United States v. Jernigan

BEA, Circuit Judge,

with whom O’SCANNLAIN, Circuit Judge, joins, dissenting:

I cannot articulate my disagreement with the majority’s reasoning better than did our colleague Judge Farris: “My brother and I differ on what is the appropriate appellate function. He would retry. I am content to review.” Li v. Ashcroft, 378 F.3d 959, 964 n. 1 (9th Cir.2004).

The majority gives an exquisitely detailed tour de force through the circumstantial evidence of similarity of physical appearance between the two women. That tour certainly allows the preliminary inference that the jury could have found, on that evidence alone, that Rodriguez-Gallegos looked enough like Jernigan; the jury could have found Rodriguez-Gallegos was probably the September 20, 2000 robber on that evidence. But our task is not to imagine ourselves as jurors given only the circumstantial evidence of the women’s descriptions. We must consider whether the district court erred in rejecting that circumstantial evidence in view of what he saw: that the women did not look alike.1

When we focus on the proper scope of review, our obligation is solely to determine whether the trial court erred in finding the non-disclosed evidence of the later bank robberies was not material; that no Brady violation occurred. See United States v. Bagley, 473 U.S. 667, 676-78, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985) (citing Brady v. Maryland, 373 U.S. 83, 87, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963)). Determination of the facts upon which Brady *1058materiality is decided is properly confided to the trial judge and should be subject to review by us only for clear error. See section II.A, infra.

The only Brady issue2 before us should be whether Judge Carroll’s determination that the women were not of similar appearance was clear error. If not, the undisclosed evidence was not material under Brady.

Judge Earl H. Carroll3 presided over a four-day trial of the bank robbery charge against Jernigan. Jernigan, sans the “hat that obscures her face”4 in the September 20, 2000 bank surveillance video, sat at counsel table, several feet from Judge Carroll. After about 25 hours of such proximity and on the motion for new trial, Judge Carroll compared a photograph of Rodriguez-Gallegos taken from the November 28, 2000 bank surveillance video with a photograph of Jernigan and with Jernigan herself. The clear photograph of Rodriguez-Gallegos was a full face black and white; she was not wearing a hat of any sort nor, Heaven forfend, a mask. But that was not all Judge Carroll had to help him determine whether there was any chance Jernigan could be mistaken for Rodriguez-Gallegos. No, he also had the written admission of Jernigan’s counsel that Jernigan had never argued the two women “looked alike.”5

Judge Carroll found Jernigan’s Brady claim meritless and delivered a written decision denying Jernigan’s motion for new trial, in which decision he made specific findings regarding whether Jernigan looked enough like Rodriguez-Gallegos, so that the former could be mistaken for the latter:

“The simple fact is that [Jernigan] and Rodriguez[-Gallegos] do not look alike....”
“The defense has not argued the two women ‘look alike.’ ”
“Common sense says that if Defendant and Rodriguez do not look alike, the fact that they may share similar physical characteristics such as height, racial characteristics and poor complexions, does not support a finding or an ‘inference’ that Rodriguez-Gallegos robbed the Bank of America branch in Gilbert on September 20, 2000.”6

Judge Carroll’s written decision is consistent with his earlier finding that Jernigan’s “appearance, physical appearance, her face, mouth, hair and that of this Miss Rodriguez[-Gallegos] ... are markedly different.” 7

The majority labels Judge Carroll’s finding of dissimilarity a “red herring.” See note 6, Majority Opinion. Is the finding that Jernigan and Rodriguez-Gallegos do not look alike really a “deliberately misleading object,” like the smoked and smelly fish that poachers would trail on the ground, away from their game, to throw the owner’s dogs off, and leave the game to the poachers?

No, to call Judge Carroll’s visual determination the women did not look alike a “red herring” is to invoke the last retort of the knave caught red-handed: “Who are you going to believe, me or your lying eyes?”

