month for his efforts.

Approximately one year after the regular deliveries began, agents of the FBI and Naval Investigative Service (NIS) stopped Pollard as he was leaving work and questioned him concerning the unauthorized removal of classified information from his office. Twice during the interview, Pollard received permission to call home to his wife, Anne Henderson Pollard. In those conversations, he used the prearranged code word “cactus,” whereupon Mrs. Pollard removed a suitcase full of classified U.S. intelligence information from the Pollards’ apartment and contacted Pollard’s Israeli handlers to tell them that Pollard was in trouble. This was her only active involvement in Pollard’s espionage.

During the next two days the FBI and NIS conducted further interviews of Pollard. Pollard, in the course of these interviews, lied to his interrogators, keeping secret his involvement with the Israelis. He stalled for time, as his Israeli handlers had instructed him, and one of his handlers during this period managed to leave the country.

Jonathan Pollard was finally arrested on November 21,1985, and charged with violations of 18 U.S.C. §§ 794(a) and 793(e). Mrs. Pollard was arrested a day later as an accessory. The Pollards, prior to their arrests, had sought asylum at the Israeli embassy, actually gaining entrance to the embassy compound at one point, but the Israelis turned them away. Nonetheless, during his initial post-arrest interviews, Pollard continued to protect his Israeli handlers. Although he admitted that he had lied in his earlier interviews and had in fact been delivering classified information to a foreign government, he refused to identify the government or the names of the foreign intelligence agents controlling him. Pollard’s other Israeli handler departed the United States during this time.

Following their arrests, the Pollards were jailed without bail. Mrs. Pollard had, for several years, suffered from a debilitating gastrointestinal disorder that had not been accurately diagnosed prior to her arrest. During her stay in the D.C. jail, she was seriously ill, losing forty pounds over a period of three months. In February of 1986, Mrs. Pollard was released on bail.

Pollard had, in the meantime, begun plea discussions with the government. He sought to plead guilty both to minimize his chances of receiving a life sentence and to enable Anne Pollard to plead as well, which the government was otherwise unwilling to let her do. The government, however, was prepared to offer Pollard a plea agreement only after Pollard consented to assist the government in its damage assessment and submitted to polygraph examinations and interviews with FBI agents and Department of Justice attorneys. Accordingly, over a period of several months, Pollard cooperated with the government investigation, and in late May of 1986, the government offered him a plea agreement, which he accepted.

By the terms of that agreement, Pollard was bound to plead guilty to one count of conspiracy to deliver national defense information to a foreign government (18 U.S.C. § 794(c)), which carried a maximum prison term of life, and to cooperate fully with the government's ongoing investigation. He promised not to disseminate any information concerning his crimes without submitting to pre-clearance by the Director of Naval Intelligence. His agreement further provided that failure by Anne Pollard to adhere to the terms of her agreement entitled the government to void his agreement, and her agreement contained a mirror-image provision.

In return for Pollard’s plea, the government promised not to charge him with additional crimes, entered into a plea agreement with Anne Pollard, and made several specific representations that are very much at issue in this case. The critical provisions are paragraphs 4(a) and 4(b) of the agreement, in which the government “agree[d] as follows”:

(a) When [Pollard] appears before the Court for sentencing for the offense to which he has agreed to plead guilty, the Government will bring to the Court’s attention the nature, extent and value of his cooperation and testimony. Because of the classified nature of the information Mr. Pollard has provided to the Government, it is understood that particular representations concerning his cooperation may have to be made to the Court in camera. In general, however, the Government has agreed to represent that the information Mr. Pollard has provided is of considerable value to the Government’s damage assessment analysis, its investigation of this criminal case, and the enforcement of the espionage laws.

(b) Notwithstanding Mr. Pollard’s cooperation, at the time of sentencing the Government will recommend that the Court impose a sentence of a substantial period of incarceration and a monetary fine. The Government retains full right of allocution at all times concerning the facts and circumstances of the offenses committed by Mr. Pollard, and will be free to correct any misstatements of fact at the time of sentencing, including representations of the defendant and his counsel in regard to the nature and extent of Mr. Pollard’s cooperation. Moreover, Mr. Pollard understands that, while the Court may take his cooperation into account in determining whether or not to impose a sentence of life imprisonment, this agreement cannot and does not limit the court’s discretion to impose the maximum sentence.

The district court accepted Pollard’s plea at a hearing held on June 4, 1986. Chief Judge Robinson addressed Pollard in open court and questioned him about his understanding of the rights he was surrendering, including the potential sentences he faced and the terms of the plea agreement. After being assured by both Pollard and his attorney that there was no reason to be chary of the plea, the district court accepted it.

Sentencing took place nine months later. Both the government and Pollard submitted extensive pre-sentencing memoran-da to the district judge. In general, the government argued that Pollard had done grievous damage to U.S. national security interests in the pursuit of financial reward and that Pollard was unremorseful and a continuing danger to national security. The government mentioned that Pollard had violated his plea agreement while in prison awaiting sentencing by giving several interviews to a journalist for the Jerusalem Post, Wolf Blitzer, without first submitting his comments to the Director of Naval Intelligence. This again, according to the government, demonstrated that Pollard could not be trusted to refrain from disclosing the various national secrets in his possession and that he still considered himself the best judge of when to follow rules that had been imposed upon him. In addition to its principal sentencing memo-randa urging the district court to impose a substantial sentence, the government also submitted a highly classified declaration by Secretary of Defense Caspar Weinberger cataloguing the damage Pollard had done and opining that the damage had been “substantial and irrevocable.” The day before sentencing took place, Secretary Wein-berger submitted an unclassified supplemental declaration in response to Pollard’s submissions that stated, in a rather polemical tone, the Secretary’s contentions.

In response to the government’s submissions, Pollard argued that he had done little damage to national security because the information he had misappropriated had been delivered to one of the United States’ closest allies. He denied that he had been motivated by greed, instead claiming that he had sought to aid Israel because he believed that his aid to Israel would also benefit United States’ security interests. He stressed the extent and value of his cooperation and the sincerity of his contrition. And he emphasized the special hardships prison life would impose on him and the psychological and emotional deterioration he had experienced during confinement.

After hearing oral allocution, the district judge sentenced Pollard to life in prison and Mrs. Pollard to five years.² Pollard did not appeal but made a timely Rule 35 motion seeking to have his sentence reduced on the ground that it was disproportionate to the sentences received by other spies whose espionage was arguably more damaging to United States’ interests. He repeated the arguments he had made at sentencing and maintained in particular that the district court had failed to take proper account of his cooperation. The district court denied the motion, and Pollard did not appeal.

Anne Pollard was released on parole after serving three years in prison; she subsequently moved to Israel in August, 1990. Some months earlier, a professor at Harvard Law School, Alan Dershowitz, whose role as Pollard’s representative in this case is not entirely clear, had asked retired Supreme Court Justice Arthur Goldberg to investigate the reasons for Pollard’s life sentence. Mr. Goldberg, according to Professor Dershowitz’s affidavit, discussed the matter with Chief Judge Robinson. The Chief Judge purportedly told Mr. Goldberg that Pollard had provided Israel with information demonstrating American knowledge of certain details of Israeli-South African defense cooperation and that this had “weighed heavily” in the sentence. Der-showitz claims that Pollard never provided the Israelis that sort of information, so he asserts in his affidavit that Mr. Goldberg inferred that the government may have made improper ex parte submissions to Chief Judge Robinson.³

The present litigation began in March 1990, when Pollard filed the motion under 28 U.S.C. § 2255 to withdraw his guilty plea. Pollard sought a hearing on the allegations contained in the Dershowitz affidavit, as well as on Pollard’s claims that the government coerced his plea and breached the subsequent plea agreement. He moved to disqualify Chief Judge Robinson pursuant to 28 U.S.C. § 455 on the grounds that Chief Judge Robinson’s impartiality might reasonably be questioned and that the judge possessed personal knowledge of disputed evidentiary facts and might therefore be a material witness at the hearing. Pollard’s new counsel also sought access to the highly classified materials originally submitted by the government at sentencing, including the Weinberger declaration, in order to aid in the preparation of the § 2255 motion.

The district court rejected all of Pollard’s claims without a hearing, finding that his contentions could be resolved adequately on the files and records of the case and on the court’s judicial knowledge and recollection. See Pollard, 747 F.Supp. at 801. It found that the government had complied in all relevant respects with the terms of the plea agreement and that Pollard’s plea had been knowing and voluntary. See id. at 802-06. It denied Pollard’s counsel access to the classified sentencing materials for the same reasons that other courts have sometimes refused to permit access to pre-sentence reports: the district court found that the merits of Pollard’s substantive claims “simply do not withstand scrutiny” and that Pollard’s new counsel could, in any event, learn the information from Pollard and his former counsel. Id. at 807. The district judge refused to recuse himself, stating flatly that the allegations of ex parte contacts were untrue and that the district court’s personal knowledge of the disputed evidentiary facts was gained in a judicial capacity and so provided no basis for disqualification. See id. at 799-801.

II.

We are met at the outset with the government’s argument that some of Pollard’s claims, particularly that the government violated the plea agreement by the nature of its arguments to Chief Judge Robinson, are not properly before us in a § 2255 proceeding because they were not raised at sentencing, or in the subsequent Rule 35 motion, and because Pollard did not take a direct appeal. The government’s contention is, in essence, that those claims were waived. We think, however, that Pollard’s argument that he was unconstitutionally coerced into a plea agreement because his wife’s plea was wired with his is properly brought in the first instance in a § 2255 motion since, accepting Pollard’s allegations as true, he would have been deterred from raising that argument until his wife was out of harm’s way. If he had been induced to plead guilty because of the government’s posture vis-a-vis his wife, he could not be expected to openly challenge that very arrangement as unlawful while his wife was still vulnerable. If he had prevailed — and the plea agreement was nullified — he would be back in the same position he was in when he was “forced” to enter into the plea in the first place. Similarly, the claim that Chief Judge Robinson considered information brought to him ex ;parte by the government, based as it is on newly discovered “evidence,” could not have been raised before.

Pollard’s claims of breach of the plea agreement stand on somewhat different footing. We cannot as easily excuse his failure to raise all of those arguments. At the time of sentencing, such claims would require specific performance of the agreement, not its nullification, so his wife’s alleged vulnerability if the plea were withdrawn is not relevant. Pollard did in fact raise one of his breach arguments at sentencing and again in a Rule 35 motion (although he appeared to abandon it at sentencing), perhaps illustrating the point.

We note that the law is rather unclear as to what claims can be raised in a § 2255 motion that were not raised at sentencing, in a Rule 35 motion, or in a direct appeal from sentencing. See, e.g., Theodorou v. United States, 887 F.2d 1336, 1338-40 (7th Cir.1989); Williams v. United States, 805 F.2d 1301, 1303-06 (7th Cir.1986), cert. denied, 481 U.S. 1039, 107 S.Ct. 1978, 95 L.Ed.2d 818 (1987); United States v. Baylin, 696 F.2d 1030 (3d Cir.1982); United States v. Corsentino, 685 F.2d 48, 50-51 (2d Cir.1982). To complicate the matter even further, the government was quite imprecise in the proceeding below regarding its waiver arguments. It may well be that the government thus waived its waiver claim. See United States v. Hall, 843 F.2d 408, 410 (10th Cir.1988). The government, to be sure, repeatedly complained that Pollard’s § 2255 motion had been filed three years after he was sentenced and long after the basis for his claims should have become apparent to him, and that Pollard’s counsel had not protested the alleged breaches of the plea agreement at the sentencing hearing. But the government explicitly used the term “waiver” only with respect to Pollard’s claim that the government failed adequately to describe the nature of Pollard’s cooperation.

Ordinarily, we would feel obliged to decide whether Pollard had waived the breach of the plea agreement arguments that he presents here (although we do not see the issue as jurisdictional), but because the government was so unfocused in its argument on this issue before the district court, and because we reject Pollard’s claims on the merits, we do not think it necessary to resolve the difficult question of Pollard’s right to bring all of his claims in a § 2255 proceeding.⁴ We do, however, give due consideration to Pollard’s (and his counsel’s) reactions upon hearing the govern-mentis allocution at sentencing in determining whether the government’s allocution breached the agreement.

We are also mindful that in a § 2255 collateral challenge, an appellant, in order to gain relief under any claim, is obliged to show a good deal more than would be sufficient on a direct appeal from his sentence. Section 2255 is not a substitute for a direct appeal. See United States v. Frady, 456 U.S. 152, 165, 102 S.Ct. 1584, 1593, 71 L.Ed.2d 816 (1982). As both parties agree, we are authorized to grant relief only if we determine that the challenged sentence resulted from “ ‘a fundamental defect which inherently results in a complete miscarriage of justice,’ or ‘an omission inconsistent with the rudimentary demands of fair procedure.’ ” Fed. R.Crim.P. 32, Advisory Committee’s Note on 1983 Amendment (quoting Hill v. United States, 368 U.S. 424, 428, 82 S.Ct. 468, 471, 7 L.Ed.2d 417 (1962)).

A.

Jonathan Pollard claims that by refusing to offer Anne Pollard a plea agreement unless he also entered into a plea agreement, the government overreached; it used improper, indeed unconstitutional, pressure to force him to plead guilty. We note, however, that Pollard does not contest his guilt. He does not ask for a new trial to establish his innocence. Cf. United States v. Barker, 514 F.2d 208, 220 (D.C.Cir.) (en banc) (in deciding whether to permit withdrawal of a guilty plea, whether defendant claims innocence is an “important factor” to consider), cert. denied, 421 U.S. 1013, 95 S.Ct. 2420, 44 L.Ed.2d 682 (1975). We are not faced with the prospect that an innocent man was involuntarily compelled to plead guilty. Still, that is not dispositive. Government coercion of a eer-tain degree and kind would invalidate the plea agreement even though Pollard does not contest his guilt. See Fontaine v. United States, 411 U.S. 213, 214-15, 93 S.Ct. 1461, 1462-63, 36 L.Ed.2d 169 (1973) (per curiam).

Pollard first argues that the district judge did not strictly follow the dictates of Fed.R.Crim.P. 11, which requires a district judge accepting a guilty plea to advise the defendant in open court of the rights he is surrendering and to determine that his plea is voluntary and “not the result of force or threats or of promises apart from a plea agreement.” Fed.R.Crim.P. 11(c) & (d). Chief Judge Robinson, although he questioned Pollard at great length as to his understanding of his rights and the significance of the plea, never in haec verba asked Pollard whether his plea was voluntary. United States v. Timmreck, 441 U.S. 780, 99 S.Ct. 2085, 60 L.Ed.2d 634 (1979), holds, however, that failure to comply with the literal language of Rule 11 does not itself constitute the “complete miscarriage of justice” required before a court will grant a § 2255 motion.⁵ Id. at 784, 99 S.Ct. at 2087.

Pollard’s more substantial involuntariness argument is that wired pleas are unconstitutional. The Supreme Court has specifically reserved judgment on “the constitutional implications of a prosecutor’s offer during plea bargaining of adverse or lenient treatment for some person other than the accused.” Bordenkircher v. Hayes, 434 U.S. 357, 364 n. 8, 98 S.Ct. 663, 668 n. 8, 54 L.Ed.2d 604 (1978) (emphasis in original). The circuits that have considered the question, however, while occasionally expressing distaste for the practice, have uniformly agreed that it does not, per se, offend due process or the privilege against compulsory self-incrimination. See United States v. Marquez, 909 F.2d 738, 742 (2d Cir.1990) (citing cases from First, Fourth, Fifth, Sixth, Seventh, Eighth, Tenth, and Eleventh Circuits), cert. denied, — U.S. -, 111 S.Ct. 957, 112 L.Ed.2d 1045 (1991).

We agree with our sister circuits that plea wiring does not violate the Constitution. The question, of course, is whether the practice of plea wiring is so coercive as to risk inducing false guilty pleas. See Bordenkircher, 434 U.S. at 364 n. 8, 98 S.Ct. at 668 n. 8. To say that a practice is “coercive” or renders a plea “involuntary” means only that it creates improper pressure that would be likely to overbear the will of some innocent persons and cause them to plead guilty. Only physical harm, threats of harassment, misrepresentation, or “ ‘promises that are by their nature improper as having no proper relationship to the prosecutor’s business (e.g., bribes)’ ” render a guilty plea legally involuntary. Brady v. United States, 397 U.S. 742, 750, 755, 90 S.Ct. 1463, 1470, 1472, 25 L.Ed.2d 747 (1970) (quoting Shelton v. United States, 246 F.2d 571, 572 n. 2 (5th Cir.1957) (en banc), rev’d on other grounds, 356 U.S. 26, 78 S.Ct. 563, 2 L.Ed.2d 579 (1958)). Almost anything lawfully within the power of a prosecutor acting in good faith can be offered in exchange for a guilty plea. No constitutionally impermissible compulsion arises, for instance, when a defendant is forced to choose between the possibility of a mandatory minimum sentence of ten years in prison if he goes to trial or a suspended sentence on a reduced charge if he pleads. See Brady, 397 U.S. at 751, 90 S.Ct. at 1470. In Brady, the Supreme Court held that “a plea of guilty is not invalid merely because entered to avoid the possibility of a death penalty.” Id. at 755, 90 S.Ct. at 1472. Even where the defendant continues to maintain his innocence, having to face the death penalty as the price of trial does not invalidate a guilty plea, as long as the record contains adequate evidence of actual guilt. See North Carolina v. Alford, 400 U.S. 25, 37-38, 91 S.Ct. 160, 167-68, 27 L.Ed.2d 162 (1970).

We can understand how it might be thought that a threat of long imprisonment for a loved one, particularly a spouse, would constitute even greater pressure on a defendant than a direct threat to him. Whether one could generalize as to that proposition depends, we suppose, on one’s view of human nature. But it does not seem to be the sort of widely-shared intuition upon which a constitutional rule should be based. We must be mindful, moreover, that if the judiciary were to declare wired pleas unconstitutional, the consequences would not be altogether foreseeable and perhaps would not be beneficial to defendants. Would Pollard, for instance, have been better off had he not been able to bargain to aid his wife? Would his wife have been better off? Would the bargaining take place in any event, but with winks and nods rather than in writing?

Nor do we believe that Mrs. Pollard’s medical condition makes an otherwise acceptable linkage of their pleas unconstitutional. The appropriate dividing line between acceptable and unconstitutional plea wiring does not depend upon the physical condition or personal circumstances of the defendant; rather, it depends upon the conduct of the government. Where, as here, the government had probable cause to arrest and prosecute both defendants in a related crime, and there is no suggestion that the government conducted itself in bad faith in an effort to generate additional leverage over the defendant, we think a wired plea is constitutional. See, e.g., Politte v. United States, 852 F.2d 924, 930 (7th Cir.1988) (emphasizing “good faith prosecution”); Harman v. Mohn, 683 F.2d 834, 837 (4th Cir.1982) (probable cause and good faith); United States v. Nuckols, 606 F.2d 566, 569 (5th Cir.1979) (same). Once the government had probable cause to prosecute Mrs. Pollard and had obtained a valid indictment, it was entitled, despite her illness, to prosecute her fully — or to offer lenience for her in exchange for Pollard’s plea. See United States v. Clark, 931 F.2d 292, 294-95 (5th Cir.1991) (plea offered by man, who maintains his innocence, in order to help his “sick, pregnant and innocent” wife held not involuntary); Bontkowski v. United States, 850 F.2d 306, 313 (7th Cir.1988) (threat to prosecute validly indicted pregnant woman does not constitute unconstitutional coercion of her husband).

At minimum, Pollard argues, wired pleas raise special dangers of coercion, so that a district court faced with such a plea must undertake a more searching inquiry into the voluntariness of the plea than would normally be required. See, e.g., Nuckols, 606 F.2d at 569. Even if that were so, we are satisfied that the district court adequately discharged its obligations here. The colloquy between the court and Pollard was so extensive that there could be little doubt about Pollard’s willingness to plead. Pollard had several opportunities to confess any misgivings to the judge, but he never gave the slightest hint that his plea was anything other than voluntary. In fact, Richard Hibey, who was Pollard’s attorney at the time, brought to the court’s attention at the end of the plea proceeding a document captioned “Waiver of Trial by Jury,” which Pollard had executed. And when counsel did so, he specifically stated, “I think the Court has taken care of [the waiver] under Rule 11.”

B.

Pollard places greatest emphasis in this appeal on his arguments that the government breached the plea agreement in its allocution before Chief Judge Robinson. According to Pollard, the government made three promises to him, and it broke them all. The government stated it would seek a “substantial period of incarceration,” but agreed not to ask for a life sentence. It agreed to describe to the judge the extent of Pollard’s post-arrest cooperation and to say explicitly that it was of “considerable value” to the government’s damage assessment, as well as to its investigation of this case and enforcement of the espionage laws. And, perhaps most problematic, Pollard claims that the government agreed to limit its allocution to the “facts and circumstances of the offenses committed.”