*1059Unlike the majority, I cannot ignore Judge Carroll’s eyes and their finding; they determine the undisclosed circumstantial evidence was not material.

Let me illustrate the effect of such a finding. Assume the prosecutor had disclosed the circumstantial description evidence which suggested Jernigan looked like the November 28, 2000 robber, later identified as Rodriguez-Gallegos. But also assume the prosecutor had disclosed evidence that convinced Judge Carroll that Rodríguez-Gallegos was in fact incarcerated throughout the day of September 20, 2000, and thus unable to have committed the September 20, 2000 bank robbery. Exercising his gate-keeping function to bar confusing and misleading testimony, Judge Carroll would exclude the circumstantial description testimony as irrelevant. See Fed.R.Evid. 104, 402 and 403. We would affirm. We would also affirm if the evidence had not been disclosed at all, see United States v. Sarno, 73 F.3d 1470, 1506 (9th Cir.1995), unless Judge Carroll had committed reversible error in his evidentiary ruling that Rodriguez-Gallegos was in jail and thus unable to have committed the September 20, 2000 bank robbery.

Here, the November 28, 2000 surveillance video photograph dictates the same conclusion because it-supports Judge Carroll’s finding that Jernigan and the November 28, 2000 robber, Rodríguez-Galle-gos, do not look alike. Judge Carroll’s ruling was that even if disclosed, evidence of “the robberies in November 2000 would not have been admissible in this case [.]” See Order at 7, January 21, 2005 (emphasis added). In short, the fact that Jernigan and the November 28, 2000 robber do not look alike, far from being a “red herring,” is the central judicial ruling which renders the undisclosed evidence irrelevant, hence not “material”—unless the Judge who so ruled committed reversible error in his ruling. That is the only issue properly before us.

Applying the correct scope of review, Judge Carroll’s factual findings—carefully ignored by the majority—are not clearly erroneous8 and, in my view, are entitled to deference. Accordingly, since the claimed Brady evidence would not have undermined confidence in the jury verdict of conviction, I would affirm.

The majority, however, refuses to accord deference to these findings, departs from binding Supreme Court precedent in that regard,9 engages in appellate fact finding without the benefit of ever having seen Jernigan (except in the video where she wore a face obscuring hat), departs from Kyles v. Whitley, 514 U.S. 419, 437, 115 S.Ct. 1555, 131 L.Ed.2d 490 (1995), by not determining materiality on the basis of all the non-disclosed evidence, and reverses.

I.

On November 10, 2000, Jernigan was arrested on suspicion of robbing three banks in the East Valley area of Arizona: one on September 20, 2000, one on October 11, 2000, and one on October 25, 2000. While Jernigan was in custody pending trial, two additional banks were robbed in the East Valley: one on November 28, 2000 and one on November 30, 2000. Wit*1060ness statements from all five bank robberies describe the robber as a short Hispanic woman with a poor complexion.

Agent Kyle Richard, the bank robbery coordinator at the Federal Bureau of Investigation’s Phoenix division, was the case agent for all five bank robberies. Agent Richard determined the bank robberies in September and October were not committed by the same woman who committed the bank robberies in November. This determination was based on comparisons of the modus operandi, witness statements, and a photograph of Jernigan with photographs taken from the surveillance video of the November robberies. Owing to his determination, Agent Richard did not inform the Assistant United States Attorney prosecuting Jernigan’s case of the November robberies.10 Thus, the Assistant United States Attorney did not disclose the November bank robberies to Jernigan’s defense counsel.

Five eyewitnesses independently identified Jernigan as having committed the September 20, 2000 bank robbery. At Jernigan’s trial, two of the eyewitnesses described close range contacts with Jerni-gan. One, the victim teller, who was in the immediate presence of Jernigan at the time of the robbery for several minutes, described Jernigan accurately thereafter, and had no trouble identifying Jernigan when she picked her out from the photo-spread two days after the robbery.11 The other testified Jernigan resembled a friend, which made identification for her much easier. The testimony of these two eyewitnesses was corroborated by the testimony of three additional eyewitnesses and the bank surveillance video. As the majority notes, the surveillance video from the September 20, 2000 bank robbery suffers from a lack of quality and was used primarily to bolster the eyewitnesses’ testimony.

Jernigan’s trial defense counsel argued mistaken identify to no avail, and Jernigan was convicted of the September 20, 2000 bank robbery.

On December 11, 2001, Rodríguez-Galle-gos was arrested shortly after robbing the same bank Jernigan was convicted of robbing on September 20, 2000. In addition to the December 11, 2001 bank robbery, Rodríguez-Gallegos was charged with the November 28, 2000 and November 30, 2000 bank robberies.

After learning of Rodriguez-Gallegos’ arrest, Jernigan filed a motion for a new trial, asserting a Brady violation on the basis the Government should have disclosed the November 28, 2000 and November 30, 2000 bank robberies. Judge Earl H. Carroll, the same judge who had presided over Jernigan’s four-day trial, presided over Jernigan’s motion for a new trial. During the evidentiary hearing, Judge Carroll found:

[W]ith respect at least to my observation of the defendant in this case, Miss Jernigan, and the photograph, which is all I have, of the—Miss Rodriguez[Gallegos], I would observe from what I see there that they are different people, and that someone having looked at them under these circumstances would have been able to make such a determination, if they had been presented with it at the same time or close to that time.
The photograph of Miss Jernigan that was used on the photospread and her *1061appearance, physical appearance, her face, mouth, hair and that of this Miss Rodriguez [-Gallegos] I think are markedly different.

Mot. New Trial Tr. 87, May 12, 2004 (Docket No. 117) (emphasis added).

The photograph of Rodriguez-Gallegos to which Judge Carroll compared Jernigan was taken from the surveillance video of the November 28, 2000 bank robbery. This photograph, unlike the surveillance video of the September 20, 2000 bank robbery, does not suffer from a lack of quality. Indeed, this photograph clearly portrays a woman bearing little resemblance to Jernigan. Compare PE 15, with PE 2, and PE 3.

Judge Carroll denied Jernigan’s motion for a new trial in a written order stating:

The simple fact is that [Jernigan] and Rodriguez[-Gallegos] do not look alike, whatever similarities may be in their complexions or Hispanic appearance. It is at best an oxymoron for the defense to claim that Rodriguez[-Gallegos] is “probably” the September 20, 2000 robber, given the admission “that the two women do not “ ‘look alike’
The Government’s arguments about the differences in appearance between Jernigan and Rodríguez-Gallegos are equally thin. In the first place, the Government’s repeated insistence that ‘Jernigan and Rodriguez[-Gallegos] do not in fact look alike,’ e.g. Response at 9, 11, is a straw man. The defense has not argued the two women ‘look alike. ’ (Defendant’s Reply Memorandum, p. 4, Dkt. 104).
Common sense says that if Defendant and Rodriguez do not look alike, the fact that they may share similar physical characteristics such as height, racial characteristics and poor complexions, does not support a finding or an “inference” that Rodríguez-Gallegos robbed the Bank of America branch in Gilbert on September 20, 2000.

Order, January 21, 2005 (emphasis added).

If it is an “oxymoron” for the defense to claim Rodríguez-Gallegos committed the earlier bank robberies, all the while admitting Rodríguez-Gallegos does not look like Jernigan, where does it leave the majority here? It leaves them with the necessity to make appellate findings of fact directly contrary to those made by Judge Carroll and conceded by the defense.