As noted, these claims of breach of the plea agreement were brought before the sentencing district judge three years after the breaches allegedly occurred. A § 2255 motion must be brought before the same district judge who sentenced the defendant. See 28 U.S.C. § 2255. That judge had before him the plea agreement when he listened to the government’s original argument at sentencing, so he could compare the government’s arguments with the promises in the plea agreement. Looking back again, in response to appellant’s § 2255 motion, the judge determined there was no breach. The standard of appellate review of such a determination — typically employed on direct appeal from sentencing — is unsettled. Several circuits review district court determinations of whether or not a plea agreement has been breached de novo. See, e.g., United States v. Jimenez, 928 F.2d 356, 363 (10th Cir.), cert. denied, — U.S. -, 112 S.Ct. 164, 116 L.Ed.2d 129 (1991); United States v. Moscahlaidis, 868 F.2d 1357, 1360 (3d Cir.1989). Several others will set aside a district court’s finding concerning breach of a plea agreement only if clearly erroneous. See, e.g., United States v. Conner, 930 F.2d 1073, 1076-77 (4th Cir.), cert. denied, — U.S. -, 112 S.Ct. 420, 116 L.Ed.2d 440 (1991); Raulerson v. United States, 901 F.2d 1009, 1012 (11th Cir.1990); United States v. Ataya, 864 F.2d 1324, 1327 (7th Cir.1988). A plea agreement is a form of contract, but it is a rather unusual contract, because the judge plays an active role in overseeing its performance. Not only does the judge review and accept the agreement and observe both parties’ conduct under the agreement, see Fed.R.Crim.P. 11(e), the judge’s determination of the sentence is the ultimate action to which the agreement is directed. To argue, as Pollard does, that the government breached a plea agreement through excessive allocution implies that the judge tolerated, and presumably allowed himself to be influenced by, the excessive nature of the government’s arguments. The appellant’s claim, in other words, is not only that the government breached the agreement but that the district judge was swayed by the breach.

We review factual determinations of a district court (clearly erroneous) more deferentially than legal ones (efe novo), in large part because the judge who hears the evidence and observes the demeanor of witnesses has a comparative institutional advantage over the appellate court. We think the same sort of considerations typically make a deferential standard appropriate in reviewing a district judge’s determination of whether the government’s allocution violated limitations in a plea agreement. Such a determination presents a mixed question of law and fact in which the factual aspects usually predominate. The district judge is surely in the best position to determine whether the government presented an argument that, perhaps subtly, exceeded the bounds of the agreement.

When a § 2255 motion is presented challenging the government’s allocution at sentencing, the district judge inevitably examines the claim in light of his own recollection of how the government’s arguments struck him at the time they were presented. In deferring to his judgment, we defer not to his construction of the terms of the agreement but to his assessment of whether the government’s conduct violated those terms. The meaning of a plea agreement—at least its facial meaning—is of course reviewed de novo. Cf. United States v. Western Electric Co., 900 F.2d 283, 293 (D.C.Cir.1990) (reviewing construction of a consent decree efe novo). That is strictly a question of law. Once it is determined, however, that nothing the government said explicitly transgressed the promises contained in the agreement, the more subtle question, whether an illegal implicit appeal was made, which could turn on voice inflections or even facial expressions, is one concerning which we must afford deference to the district court.

It might be thought that the judge would find it difficult subsequently to recognize as improper an argument by the govemment that might have persuaded the judge. But district judges are expected to withstand improper appeals and, even when they succumb, to exercise the discipline to recognize them in hindsight. Presumably, the sentencing judge’s unique vantage point is in part why Congress provided that § 2255 motions were to be brought before the sentencing judge. See Blackledge v. Allison, 431 U.S. 63, 74 n. 4, 97 S.Ct. 1621, 1629 n. 4, 52 L.Ed.2d 136 (1977) (noting that the district judge’s “recollection of the events at issue” enable him to apply the standards for collateral relief “in a somewhat different fashion” in a § 2255 proceeding). To be sure, in Santobello v. New York, 404 U.S. 257, 92 S.Ct. 495, 30 L.Ed.2d 427 (1971), the Supreme Court said that resentencing had to take place before a different judge once it determined the government had plainly breached the plea agreement, notwithstanding the judge’s statement that he was unaffected by the government’s argument. See id. at 263, 92 S.Ct. at 499. The extraordinary nature of that relief (disqualification of a judge is not lightly granted) suggests to us, however, the appropriateness of according a large measure of respect to the district court’s judgment in deciding whether a violation occurred in the first place.

Reviewing this sort of district court finding only for clear error is entirely consistent with our approach, and the approach of other circuits, to the review of other sorts of decrees or orders in which there is substantial judicial involvement.⁶ Western Electric held only that “construction of a consent decree”—in other words “the meaning of the Decree’s terms”— should be reviewed de novo. Id. There we rejected a claim for deference based upon the fact that the district judge “was also the drafter of the language at issue.” Id. at 294. Where the issue is instead the application of that language, i.e., violation of the terms of a consent decree, courts typically review district court findings under a clearly erroneous standard. See, e.g., Kendrick v. Bland, 931 F.2d 421 (6th Cir.1991); Hayden v. Oak Terrace Apts., 808 F.2d 1269 (7th Cir.1987); United States v. Lulac, 793 F.2d 636, 643 (5th Cir.1986). Similarly, numerous courts, including our own, review findings of violations of other types of court orders in contempt proceedings under a clearly erroneous standard. See, e.g., Dallas Gen’l Drivers, Warehousemen & Helpers, Local No. 745 v. NLRB, 500 F.2d 768 (D.C.Cir.1974) (civil contempt finding by Special Master for violation of NLRB order); Glasser v. A.H. Robins Co., 950 F.2d 147 (4th Cir.1991) (violation of protective order); United States v. Costa, 947 F.2d 919 (11th Cir.1991) (violation of district court order); United States v. National Medical Enters., 792 F.2d 906, 911 (9th Cir.1986) (violation of scheduling orders); NLRB v. Bancroft Mfg. Co., 635 F.2d 492 (5th Cir.) (civil contempt finding by Special Master for violation of NLRB order), cert. denied sub nom. Croft Metals, Inc. v. NLRB, 452 U.S. 917, 101 S.Ct. 3053, 69 L.Ed.2d 421 (1981).

Here the meaning of the agreement is not in dispute; at least with respect to the first two of Pollard’s claims of breach, the question is only whether the government’s arguments transgressed the agreed norm. As to Pollard’s first claim, it is undisputed that the government never asked, in so many words, for a life sentence. Pollard contends, however, that the very force of the government’s allocution sent the district judge an implicit request for a life sentence. Pollard particularly objects to the two written submissions by Secretary of Defense Weinberger which, Pollard claims, contain numerous exaggerations and several veiled appeals to the sentencing judge to impose the maximum penalty.⁷

Secretary Weinberger’s public statement asserted that even in a year in which several spies for the United States’ main adversaries had been apprehended, it was difficult to “conceive of greater harm to national security” than that done by Pollard and that Pollard’s punishment “should reflect the perfidy of his actions, the magnitude of the treason committed, and the needs of national security.” (emphasis added). The Secretary had detailed exhaustively in his classified submission the national security information that Pollard had compromised, and he urged the court to impose a sentence “commensurate with the enduring quality of the national defense information he can yet impart.” Pollard argues that by bringing to bear its “heaviest artillery”— the nation’s top national security official— to provide the damage assessment, the prosecutors suggested to the sentencing judge the government’s interest in seeing the maximum sentence imposed.

It is true that the government pressed its case against Pollard at sentencing with considerable force. But we do not think that in doing so the government violated its promise not to recommend a life sentence; certainly we find no clear error in the district judge’s determination that the government adhered to its promise. Cf. United States v. Benchimol, 471 U.S. 453, 455, 105 S.Ct. 2103, 2104, 85 L.Ed.2d 462 (1985) (per curiam) (agreement to recommend a particular sentence does not require the government to do so “enthusiastically”). We disagree with our dissenting colleague, who believes that the Weinberger memoranda calling for “severe punishment”—reflecting the perfidy of Pollard’s actions and even referring to treason—was equivalent to an appeal for a life sentence. The Secretary of Defense’s statement was, in our view, just as consistent with a request for a long prison sentence that would be short of a life term. The government did not ask for a “life sentence in all but name,” Dissent at 1036, because it never introduced explicitly or implicitly the notion or concept of the maximum sentence; nor did it ever use words that could be thought synonymous with a life term.

The plea agreement did not commit the government to treat Pollard gently or to minimize the extent of the damage that he caused to national security. It provided: “Notwithstanding Mr. Pollard’s cooperation, at the time of sentencing the Government will recommend that the Court impose a sentence of a substantial period of incarceration_” And the same paragraph of the plea agreement stressed that “this agreement cannot and does not limit the court’s discretion to impose the maximum sentence.” Since the government reserved the right to argue for a substantial period of incarceration and explicitly preserved the district court’s discretion to impose a life sentence, Pollard’s argument that the government made an implicit plea for a life sentence necessarily rests on nuance. In this regard, we, unlike our dissenting colleague, think it telling that Pollard’s counsel, who reviewed and responded to Secretary Weinberger’s submissions in detail and heard the government’s argument, never claimed an implicit breach of the agreement not to seek a life sentence. Counsel did dispute Secretary Wein-berger’s evaluation of the gravity of Pollard’s misdeed, but characterized the Secretary’s words—perhaps accurately—only as “rank hyperbole.” The district court was, therefore, justified when it described the government as “argupng] for a substantial term of imprisonment, and no more.” Pollard, 747 F.Supp. at 803.

Pollard’s second claim, it will be recalled, is that the government breached its agreement with him by failing to outline adequately the extent and value of his cooperation. In paragraph 4(a) of the plea agreement, the government promised to “bring to the Court’s attention the nature, extent and value of [Pollard’s] cooperation and testimony,” which the government would represent was “of considerable value to the Government’s damage assessment analysis, its investigation of this criminal case, and the enforcement of the espionage laws.” In its principal sentencing memorandum, the government began its discussion of Pollard’s crimes by noting that “[i]n the numerous interviews of defendant ..., defendant has detailed the origins and scope of the espionage operation.” The government then devoted twelve pages to a detailed recitation of the conspiracy, which the government acknowledged was based largely on these interviews. Following this recitation, Pollard’s cooperation was described further. “The defendant has submitted to numerous post-plea debrief-ings_ During those debriefings, defendant revealed a substantial amount of information regarding the formation, conduct and extent of the espionage operation which was previously unknown to the government.” The government then described in general terms the value of Pollard’s cooperation: “For this reason, defendant’s post-plea cooperation has proven to be of considerable value to the government’s damage assessment analysis, and the ongoing investigation of the instant case.” And in the following paragraph it was acknowledged “that defendant has been candid and informative in describing his wrongdoing, and that [the government] has derived benefit from the information defendant has provided.”