Consider, slowly, the import of such a view. All the appearance characteristics of the two women contained in the witness descriptions are trotted out as circumstantial evidence to show the Brady materiality of the November bank robberies. These appearance characteristics are relevant to prove one point and one point only: the two women look so much alike that the jury should have been informed of the November robberies because the jury could find that five eyewitnesses mistook Jernigan for Rodríguez-Gallegos, who actually committed the earlier heists. But the importance—the “materiality”—of this circumstantial evidence is abandoned by Jernigan’s own counsel’s admission that even he does not think the woman in the September 20, 2000 bank surveillance video and Rodríguez-Gallegos look alike, and he affirmatively points out that he does not even argue that point.12

Not to worry. The majority’s interpretation of Brady is that to be material *1062under Brady, evidence must merely be distracting, when considered alone. Perhaps the majority’s idea is based on a notion that accuracy and truth are irrelevant considerations in determining what arms should be given the defendant to make the trial more of an even joust. That may not be an uncommon notion in certain circles, but it has no support in the law, as is obvious from the lack of any citation to buttress this view.

II.

A.

The majority and I part company over the applicable standard of review. The majority affords the district court’s factual findings no deference. The majority follows our precedent as' to legal issues involving Brady materiality; such issues of law are to be reviewed de novo. See United States v. Lehman, 792 F.2d 899, 901 (9th Cir.1986) (“We review these questions of law de novo and may affirm the district court on any ground supported by the record.” (citations omitted)). But, we have not had the opportunity to consider what, if any, deference should be afforded to a district court’s factual findings that bear on Brady materiality. This case presents the now squandered opportunity to do just that, and to adopt the proper standard of review.

Those of our sister circuits which have considered this question in connection with Brady rulings afford deference to the district court’s findings of fact. The First, Second, Third, Fifth, Seventh, Eighth, Tenth, and Eleventh Circuits and the United States Court of Appeals for the District of Columbia all afford some level of deference to a district court’s factual findings bearing on Brady materiality.13

*1063As the United States Court of Appeals for the District of Columbia has explained:

Generally, this court reviews the district court’s grant of a new trial for abuse of discretion. However, when confronted with a “purely legal question,” our review is de novo. Brady claims present something of a special situation. Thus, as to findings of fact made by the district court, including determinations of credibility made both at trial and in post-trial proceedings, this court would defer under an abuse of discretion standard. But once the existence and content of undisclosed evidence has been established, the assessment of the materiality of this evidence under Brady is a question of law. In this inquiry, the question of prejudice is folded into the determination of whether a violation has occurred. As the Supreme Court has explained, “strictly speaking, there is never a real ‘Brady violation’ unless the non-disclosure was so serious that there is a reasonable probability that the suppressed evidence would have produced a different verdict.” Therefore, once a court finds a Brady violation, a new trial follows as the prescribed remedy, not as a matter of discretion.

United States v. Oruche, 484 F.3d 590, 595-96 (D.C.Cir.2007) (citations omitted) (emphasis added).

Likewise, the Fifth Circuit has explained:

The confusion stems in part from the mixed nature of the Brady inquiry. Whereas we typically analyze legal issues de novo, a Brady determination is inevitably a contextual inquiry, involving questions of both law and fact. Moreover, it is intimately intertwined with the trial proceedings: because the court must judge the effect of the evidence on the jury’s verdict, the Brady decision can never be divorced from the narrative of the trial. In addition, the court must consider not simply the withheld evidence in isolation, but also the quantity and quality of other evidence in the record.
In comparison to a district court ruling on a motion for new trial, an appellate court reviewing a Brady violation is at an inherent disadvantage. Gauging the effect that undisclosed evidence might have had on the outcome of the trial is difficult in any event, but it is made more so when it must be based on a cold record. The district judge, by contrast, has at least had the opportunity to hear the testimony at trial firsthand, view the demeanor of the witnesses, observe the ebb and flow of the evidence at trial, and evaluate the strengths and weaknesses of the government’s case. When, as here, the balance of evidence presented is close, the outcome of the case will often hinge on a subjective and personal evaluation of the evidence and the witnesses. In such a context, some degree of appellate deference makes sense.
*1064We think there is a reconciling theme in our facially competing approaches to Brady—based new trial questions— adhering to decisions that examine the Brady question anew, while acknowledging that we must proceed with deference to the factual findings underlying the district court’s decision. This gives play to the trial court’s superior understanding of the trial, evidence, and witnesses, while reviewing the ultimate constitutional question afresh. It also recognizes that in the new trial context concerns respecting finality are less strong.

United States v. Sipe, 388 F.3d 471, 479 (5th Cir.2004).

I would adopt the standard of review applied by our sister circuits. Specifically, I would defer to the district court’s factual findings unless clearly erroneous, “[b]ut once the existence and content of undisclosed evidence has been established, [treat] the assessment of the materiality of this evidence under Brady [as] a question of law.” Oruche, 484 F.3d at 595. Here, there is no basis for concluding Judge Carroll was “clearly erroneous” in his finding that Jernigan’s appearance is markedly different from that of Rodríguez-Gallegos. Indeed, the majority does not even suggest a basis for finding such error. Accordingly, as discussed below, there is no basis for finding the non-disclosure of the November bank robberies was “material” under Brady.

B.

The majority and I also part company over the majority’s willingness to depart from precedent requiring Brady materiality to be determined on the cumulative effect of the non-disclosed evidence. The majority considers the undisclosed witness descriptions from the November robberies in isolation and concludes a Brady violation occurred. The question before us, however, is whether the cumulative effect of the undisclosed evidence—including the November 28, 2000 surveillance video and photographs derived therefrom—was a reasonable probability of a different result.

“[T]he state’s obligation under Brady v. Maryland to disclose evidence favorable to the defense, turns on the cumulative effect of all such evidence suppressed by the government.” Kyles, 514 U.S. at 421, 115 S.Ct. 1555 (citation omitted). Thus, although the tendency and forcé of undisclosed evidence must be evaluated “item by item; there is no other way,” the materiality of such evidence is a separate determination, based on the effect of all the evidence, disclosed and undisclosed. Id. at 436, 115 S.Ct. 1555 n. 10. Accordingly, the majority is correct in explaining that when considering whether non-disclosed evidence would have created a reasonable probability of a different result, judges must “ ‘undertake a careful, balanced evaluation of the nature and strength of both the evidence the defense was prevented from presenting and the evidence each side presented at trial.’ ” Bailey v. Rae, 339 F.3d 1107, 1119 (9th Cir.2003) (quoting Boss v. Pierce, 263 F.3d 734, 745 (7th Cir.2001)).

The majority, however, fails to conduct a balanced evaluation of the non-disclosed evidence in this case. Specifically, in determining Brady materiality the majority considers only the undisclosed circumstantial evidence that suggests Jernigan and Rodríguez-Gallegos are of similar appearance, i.e., the witness descriptions. The majority fails to consider the direct evidence establishing that Jernigan and Rod-ríguez-Gallegos are “markedly different” in appearance, i.e., the November 28, 2000 surveillance video, the photograph of Jer-nigan, and Jernigan herself. The majority’s failure to consider the last item comes as no surprise because, unlike Judge Carroll, the majority has never had the benefit of seeing Jernigan in person.

*1065Freed from Kyles ’ requirement that the non-disclosed evidence be evaluated “item by item” and that materiality be determined based on the cumulative effect of all the evidence, the majority rests its materiality determination on a finding that the November robberies were committed by a woman “whose description bore an uncanny physical resemblance” to Jernigan. Having thus erred, the majority speculates the five eyewitnesses who independently identified Jernigan as committing the September 20, 2000 bank robbery “simply picked Jernigan out of the photospread because she was the woman who looked the most like [Rodriguez-Gallegos].”

III.