Pollard’s claim on appeal is essentially that the government’s obligation to represent his cooperation as valuable implied a corollary promise not to discount that cooperation and not to tell the judge that his cooperation was limited. Pollard accuses the government of disingenuity by including the bargained-for language describing his cooperation in a section of the sentencing memorandum entitled “FACTORS COMPELLING SUBSTANTIAL SENTENCE.” Primarily, Pollard objects to the government telling the judge that his delay in cooperating with the investigators allowed the Israeli handlers to escape and also to the government’s suggestion that Pollard’s motive for cooperating was the desire for leniency in sentencing, not remorse for his crimes.

The government, however, never promised not to give the court a complete appraisal of its view of Pollard’s cooperation. It was of considerable value, and the government said so, but it was not wholehearted; the Israeli handlers were given an opportunity to flee the country. We disagree with the dissent’s statement that the government conveyed the impression that the value of Pollard’s cooperation was only “slight,” Dissent at 1035; instead, the government legitimately pointed out that there were certain matters of considerable importance, i.e., the apprehension of his handlers (which, of course, goes a long way to deter future espionage) concerning which Pollard did not cooperate.⁸ Pollard contends that at the time the plea agreement was reached, the government well knew the handlers had been given an opportunity to escape, and by not reserving explicitly the right to mention that to the judge, the government in effect waived the point. The argument, however, can just as easily be turned against Pollard. He could have bargained to exclude reference to the Israeli handlers. And the agreement does provide, in paragraph 4(b), that the government reserves the right “to correct any misstatements of fact at the time of sentencing, including representations of the defendant and his counsel in regard to the nature and extent of Mr. Pollard’s cooperation.” That provision at least suggests that the parties anticipated the judge would be given the full story of Pollard’s cooperation, its value to the country and where it fell short, as well as the government’s view of Pollard’s motive.

Nor do we think that the manner of the government’s description of Pollard’s cooperation or its questioning of his motive for cooperation is determinative. The form of the memorandum, to be sure, bespeaks the grudging nature of the government’s compliance, but we do not believe that the form significantly detracted from the facts presented regarding the nature of Pollard’s cooperation or from the government’s explicit statement of its value. And it is normally to be assumed that a defendant cooperates with the government largely — if not entirely — in hopes of a reduced sentence.

Pollard’s counsel initially objected at sentencing to the government’s description of his cooperation, but he subsequently abandoned this objection. See Pollard, 747 F.Supp. at 804 n. 6. Although he reasserted it in the Rule 35 motion, in neither proceeding did he make the argument presented here. He did not claim that the plea agreement barred the government from commenting on Pollard’s motive for cooperation or on Pollard’s delay that permitted the Israeli intelligence agents to escape. He contended only that the affirmative value of Pollard’s cooperation was understated. We think that although the government’s presentation was certainly not generous — it could well be thought stingy — it did not explicitly or implicitly violate the plea agreement. Again, we do not believe that the district court’s ruling on this point can reasonably be thought clearly erroneous.⁹

Pollard’s third claim of breach is, as we have indicated, more troublesome. The plea agreement in paragraph 4(b) includes the statement that the government “retains full right of allocution at all times concerning the facts and circumstances of the offenses committed by Mr. Pollard.” Pollard argues that by this language the government clearly, if implicitly, agreed not to comment on his motives for committing espionage or on his character. On numerous occasions during the sentencing hearing, the government referred to Pollard’s “arrogance and deception,” and employed other unfavorable adjectives describing his character, such as “vengeful,” “unworthy of trust,” or “contemptuous.” The government consistently suggested, moreover, that Pollard’s spying had been motivated by greed rather than ideology.

The government disputes the meaning Pollard gives to the sentence; it claims that the language is not intended to limit the government’s allocution. The only limitation the government agreed to (which itself is not explicit in the agreement) was not to ask for a life sentence or a specific term of years. One difficulty with the government’s position is the comparison with Anne Pollard’s plea agreement, which, in a parallel paragraph, states that the government “retains full right of allocution at all times, including the right to detail the facts and circumstances of the offenses committed_” (emphasis added). That language in the Anne Pollard agreement certainly suggests that a full right of allo-cution extends to something beyond facts and circumstances. Pollard also notes that in his own plea agreement at paragraph 4(c) the government reserved its “full right of allocution in connection with any Rule 35 motion,” thus indicating the parties may have meant something less than a full right by adding the facts and circumstances language in the earlier clause.

Appellant relies heavily on United States v. Moscahlaidis, 868 F.2d 1357 (3d Cir.1989), to support his interpretation of the “facts and circumstances” language as excluding discussion of his motive and character. In Moscahlaidis, the Third Circuit held that a plea agreement limiting the government’s allocution to “ ‘the full nature and extent of [the defendant’s] activities with respect to this case’ ” had been violated when the government commented at length upon the defendant’s character. Id. at 1359 (quoting plea agreement). The critical factor in Moscahlaidis, however, was that in that case, unlike our own, the government had agreed to take no position concerning defendant’s sentence, a promise which the Third Circuit held bound the government to make “ ‘no attempt at all to influence the defendant’s sentence.’ ” Id. at 1362 (quoting United States v. Miller, 565 F.2d 1273, 1275 (3d Cir.1977) (per curiam)). It was in that context that the prosecutor’s description of the defendant’s character was thought to go beyond the terms of the plea agreement.

The wording of the “facts and circumstances” clause before us, in contrast, is ambiguous. It does not, on its face, exclude commentary on Pollard’s motive for committing his crime. And, as the government correctly points out, motive is generally thought — at least for any crime that requires mens rea — to go to the very heart of a crime, and thus is presumably included in the “facts and circumstances” of the crime. The government’s discussion of Pollard’s character is, perhaps, another matter. Special protections attend the admissibility of character evidence at trial, see Fed.R.Evid. 404 and 608, and it would not be unreasonable to suppose that this type of allocution is what “facts and circumstances” excludes. However, limiting the government’s allocution to the facts and circumstances of the offenses does not necessarily preclude any comment on character, because some consideration of character is inseparable from the motivation of the defendant. In this case, for instance, it would have been artificial to permit the government to argue that Pollard committed his espionage for financial reward but rigidly exclude the suggestion that Pollard was thus greedy and self-regarding.

Given the ambiguity, Pollard contends the agreement should be construed strictly against the government, which drafted its terms, and in his favor, especially since it operates as a waiver of his constitutional jury trial rights. See, e.g., United States v. Jefferies, 908 F.2d 1520, 1523 (11th Cir.1990); United States v. Harvey, 791 F.2d 294, 301, 303 (4th Cir.1986). On the other hand, a good deal of weight must be placed on the contemporaneous interpretation of Pollard’s counsel, who apparently thought nothing amiss when the government’s allo-cution included an unflattering presentation of Pollard’s character and motive. Moscahlaidis is distinguishable on those grounds as well. The defendant in that case asserted at the time of sentencing that the government’s allocution breached the plea agreement. He also took a direct appeal from his sentence. See Moscahlaidis, 868 F.2d at 1359-60.

We find it unnecessary to decide whether the government breached the “facts and circumstances” limitation, because there is a higher and decisive barrier against which this claim of breach of the plea agreement collides. If we were to conclude that the government’s allocution extended beyond “facts and circumstances” and thus breached the agreement, we do not think Pollard would be entitled to relief under § 2255. It is settled that a § 2255 motion is not meant to be a substitute for a direct appeal and that “it does not encompass all claimed errors in conviction and sentencing.” United States v. Addonizio, 442 U.S. 178, 184-85, 99 S.Ct. 2235, 2239-40, 60 L.Ed.2d 805 (1979). The government’s allocution in Pollard’s case, even if it had crossed the limits of the plea agreement, falls far short of a “fundamental defect” in Pollard’s sentencing that resulted in a “complete miscarriage of justice”; nor was it “an omission inconsistent with the rudimentary demands of fair procedure.” Hill, 368 U.S. at 428, 82 S.Ct. at 471; cf. United States v. McKoy, 645 F.2d 1037, 1040 n. 3 (D.C.Cir.1981) (“The stringent standard for post-sentence plea withdrawal motions is intended to prevent a defendant from testing the weight of potential punishment, and then withdrawing the plea if he finds the sentence unexpectedly severe.”).

It is of course true that the government must keep the plea agreements it makes. Any breach of a promise that induced the guilty plea ordinarily entitles the defendant on direct review either to specific performance and resentencing before a different judge or to withdrawal of the guilty plea, as the court deems appropriate. See Santobello v. New York, 404 U.S. 257, 262-63, 92 S.Ct. 495, 498-99, 30 L.Ed.2d 427 (1971). But not all breaches of plea agreements can be said to result in complete miscarriages of justice; not all call for relief under § 2255. See United States v. Griffin, 816 F.2d 1, 7 (D.C.Cir.1987). Only where there has been a clear violation of a definite promise that was a significant inducement to the plea can a § 2255 court permit a prisoner to withdraw his plea. See Machibroda v. United States, 368 U.S. 487, 493, 82 S.Ct. 510, 513, 7 L.Ed.2d 473 (1962) (guilty plea is void where induced by broken promises that deprive it of the character of a voluntary act).

Our dissenting colleague suggests that the notions of “fundamental defect” and “miscarriage of justice” are elastic terms that shift in meaning depending on the nature of the relief sought (or, more accurately in this case, offered). Thus, Judge Williams contends that for the remedy he would provide, resentencing before a new judge, the test for relief under § 2255 (“fundamental defect” that results in a “complete miscarriage of justice”) can be satisfied by a lesser showing than would be required if the appellant sought rescission of the plea agreement. This is so because ordering resentencing before a new judge would not be “markedly more burdensome” than a resentencing remedy on direct appeal. Dissent at 1038. But there is no indication that Congress, when it accepted this strict standard for relief through Fed. R.Crim.P. 32(d),¹⁰ meant to delegate to judges the power to lessen it if the nominal costs to the government of a particular remedy were thought to be relatively low. (We say nominal because the disqualification of the district judge always carries immense indirect costs to the judicial system.) Even the appellant, who sought before the district court and in his opening brief to this court complete rescission of the plea agreement, seems not to have perceived the dissent’s distinction. Certainly there is no support in the case law for such a proposition.