As the majority explains, the touchstone of Brady materiality is whether admission of the suppressed evidence would have created a “ ‘reasonable probability’ of a different result.” Kyles, 514 U.S. at 434, 115 S.Ct. 1555. “[T]he prosecution, which alone can know what is undisclosed, must be assigned the consequent responsibility to gauge the likely net effect of all such evidence and make disclosure when the point of ‘reasonable probability’ is reached.” Id. at 437, 115 S.Ct. 1555 (emphasis added). Here, the point of “reasonable probability” was never reached. The non-disclosed evidence in this case does not create a reasonable probability of a different result because it does not support the inference that the five eyewitness who independently identified Jernigan as having committed the September 20, 2000 bank robbery were mistaken. To do so, the non-disclosed evidence would have to permit the inference that Jernigan sufficiently resembles Rodriguez-Gallegos such that Jernigan could be mistaken for Rodriguez-Gallegos.

Notwithstanding the witness statements, the district court found the non-disclosed photographs taken from the surveillance video of the November 28, 2000 bank robbery establish that Jernigan and Rodriguez-Gallegos are “markedly different” in appearance and “do not look alike, whatever similarities may be in their complexions or Hispanic appearance.” In addition, the district court found “someone having looked at them under these circumstances would have been able to make such a determination.” These findings are not clearly erroneous. Moreover, Judge Carroll, who had the benefit of observing Jer-nigan during the course of her four-day trial, was in a better position than this court to determine whether Jernigan resembles Rodriguez-Gallegos.

In short, I would defer to Judge Carroll’s factual findings, conduct a de novo review of Brady materiality based on Judge Carroll’s findings, and affirm.

. The majority concludes that "what Rodriguez-Gallegos actually looks like is completely irrelevant to the Brady analysis” because the Government did not know the identity of the November 28, 2000 robber prior to Jerni-gan’s trial. See note 6, Majority Opinion. The majority errs in failing to distinguish between identity and appearance; they are not synonymous for purposes of determining the materiality of the claimed Brady evidence. Although the Government did not know the identity of the November 28, 2000 robber prior to Jernigan's trial, the Government knew what the November 28, 2000 robber looked like prior to Jernigan's trial; the Government possessed and viewed surveillance video of the November 28, 2000 bank robbery prior to Jernigan’s trial. FBI Agent Kyle Richard had compared the photographs taken from the November 28, 2000 surveillance video (of Rodriguez-Gallegos) to those of Jerni-gan, and had determined—as Judge Carroll later did—he was dealing with two different female bank robbers. Significantly, Judge Carroll’s determination that Jernigan did not look like the November 28, 2000 robber was based solely on evidence possessed by the government prior to Jernigan’s trial, i.e., a photograph taken from the November 28, 2000 surveillance video. See Mot. New Trial Tr. 87, May 12, 2004 (Docket No. 117). Accordingly, Judge Carroll’s finding that Jerni-gan did not look like the November 28, 2000 robber is hardly irrelevant. Indeed, if by some happenstance the two women had the same names—further complicating matters from an identity point of view—this coincidence would not have made the evidence "material” under Brady for the simple reason the two women do not look alike. Hence, the non-disclosed evidence does not permit the inference that the eyewitnesses mistook Jerni-gan for Rodriguez-Gallegos.

. The Government concedes the evidence was favorable and undisclosed.

. Appointed to the district court in 1980.

. See note 3, Majority Opinion.

. Def.'s Reply on Mot. for New Trial 4, April 12, 2004.

. Order, January 21, 2005.

. Mot. New Trial Tr. 87, May 12, 2004 (Docket No. 117).

. There is no claim that Judge Carroll did not actually have a photograph of Rodriguez-Gallegos to compare to Jernigan or that Judge Carroll was indeed blind—claims that might be grounds for clear error in fact finding.

. Similar appellate fact finding contrary to what the trial judge has observed and has deduced from his own observations have recently resulted in reversals of our decisions. See Uttecht v. Brown, - U.S. -, 127 S.Ct. 2218, 167 L.Ed.2d 1014 (2007); Rice v. Collins, 546 U.S. 333, 126 S.Ct. 969, 163 L.Ed.2d 824 (2006).