Society’s interest in bringing criminal appeals to an end is the reason for the high standard for relief in a collateral proceeding. See, e.g., Timmreck, 441 U.S. at 784, 99 S.Ct. at 2087. This interest in finality is a good deal broader than merely avoiding the direct costs of supplemental proceedings. See, e.g., Coleman v. Thompson, — U.S. -, 111 S.Ct. 2546, 2563, 115 L.Ed.2d 640 (1991) (quoting Sanders v. United States, 373 U.S. 1, 24-25, 83 S.Ct. 1068, 1081-82, 10 L.Ed.2d 148 (1963) (Harlan, J., dissenting)). Judge Williams’ bi-level standard would disserve this interest by making it considerably more attractive for defendants to pursue collateral attacks on their sentences, and perhaps would create temptation for appellate judges to second guess trial judges’ sentencing determinations (especially in non-Guidelines cases). We think, then, the appellant must meet the statutory standard for relief, and there is simply no way that Pollard’s sentence, harsh as it is, can be thought to stem from a fundamental defect that caused a miscarriage of justice — without robbing those words of all meaning.

Judge Williams, nevertheless, would grant resentencing in this context if the government’s breach creates “a serious likelihood that a judge would have given a harsher sentence” than without the breach. Dissent at 1038-39. That approach would appear to require resentencing, regardless of the clarity or seriousness of the alleged breach, if appellate judges were to believe, for example, that a defendant would have received (from an untainted hypothetical judge) a prison term of 11 months rather than a year, 10 years rather than 11, or 40 years rather than life. In other words, once Judge Williams would conclude that the line of breach is crossed, no matter how subtly, he would find the “fundamental defect” and “miscarriage of justice” standard met by even a minor increase in the sentence over the one that would have been given by the hypothetical untainted judge. This analytical framework would either justify § 2255 relief for minor breaches of the plea agreement or it would invite appellate judges to weigh de novo the severity of the sentence imposed (or both).¹¹ For those reasons, all the courts that have considered alleged breaches of plea agreements — in a § 2255 context — have thought the statutory standard could be met only by a showing of clear breach of a significant provision in the agreement. We are aware of no case — neither appellant nor the dissent offers any — where relief was granted in a § 2255 proceeding based on an allegation that the government’s arguments amounted to an implicit breach of a partially ambiguous plea agreement. A survey of nearly one hundred appellate decisions in § 2255 cases over the last twenty years reveals only four decisions granting relief to a § 2255 petitioner based on a claimed breach of a plea agreement, and all of those cases involved clear violations of the agreement or federal statutes. See Brunette v. United States, 864 F.2d 64 (8th Cir.1988); United States v. Carbone, 739 F.2d 45 (2d Cir.1984); United States v. Corsentino, 685 F.2d 48 (2d Cir.1982); Correale v. United States, 479 F.2d 944 (1st Cir.1973). The mood, atmosphere, or “rhetoric” of the government’s allocution— upon which the dissent relies — might well justify relief on direct appeal of a sentence, but it is unlikely to satisfy the rigorous test of § 2255.

It does appear that the government was engaged in rather hard-nosed dealings with the appellant. But we think that Pollard’s claims of government breaches of the plea agreement, which appear to us to be very much the product of revisionist thinking on the part of Pollard and his new counsel, are brought far too late, in this collateral proceeding, to enable Pollard to prevail.¹² Pollard waited three years before complaining about the government’s allocution. The promise that the government arguably transgressed, to limit its allocution to the “facts and circumstances of the offenses committed,” is ambiguous with regard to the types of statements the government made. The district judge was aware of all the terms of the plea agreement and did not regard the government’s allocution as impermissible under the agreement. Even if, as the dissent argues, the contemporaneous lack of response from the district judge should be given no weight, see Dissent at 1037, the same cannot be said of Pollard’s counsel, who also made no objection to the government’s allocution. The sentence Pollard received was within the power of the district court to impose, both by the terms of the statute under which he pleaded guilty and by the explicit terms of the plea agreement. Pollard has never denied that he is guilty of the crimes for which he was imprisoned. See Barker, 514 F.2d at 220 (whether defendant asserts factual innocence is an important factor to be weighed in deciding whether to permit withdrawal of guilty plea). Nor is there any allegation that Pollard’s guilty plea was induced by the promise of a specific sentence, which he subsequently did not receive. Cf. Blackledge v. Allison, 431 U.S. 63, 97 S.Ct. 1621, 52 L.Ed.2d 136 (1977); Machibroda, 368 U.S. at 490 n. 1, 82 S.Ct. at 511 n. 1. Under such circumstances, it cannot be said that justice completely miscarried.

III.

In a quite unusual case, surely the most extraordinary claim raised by the appellant is his contention that the district judge improperly considered information conveyed to him ex parte by the government. This rather shocking charge is based on an alleged conversation between Chief Judge Robinson and former Supreme Court Justice Goldberg, who, after having been contacted by Professor Dershowitz, set out to determine the reasons for Pollard’s life sentence. The ex parte information is said to indicate that Pollard passed information to the Israelis demonstrating United States’ knowledge of secret cooperation between Israeli and South African defense agencies. The Chief Judge supposedly told Mr. Goldberg that this information “weighed heavily” in his sentencing determination. As noted earlier, Arthur Goldberg died shortly thereafter and so was not available to be questioned. But Dershowitz, we are told, has determined that Pollard did not in fact pass such information to the Israelis and that it was not included within the classified summary presented to the district judge and to Pollard’s counsel at sentencing; Dershowitz then deduces that the government must therefore have conveyed this information to Chief Judge Robinson ex parte.

Pollard sought a hearing based on the Dershowitz affidavit and, perhaps more important, asked Chief Judge Robinson to recuse himself. The Chief Judge refused to recuse and denied the hearing. Pollard claims the district court abused its discretion. We think very little of this claim.

A district judge must grant a prompt hearing under § 2255 unless “the motion and the files and records of the case conclusively show that the prisoner is entitled to no relief.” 28 U.S.C. § 2255. The decision whether to do so is committed to the district court’s discretion. See Machibroda, 368 U.S. at 495, 82 S.Ct. at 514. In making this decision, “the judge’s recollection of the events at issue may enable him summarily to dismiss a § 2255 motion”; indeed, his ability to do so is one of the advantages of § 2255 relative to habeas corpus for state prisoners. Blackledge, 431 U.S. at 74 n. 4, 97 S.Ct. at 1629 n. 4. Only where the § 2255 motion raises “detailed and specific” factual allegations whose resolution requires information outside of the record or the judge’s “personal knowledge or recollection” must a hearing be held. Machibroda, 368 U.S. at 495, 82 S.Ct. at 514. Even if the files and records of the case do not clearly rebut the allegations of the prisoner, no hearing is required where his claims are “vague, conclusory, or palpably incredible.” Id.

The district judge in this case was, of course, intimately familiar with the record and knew as a matter of fact whether or not the government had submitted ex parte material to him. The Chief Judge attested that it had not. He accordingly denied the motion for a hearing.¹³ We believe the Chief Judge was well within his discretion to do so. The Dershowitz affidavit was a very weak submission. It did not suffice to require a hearing into the question it raised, even if the district judge had not been personally aware that Professor Der-showitz’s “deductions” were fallacious. The affidavit included no direct or even hearsay evidence that an ex parte submission had been received. Furthermore, after examining the classified material submitted to the district judge, we understand how the subject of South Africa (if not the Jericho missiles mentioned in the affidavit) could have been mentioned in a conversation between the Chief Judge and Arthur Goldberg, without recourse to any ex parte materials.¹⁴

For similar reasons, we believe that Chief Judge Robinson did not abuse his discretion in denying Pollard’s motion for disqualification. Pollard argues that Chief Judge Robinson was required to recuse himself under 28 U.S.C. § 455(b)(1), because he had personal knowledge of disputed evidentiary facts concerning the alleged ex parte contacts; under § 455(b)(5)(iv), because he might have been a material witness at a hearing on those allegations; and under § 455(a), because his deciding the motion alleging misconduct on his part gave the appearance of bias. The district court did not need to recuse itself under § 455(b)(1), because only personal knowledge of disputed evidentiary facts gained in an extrajudicial capacity is grounds for recusal, and any knowledge Chief Judge Robinson had concerning the government’s sentencing submissions came directly from his participation in the case. See United States v. Heldt, 668 F.2d 1238, 1274 (D.C.Cir.1981) (per curiam), cert. denied, 456 U.S. 926, 102 S.Ct. 1971, 72 L.Ed.2d 440 (1982). Likewise, the district court could not be disqualified for bias, because the bias alleged “must stem from an extrajudicial source and result in an opinion on the merits on some basis other than what the judge learned from his participation in the case.” United States v. Grinnell Corp., 384 U.S. 563, 583, 86 S.Ct. 1698, 1710, 16 L.Ed.2d 778 (1966); Heldt, 668 F.2d at 1274. Finally, Chief Judge Robinson was not required to recuse himself as a potential material witness, because, as we noted above, he acted within his discretion in denying a hearing.¹⁵

* * * * * *

The issue before us as appellate judges is not whether a life sentence was appropriate punishment for Pollard’s crime, still less whether we ourselves would have imposed such a sentence. It is rather whether the appellant has mounted a sufficient challenge to the actions of the government and the district judge to clear the formidable barriers to relief in a collateral attack on his sentence under § 2255. We think not. Accordingly, we affirm the judgment of the district court.

.It is so ordered.

1 . 28 U.S.C. § 2255 provides in pertinent part: A prisoner in custody under sentence of a court established by Act of Congress claiming the right to be released upon the ground that the sentence was imposed in violation of the Constitution or laws of the United States, or that the court was without jurisdiction to impose such sentence, or that the sentence was in excess of the maximum authorized by law, or is otherwise subject to collateral attack, may move the court which imposed the sentence to vacate, set aside or correct the sentence.

2 . It should be noted that this case arose, and sentence was imposed, prior to the effective date of the United States Sentencing Guidelines.

3 . Mr. Goldberg was unable to prepare his own affidavit or to testify to the matters Professor Dershowitz described, because he died on January 18, 1990, a few days after he received a letter from Dershowitz in which Dershowitz informed him of the results of the inquiries Der-showitz had made.

4 . Our dissenting colleague sidesteps the knotty problems involved in finding cause and prejudice sufficient to excuse Pollard’s failure to object at sentencing, because he contends the government "waived the defense" below. Dissent at 1032. Judge Williams may be correct that the government merely "grazed the issue,” id., as to some of Pollard’s claims, but the government addressed it directly as to Pollard’s claim that the government failed to describe adequately his cooperation. No mere "vague footnote throwaway line,” id., the government’s statement concerning this claim of Pollard’s said:

"We doubt that the defendant has preserved this point for review. As he raised it explicitly during the sentencing hearing, and then abandoned it, we believe he has waived the right to argue it on collateral attack under § 2255. Nonetheless, we respond to these claims on the merits.”

Opposition to Defendant’s Motion to Withdraw His Guilty Plea at 24 n. 10.

5 . Pollard further suggests that his wife’s illness was an "aggravating factor" of the type the Court suggested in dicta in Timmreck might expand a § 2255 court’s scope of review. See Timmreck, 441 U.S. at 784-85, 99 S.Ct. at 2087-88 ("we find it unnecessary to consider” whether aggravating circumstances might change the result). Case law does not indicate that her illness is the kind of factor the Court contemplated. Cf. Johnson v. United States, 805 F.2d 1284, 1287 (7th Cir.1986) (stating that “aggravating factor,” if it means anything at all, means only factors that rise to the level of a due process violation); United States v. Laura, 667 F.2d 365, 379 (3d Cir.1981) (Stern, J., dissenting) (“being advised by a lawyer with a conflict of interest relating directly to [the] plea” is an aggravating circumstance).

6 . Indeed, it comports with our review of ordinary contracts. When issues of interpretation and construction predominate in an ordinary breach of contract claim, we regard the issue of breach as a question of law to be reviewed de novo. When, however, a breach determination rests primarily on an analysis of facts, we will reverse the determination only if clearly erroneous. See, e.g., ARB, Inc. v. E-Systems, Inc., 663 F.2d 189 (D.C.Cir.1980); W.G. Cornell Co. v. Ceramic Coating Co., Inc., 626 F.2d 990 (D.C.Cir.1980).

7 . Pollard, for purposes of this appeal, received a redacted version of Secretary Weinberger’s classified memorandum.

8 . The dissent’s suggestion that the district judge in the § 2255 proceeding may have improperly interpreted the plea agreement to have permitted the government to "convey the message that Pollard’s cooperation, while containing some elements of considerable value, was on an overall basis not worth much," Dissent at 1035, is not a fair reading of the district court’s opinion. After noting that the government agreed to represent that Pollard’s cooperation had been of considerable value, see Pollard, 747 F.Supp. at 802, the district court went on to observe that

Defendant claims that by “casting aspersions” on defendant’s cooperation, the Government completely undercut the statements it made earlier. For example, the Government noted that defendant’s delay in cooperating allowed certain co-conspirators to flee the United States. It was no violation of the plea agreement for the Government to explain the positive value of the cooperation in one sense (damage assessment), while also noting that defendant had frustrated Government efforts in another sense (law enforcement). The record in this case does not support the contention that the Government failed in its obligation. It did not dryly recite in a few declarative sentenced] that defendant had cooperated. It did much more, as it had said it would.

Id. at 804.

9 . Although our dissenting colleague agrees that the district court’s judgment concerning these claimed violations of the plea agreement must be reviewed for clear error, his application of that standard appears indistinguishable from de novo review. See Dissent at 1035-36.

10 . The "complete miscarriage of justice” standard was originally a judicial gloss on § 2255. See Hill v. United States, 368 U.S. 424, 428, 82 S.Ct. 468, 471, 7 L.Ed.2d 417 (1962). When rule 32(d) was amended in 1983, the Advisory Committee Notes expressly approved that standard for § 2255 proceedings, and Congress authorized the amendment to take effect without change.

11 . In a direct appeal of a sentencing determination, if the appellate court concludes that the government breached a plea agreement, it grants relief. The appellate court does not try to decide whether the breach caused the judge to give a greater sentence than would have been levied otherwise. As the Santobello Court recognized, that is an unprofitable line of inquiry because the trial judge — as did the district judge here — may well state that he was unaffected by the arguably excessive nature of the government's argument. See Santobello, 404 U.S. at 262-63, 92 S.Ct. at 498-99. An appellate court should not be obliged to weigh the trial judge’s credibility in the course of considering whether the breach had any effect.

12 . Pollard also claims that the government breached its implied covenant of good faith and fair dealing by "permitting” Pollard to meet with journalist Wolf Blitzer without forcing Pollard to seek permission from the Director of Naval Intelligence, as his plea agreement called for, and then bringing up this violation of the plea agreement before the judge. Because the government was under no duty to police Pollard's compliance with the plea agreement, and because there is no evidence at all that the government connived to bring about this violation, we do not see any substance to this claim.

13 . The motion may have been designed primarily to force the judge to recuse himself.

14 . Indeed, on pages 34-36 of Pollard’s own Second Memorandum In Aid Of Sentencing, there appears a section entitled “The South African Affair,” discussing Pollard’s involvement in intelligence-sharing with South Africa.

15 .Assuming arguendo that the district judge’s refusal to direct the United States Attorney to provide Pollard’s new counsel with access to the Weinberger classified submission was erroneous, our examination of the material satisfies us that the error was harmless.

---

Concurrence in Part

WILLIAMS, Circuit Judge,

dissenting in part:

I agree with the majority that the “plea wiring” was not an unlawful coercion of Pollard’s guilty plea and that Chief Judge Robinson did not abuse his discretion in refusing to recuse himself or to conduct a hearing into the claim of ex parte contacts. But because the government's breach of the plea agreement was a fundamental miscarriage of justice requiring relief under 28 U.S.C. § 2255, I dissent.

Before turning to the specifics of the government’s failure to behave at sentencing as it had promised, there are some preliminary matters.

1. “Cause and prejudiceRelief is to be granted under 28 U.S.C. § 2255, the statutory replacement of common law ha-beas, if the petitioner shows a “fundamental defect [resulting in] a complete miscarriage of justice” or “an omission inconsistent with the rudimentary demands of fair procedure.” Hill v. United States, 368 U.S. 424, 428, 82 S.Ct. 468, 471, 7 L.Ed.2d 417 (1962). If the petitioner failed to lodge a contemporaneous objection to the challenged ruling, and failed to appeal, then he may also have to show “cause and prejudice”, the first to excuse his “double procedural default”, the second to show the actuality of injury from the error. See United States v. Frady, 456 U.S. 152, 167-68, 102 S.Ct. 1584, 1594-95, 71 L.Ed.2d 816 (1982). The government argues that Pollard must show cause and prejudice because he failed to object to the government’s allocution and to appeal his sentence. Because the majority believes that Pollard has not satisfied even the “fundamental defect” standard, it need not address “cause and prejudice”, see Maj.Op. at 1019-20, but I must.

As the government at no point below asserted the cause and prejudice standard, it has waived the defense. At best the government may be said to have grazed the issue. In its Opposition to Motion to Withdraw Guilty Plea it noted the three-year delay between sentencing and Pollard’s § 2255 motion, id. at 5-6, the failure of Pollard to object to the government’s allo-cution at the sentencing hearing, id., and Pollard’s “waiver” of his claim that the government failed to describe his cooperation adequately, id. at 24 n. 10. But the government made the first two points solely to boost an argument that Pollard did not at the time of allocution consider its conduct a breach, and it made the last in a vague footnote throwaway line without mention of the “cause and prejudice” standard or citation to any case invoking that standard. A party demanding strict adherence to principles of finality is in a weak position to ask that its own lapses be disregarded, and accordingly the circuit courts have treated government silence as a waiver. See, e.g., United States v. Hicks, 945 F.2d 107, 108 (5th Cir.1991); United States v. Hall, 843 F.2d 408, 410 (10th Cir.1988); United States ex rel. Bonner v. DeRobertis, 798 F.2d 1062, 1066 (7th Cir.1986); but cf. Titcomb v. Virginia, 869 F.2d 780, 782-84 (4th Cir.1989) (no waiver where “it would appear” that the state raised the issue in its second response and where the evidence of the procedural default was part of record). Finding waiver is especially sensible where, as here, relief can take the form of an order merely requiring resen-tencing, not a vacation of the plea, so that the government is most unlikely to suffer any injury from loss of memories or other evidence.

2. Deference: My understanding of the scope of our review is largely similar to the majority’s. We review a trial court’s interpretation of a plea agreement de novo, just as we review an interpretation of an ordinary contract de novo (unless the interpretation turns on extrinsic evidence). See, e.g., HOH Co. v. Travelers Indem. Co., 903 F.2d 8, 12 n. 6 (D.C.Cir.1990) (contract interpretation); United States v. Western Electric Co., 900 F.2d 283, 294 (D.C.Cir.1990) (interpretation of civil settlement); see also Maj.Op. at 1022-23. We review a trial court’s findings of pure fact (what happened) under the clearly erroneous standard. And we defer to district court findings even on fact-intensive issues of characterization, such as whether particular conduct, about which there is no factual dispute, violates a norm that courts have defined with as much clarity as they can expect to achieve. Thus in Kendrick v. Bland, 931 F.2d 421 (6th Cir.1991), the court applied the clearly erroneous standard to findings as to whether specific conduct by prison officials — e.g., cancellation of club activities — constituted “institution-wide” violations of a consent decree.

The appellate deference in this last category is somewhat anomalous, as the characterization is in a way part of the law-finding process; it answers a normative question (whether the conduct was acceptable or not), and thus in some measure refines the norm. But where the characterization process is extremely fact-intensive, the appellate decision may not yield a useful contribution to the formulation of legal rules, so that appellate judges’ duplication of the trial court’s effort would consume judicial resources for very little return.¹ See Mars Steel Corp. v. Continental Bank N.A., 880 F.2d 928, 933-36 (7th Cir.1989) (en banc) (Easterbrook, J.). By the same token, however, review for clear error cannot fade off into complete deference, else the appellate court’s law-finding (or law-forming) function would be disconnected from the real world. Moreover, presumably because of a recognition of the law-forming aspect of fact-intensive characterizations, the decisions on deference appear replete with contradictions and inconsistency. See Wright & Miller, Federal Practice and Procedure §§ 2585-89 (1971 & Supp.1990). Perhaps one can do no better than to say that there is deference, but that the special character of the practice informs its nature.

Where a defendant claims that the government’s allocution breached a plea agreement, there is no deference to the sentencing judge’s view of the actual effect of the allocution on him or her (as opposed to the projected or inferred effect on a hypothetical judge). Compare Maj.Op. at 1022-23. This is for the simple reason that the actual effect on the actual judge is irrelevant. Santobello v. New York, 404 U.S. 257, 92 S.Ct. 495, 30 L.Ed.2d 427 (1971), put any such inquiry out of bounds, perhaps because the sentencing judge would be in an awkward position making “findings” about his own state of mind. The Court ordered relief despite declaring that it had “no reason to doubt” the judge’s statement that “the prosecutor’s recommendation did not influence him”. Id. at 262-63, 92 S.Ct. at 498-99.

3. Principles of plea interpretation: The majority correctly notes that Pollard’s argument in some respects “rests on nuance”, Maj.Op. at 1025, and that the government did not ask for a life sentence “in so many words”, id. at 1024, or use “words ... synonymous with a life term”, id. at 1024. Without explicitly setting forth a general approach, the majority appears to apply a principle that the government should be held only to the letter of its bond.

Such an approach puts an impossible burden on contract drafters. A party cannot anticipate every evasive move that another party may make, or every opportunity for evasion. Brittle interpretation makes for long contracts, and even the longest will have loopholes. Accordingly, as with other contracts, courts ordinarily reject literalism and read plea agreements as a reasonable person would understand them. United States v. Moscahlaidis, 868 F.2d 1357, 1361 (3d Cir.1989) (citing cases); United States v. Carbone, 739 F.2d 45, 46 (2d Cir.1984); United States v. Crusco, 536 F.2d 21, 26 (3d Cir.1976) (criticizing “stubbornly literal” reading of plea agreement).² Finding an implicit duty of good faith, though it may sound more drastic, amounts to the same thing if it is understood as “a stab at approximating the terms the parties would have negotiated had they foreseen the circumstances that have given rise to their dispute.” Market Street Associates Ltd. Partnership v. Frey, 941 F.2d 588, 595 (7th Cir.1991). Thus in United States v. Bowler, 585 F.2d 851, 853-54 (7th Cir.1978), the court construed a provision that the government’s sentencing recommendations “may be reduced, based upon [various] personal factors” to include an implicit promise to consider the factors specified. See also United States v. Brown, 500 F.2d 375 (4th Cir.1974) (prosecutor’s statement to the judge that he recommended three years because of the plea agreement, though he had “some problems” with this sentence, violates agreement to recommend no more than three years); Snowden v. State, 33 Md.App. 659, 365 A.2d 321 (1976) (promise to recommend probation breached by prosecutor who did so but, when asked by the judge whether he still adhered to this promise, said, “I believe we must, I believe we must”).

Judicial insistence on a reasonable and not a niggling interpretation for plea agreements is hardly surprising. The defendant has given up his right to a trial that satisfies due process, so if a plea rests “in any significant degree on a promise or agreement of the prosecutor ..., such promise must be fulfilled.” Santobello, 404 U.S. at 262, 92 S.Ct. at 498. If fulfillment of the promise is to mean anything, it cannot refer only to the promise pared to its literal bone. While the government is a repeat player in this field and has an interest of its own that its word be good, experience in the commercial world shows that even repeat players lapse. Here, too, reasonable judicial enforcement not only assures fairness to the individual who may be the victim of a slip-up or of an agent’s excessive zeal, but also makes it easier for the government to build confidence in its word.

# * Sjt J¡C * *

Pollard’s plea agreement required him to plead guilty and to cooperate. On its side, the government made three promises of significance here. First, it would bring to the court's attention “the nature, extent and value of [Pollard’s] cooperation and testimony” and would represent that the information supplied was of “considerable value to the Government’s damage assessment analysis, its investigation of this criminal case, and the enforcement of the espionage laws.” Maj.Op. at 1017. Second, it would not ask for a life sentence (this promise was implicit but is not contested by the government), though it would be free to recommend a “substantial period of incarceration”. Id. at 1016-17. Third, the government limited its reserved right of allocution to “the facts and circumstances” of Pollard’s crimes. Id. at 1017. The government complied in spirit with none of its promises; with the third, it complied in neither letter nor spirit.

Though the government obligated itself to call attention to the “considerable value” of Pollard’s cooperation, in its principal sentencing memorandum it buried its sole discussion of that cooperation in a section entitled “FACTORS COMPELLING SUBSTANTIAL SENTENCE”. The first paragraph of this section discussed the extent of Pollard’s offense, saying that he had compromised “thousands of pages of classified documents”, and recommended a “substantial period of incarceration”. J.A. 160. The second paragraph called Pollard’s activities a “flagrant breach of ... trust”, a breach “all the more venal in that [despite Pollard’s contrary claims] it is clear that the money and gifts provided by the Israelis were significant, if not the primary factors motivating defendant”. Id. at 161. The third paragraph said that Pollard “will undoubtedly urge the Court also to consider his post-arrest conduct, i.e., [his] submission of a plea of guilty and his cooperation ...and noted the obvious point that plea bargaining and cooperation “may be considered by courts at the time of sentencing.” Id. In paragraph four the government provided nominal compliance with its promise, saying that Pollard “revealed a substantial amount of information regarding ... the espionage operation which was previously unknown to the government” and that this cooperation “has proven to be of considerable value to the government’s damage assessment analysis, and the ongoing investigation of the instant case”.³ Id. at 162. In the fifth paragraph, though acknowledging that the defendant had been candid and informative, the government told the court that Pollard delayed his cooperation in order to assist the escape of three coconspirators, devoting more space to this caveat than to its favorable words for Pollard’s cooperation. Id. The rest of the section, of course, went on with further denunciations.

Thus the government came forth with the magic words “of considerable value”, and it even mentioned two of the three general areas of inquiry, specified by the agreement, to which Pollard’s cooperation contributed. But by placing the discussion square in the middle of its reasons why the sentence should be substantial, and by its heavy stress on the cooperation’s imperfections, it succeeded in conveying the impression that, overall, the value was not “considerable” but slight. Perhaps the value was slight, but if so, then the government should not have embraced an obligation to say the contrary. In United States v. Fisch, 863 F.2d 690 (9th Cir.1988), the court found a plea agreement violated where the government, though obliged to tell the court of defendant’s cooperation, neither offered details on its own nor endorsed the defense account, saying only that defendant had “been cooperating in helping prosecute the coconspirators”. The government’s breach here is no less.

It is not clear from the district court’s rejection of this claim whether the court construed the agreement as leaving the government free to convey the message that Pollard’s cooperation, while containing some elements of considerable value, was on an overall basis not worth much, or whether the court found that the government had said that the cooperation was of considerable value overall. See United States v. Pollard, 747 F.Supp. 797, 804 (D.D.C.1990). If the first, I think the contract interpretation was wrong; if the second, I believe the finding was clear error.

On the promise not to ask for a life sentence, the government coupled its adherence to the letter with an even more flagrant violation of the agreement’s spirit. It presented memoranda from Secretary of Defense Weinberger saying that “no crime is more deserving of severe punishment than conducting espionage activities against one’s own country”, J.A. 467, that “it is difficult for me ... to conceive of a greater harm to national security than that caused by the defendant”, id. at 263, and that “the punishment imposed should reflect the perfidy of [his] actions, the magnitude of the treason committed, and the needs of national security”, id. at 264.

While these remarks did not expressly endorse a life sentence (or use a synonym, compare Maj.Op. at 1024-25), the repeated use of superlatives implied an appeal for the maximum. Weinberger’s reference to treason took the point further. Whereas treason carries the death penalty, 18 U.S.C. § 2381, and involves aiding the nation’s enemies, U.S. Const., Art. Ill, § 3, cl. 1, Pollard was charged with espionage, carrying a maximum of life imprisonment and encompassing aid even to friendly nations — here, Israel. Of course the sentencing judge knew the difference, but the government’s barrage expressed a viewpoint that the government had promised not to express. Weinberger’s subtext was that the heaviest possible sentence was the lightest that was just. The trial court’s conclusion to the contrary, 747 F.Supp. at 803, was clearly erroneous.

That the government had reserved the right to seek “a substantial period of incarceration” does not change the analysis. Compare Maj.Op. at 1017. Of course the government remained free to lay out the details of the crime and its impact on national security. These, coupled with an explicit plea for a substantial sentence, might well have secured the government’s objective. But the availability of these methods scarcely entitled it to wheel out the heaviest rhetorical weapons, calling for a life sentence in all but name.

Finally, despite its agreement to confine its allocution to “the facts and circumstances” of the offenses, the government told the district judge that Pollard’s expressions of remorse were “both belated and hollow”, J.A. 371, and “grounded in the fact he was caught” (emphasis in original), id.; that Pollard was a “recidivist” who was “contemptuous of this Court’s authority” and “unworthy of trust”, id. at 359; that Pollard felt “blind contempt” for the U.S. military, id. at 318, and had a “warped” and “skewed” perspective, id. at 319; that Pollard was “traitorous”, id. at 369, “arrogant [and] deceitful”, id. at 315 (and see id. at 316, 320), “without remorse”, id. at 179 (and see id. at 163), and “literally addicted to the high lifestyle funded by his espionage activities”, id. at 167. The assistant U.S. Attorney noted that he (the assistant) had been brought up to regard two sins as “unforgivable”, arrogance and deception — id. at 313, precisely the two sins that he repeatedly imputed to Pollard. Pollard’s “loyalty to Israel transcends his loyalty to the United States,” said Secretary Weinberger. Id. at 265. The government devoted much space to marshalling evidence that Pollard was driven by greed (“enamored of the prospect for monetary gain”, id. at 165; motivated by “the lure of money”, id. at 168), and not materially affected by anti-terrorist concerns, id. at 164-69, or, by implied extension, by any sympathy for Israel.

The government contends that in the phrase by which it retained “full right of allocution at all times concerning the facts and circumstances of the offenses”, the limiting reference to “facts and circumstances” was a nullity. This is hard to swallow. As the majority points out, the contrast with the language in Anne Pollard’s plea agreement suggests that here the parties intended to exclude some otherwise acceptable elements of an allocution. See Maj.Op. at 1027.

I agree with the majority that “facts and circumstances” include matters going to Pollard’s motivation. But one can address motive by detailing specifics, leaving the moral and legal conclusions alone, to be settled by the judge. This is the line drawn by the court in Moscahlaidis. There the government was obliged by its agreement to “take no position” on the sentence, but was allowed to speak of “the full nature and extent of [the defendant’s] activities with respect to this case”. 868 F.2d at 1359. The court found that the agreement barred the government from asserting its opinion of the defendant’s character with phrases alluding (for example) to “the depth of [his] greed and moral bankruptcy” and a “demonic pursuit demonstrating] [his] utter contempt for the welfare of his fellow man”, id. at 1362. Though the structure of the agreement was different from Pollard’s, the Third Circuit’s line — between fact and opinion— makes equal sense here.⁴

So the government was free to relate not only the intelligence implications of Pollard’s acts, but also details supporting an inference that his motive was pecuniary. But if the limit meant anything, it could not allow the government to wrap the raw facts in an inflammatory rhetoric, endlessly alluding to its (necessarily subjective) opinions that Pollard was greedy and immoral, depicting his conduct as the apogee of espionage, naming him a traitor, and delivering a tirade on his “arrogance and deceit”.

Taken together, the government’s three promises worked a substantial restraint on the government’s allocution. Its commitments to restrict itself to facts and circumstances, and to assess Pollard’s cooperation as having considerable value, closed off a means by which it might demand a life sentence in all but name. Safely after the fact, the government’s briefing here undermines its commitment, isolating the components of the promise in order to conceal their synergies.

Is it troubling that the breaches involve matters of rhetoric? It means, of course, that the violation cannot be measured mathematically. But that is often true in contract disputes — most obviously whenever courts enforce express or implied duties of “reasonable” performance. Further, courts frequently draw the line between facts and legal or moral conclusions in classifying the testimony of lay witnesses. See Fed.R.Evid. 701; compare United States v. Slade, 627 F.2d 293, 305 (D.C.Cir.1980) (finding lay references to defendant’s drug “organization” to be equivalent of improper assertion of conclusion that it was conspiracy) with Williams Enterprises, Inc. v. Sherman R. Smoot Co., 938 F.2d 230, 233-34 (D.C.Cir.1991) (permitting lay testimony by insurance broker on relation between accident and insured’s increased premium). Is rhetoric simply too trivial? To say so would, I think, buck prevailing views of human psychology. When the majority relies on the district judge’s failure to note a breach at the time of the allocution, see Maj.Op. at 1029-30, it assumes that the judge was unaffected by the government’s efforts to disparage Pollard’s character, thereby resolving an unanswerable question and perhaps violating Santobello, which ordered relief despite the trial judge’s assertion — accepted by the Court — that the improper allocution had not influenced him. 404 U.S. at 262, 92 S.Ct. at 498.

Do the breaches amount to the “complete miscarriage of justice” required for relief under § 2255? The cases under § 2255 (and § 2254),⁵ quite frankly, leave a gap between two patterns — conduct clearly violating even the literal language of the agreement (eliciting relief), and conduct varying from the agreement only trivially (eliciting none). Typical of the first pattern is Brunette v. United States, 864 F.2d 64, 65 (8th Cir.1988), where despite an agreement to recommend an “unspecified period”, the government expressly suggested the maximum. See also United States v. Birdwell, 887 F.2d 643 (5th Cir.1989) (where incorporation of a state plea agreement into a federal one induces defendant to accept the latter, the state authorities’ failure to follow through undermines the federal plea); Smith v. Blackburn, 785 F.2d 545 (5th Cir.1986) (state’s promise that defendant would be released upon parole breached when parole was denied); Carbone, 739 F.2d 45 (prosecutor’s objection to defendant’s request for sentence allowing early parole violated promise to make no recommendation); United States v. Mercer, 691 F.2d 343 (7th Cir.1982) (failure by court to allow defendant to withdraw plea, when conditions specified in the agreement as permitting withdrawal occurred, required vacation of plea); Correale v. United States, 479 F.2d 944 (1st Cir.1973) (where prosecutor promised to recommend sentence that would make defendant eligible for federal parole when he received parole under state sentence, noncomplying recommendation breached promise and entitled defendant under § 2255 to resentenc-ing).

On the other side is United States v. Benchimol, 471 U.S. 453, 105 S.Ct. 2103, 85 L.Ed.2d 462 (1985), where the government had agreed to recommend probation, and, when defense counsel informed the court of the agreement, the assistant U.S. attorney said, “That is an accurate representation.” Id. at 455, 105 S.Ct. at 2104. The Court refused to upset the sentence under § 2255, viewing it as a case where the prosecutor simply “left an impression ... of less-than-enthusiastic support for leniency.” Id. at 456, 105 S.Ct. at 2105 (inner quotations omitted). Benchimol scarcely helps the government much here. It is one thing to say that the government need not fulfill its obligations with gusto, or with the right tone of voice, and quite another to say that it may fill its allocution with ardent declamations that wholly undermine any nominal compliance.

On the facts the closest case is Moscah-laidis, where the government similarly loaded its allocution with inflammatory rhetoric inconsistent with the reasonable meaning of the agreement. See p. 1036 above. There the court, on appeal, ordered relief. As the “miscarriage of justice” standard is more demanding than the standard of review on appeal, Moscahlaidis would not control this case even in the Third Circuit. But how much more should be required? In the nature of things the difference in standards is elusive. It arises, moreover, out of the fact that in the ordinary habeas case the defendant’s prior failure to object makes any remedy more costly for the system than it would have been: in a new trial, the government will be hobbled by the staleness of its evidence and the risk of an erroneous acquittal will be higher. Where the remedy is not markedly more burdensome than it would have been on appeal (resentencing only), the increment implicit in “fundamental defect” should be correspondingly modest.

It is hardly surprising that the exact stringency of the “fundamental defect” standard should vary with the context. The scope of a right may turn on the character of the remedy sought. Much of the law of equitable remedies revolves around the point that courts may treat the same conduct as actionable when the plaintiff seeks damages and not actionable when he seeks an injunction. See, e.g., Restatement (Second) of Torts § 941 & comment c (1979) (injunctive relief should be withheld for nuisances where balance of hardships so dictates, even though damages are to be awarded); Harrisonville v. W.S. Dickey Clay Manufacturing Co., 289 U.S. 334, 53 S.Ct. 602, 77 L.Ed. 1208 (1933) (same); York v. Stallings, 217 Or. 13, 341 P.2d 529 (1959) (same); Stuttgart Electric Co., Inc. v. Riceland Seed Co., 33 Ark.App. 108, 802 S.W.2d 484 (1991) (encroachment warrants remedy of damages, not of injunction). In ordinary contract law (the closest parallel to plea agreement rules), an obligee who like appellant fails to protest a breach immediately may lose his right to cancel the agreement (paralleling vacation of the plea) but not his right to damages (paralleling resentencing). See, e.g., UCC §§ 2-711, 2-714 (buyer has option upon delivery of rejecting nonconforming goods or of obtaining damages; within reasonable time after delivery he may only obtain damages). Further, exactly the same phrase may have different meanings even within the same rule. While appellate courts review all factual findings of district courts under the “clearly erroneous” standard, whether “based on oral or documentary evidence”, Fed.R.Civ.P. 52(a), the Rule also instructs that “due regard shall be given to the opportunity of the trial court to judge of the credibility of the witnesses”, giving the latter type of evidence a special status.

The majority suggests that as the Advisory Committee notes on the 1983 amendment to Fed.R.Crim.P. 32(d) endorsed the “complete miscarriage of justice” standard for § 2255, Congress’s failure to reject the amendment affirms that view of § 2255. Maj.Op. at 1028 n. 10. The argument is quite persuasive as to plea withdrawals, which are the sole subject of Rule 32(d) and (therefore) of the Committee’s discussion, but the sequence surely affords little evidence that Congress addressed either the standard for § 2255 requests for resen-tencing, or the character of the “fundamental defect” standard as applied to such relief. Here, then, as a remand for resen-tencing would preserve the plea, that relief should be given if the government’s misconduct in allocution — the gap between its actual and its promised conduct — was so great as to create a serious likelihood that a judge⁶ would have given a harsher sentence.⁷ The gap here seems easily broad enough to create that risk.

Pollard’s sentence should be vacated and the case remanded for resentencing. This should occur before a new judge, as Santo-bello indicates, even though “the fault here rests on the prosecutor, not on the sentencing judge.” 404 U.S. at 263, 92 S.Ct. at 499. See also Moscahlaidis, 868 F.2d at 1363 n. 7 (same); Corsentino, 685 F.2d at 52.

* * # * * *

Though I do not wish to be too critical of the government, and though the analogy is inexact on some points, the case does remind me of Macbeth’s curse against the witches whose promises — and their sophistical interpretations of them — led him to doom:

And be these juggling fiends no more believ’d,

That palter with us in a double sense; That keep the word of promise to our ear,

And break it to our hope.

Macbeth V, vii, 48-51.

ORDER

May 28, 1992.

Before: RUTH B. GINSBURG, SILBERMAN, and WILLIAMS, Circuit Judges.

It is ORDERED, by the court, that the motion of Alan M. Dershowitz, Esq., for leave to enter an appearance is granted, and the Clerk is directed to so note the docket and to file his lodged Motion to Alter or Amend Opinion. Upon consideration thereof, it is

FURTHER ORDERED, by the court, that the motion is granted, and the Opinion filed by Circuit Judge Silberman on March 20, 1992, is amended as follows:

[Editor’s Note: Amendments included in bound volume publication of the opinion.]

1 . It could be argued that the interpretation of specific contract language, in the absence of extrinsic evidence, is also a form of fact-specific characterization. But the "product" of interpretation is the articulation of a norm, i.e., a paraphrase of the contract language for application to particular types of conflicts, so that the process would seem to require appellate courts to rule de novo so as to achieve uniform rules as nearly as possible.

2 . The 9th Circuit commonly says that plea agreements should be interpreted both as a reasonable person would understand them and literally, see, e.g., United States v. Travis, 735 F.2d 1129, 1132 (9th Cir.1984), despite the apparent contradiction. But the case repeatedly cited in favor of literalness does not support such a view, United States v. Garcia, 519 F.2d 1343, 1344 (9th Cir.1975) (finding breach of plea agreement because government failed to abide even by the literal terms of the contract — no discussion of whether the violation had to be of the literal terms for it to be remediable), while the case typically cited for the reasonable person standard indeed favors the reasonable person perspective, United States v. Arnett, 628 F.2d 1162, 1164 (9th Cir.1979) (court looks to the facts of each case to decide what was "reasonably understood by [defendant] when he entered his plea of guilty,” quoting Crusco, 536 F.2d at 23, 27, and the terms of agreement are determined by “objective standards").

3 . Note the omission of any reference to the value of Pollard’s cooperation to "the enforcement of the espionage laws", the only broad purpose mentioned by the plea agreement.

4 . Although the court in Moscahlaidis relied in part on the government’s having obliged itself to "take no position" on sentencing, this hardly makes the decision irrelevant. Compare Maj. Op. at 1027. There the language allowing government commentary was broader — “the full nature and extent of [defendant's] activities”. 868 F.2d at 1362. Without the “take no position" constraint, this might well have left it free to say anything. Here the "facts and circumstances" language is narrower, and, as was true of "full nature and extent” in Moscahlaidis, its meaning is shaded by a related obligation — not to seek a life sentence. See also the passage two paragraphs below, noting the synergies in the government’s promises.

5 . The discrepancy between the majority’s survey of one hundred cases, Maj.Op. at 1029-30, and this account, which is illustrative only, is that the majority leaves out the § 2254 cases.

6 . Measured by Santobello’s objective standard, not by reference to the particular sentencing judge.

7 . The majority’s statement that this test would require resentencing where "a defendant would have received ... a prison term of 11 months rather than a year”, Maj.Op. at 1029, misconceives my position. Deciding whether the government’s breach created a serious likelihood of an effect on a hypothetical judge is completely independent of the penalty imposed.