. This in no way alleviated the Assistant United States Attorney's duty to disclose the November robberies if they were material under Brady. See Kyles, 514 U.S. at 437, 115 S.Ct. 1555 ("[T]he individual prosecutor has a duty to learn of any favorable evidence known to the others acting on the government's behalf in the case, including the police.'').

. No claim is made that the photospread was suggestive or otherwise improper.

. Indeed, on motion for new trial Jernigan’s defense counsel argued that the September 2000 videotape was of such poor quality that one could not rule out that it showed Rodriguez-Gallegos. This is a particularly weak argument. If the September 2000 tape is undecipherable, it identifies no one, rather than “possibly” someone. It no more identifies Rodriguez-Gallegos than Paris Hilton.

. See United States v. Madori, 419 F.3d 159, 169 (2d Cir.2005) ("Materiality in this context presents us with a mixed question of law and fact. While the trial judge's factual conclusions as to the effect of non-disclosure are entitled to great weight, we examine the record de novo to determine whether the evidence in question is material as a matter of law.”); United States v. Vallejo, 297 F.3d 1154, 1163 (11th Cir.2002) ("Similarly, a district court's denial of a motion for new trial based on a Brady violation is reviewed for abuse of discretion.”); United States v. Ryan, 153 F.3d 708, 711 (8th Cir.1998) ("[Abuse of discretion] standard also applies where, as here, a defendant seeks a new trial premised upon a Brady claim. This deferential standard of review is especially appropriate in the context of a lengthy, hard-fought, highly charged case like the present one.” (citing United States v. Williams, 81 F.3d 1434, 1437 (7th Cir.1996)) (citations omitted)); United States v. Boyd, 55 F.3d 239, 242 (7th Cir. 1995) ("But the other judgments that the district judge makes, signally here the judgment whether some piece (or pieces) of evidence wrongfully withheld by the government might if disclosed have changed the outcome of the trial, are to be reviewed deferentially. This is not only the rule; it is the dictate of common sense, especially in a case such as this. Forget the 29 witnesses at the evidentiary hearing; forget there was an evidentiary hearing on the motion for a new trial. Before then, during the trial, Judge Aspen had for months on end listened to witnesses—had heard, had not merely read, their testimony, and had watched them as they gave it. And he had observed the jurors as they listened to the witnesses. A trial judge of long experience, he would have developed a feel for the impact of the witnesses on the jury—and how that impact might have been different had the government played by the rules—that an appellate court, confined to reading the transcript,' cannot duplicate. Judge Aspen may have been mistaken; we might suspect that he was mistaken; but unless we are convinced that he was mistaken, we have no warrant to reverse. That is what it means to say that appellate review is deferential. It is not abject, but it is deferential.” (citations omitted)); United States v. Thornton, 1 F.3d 149, 158 (3rd Cir.1993) (“In considering a district court's ruling on a motion for a new trial based on the failure to disclose Brady materials, we will conduct a de novo review of the district court’s conclusions of law as well as a 'clearly erroneous' review of any findings of fact where appropriate. Where the district *1063court applies the correct legal standard, its weighing of the evidence merits deference from the Court of Appeals, especially given the difficulty inherent in measuring the effect of a non-disclosure on the course of a lengthy trial covering many witnesses and exhibits.” (internal quotation marks and citations omitted)); United States v. Sanchez, 917 F.2d 607, 618 (1st Cir.1990) ("Due to its inherently fact—bound nature the district court's determination on the materiality of newly discovered evidence in prosecutorial non-disclosure cases is ordinarily accorded deference.”); United States v. Buchanan, 891 F.2d 1436, 1440 (10th Cir.1989) (“We review the factual findings of a district court acting pursuant to 28 U.S.C. § 2255 under the clearly erroneous standard. However, the materiality of withheld evidence under Brady and its possible effect on the verdict are mixed questions of fact and law reviewed de novo." (citations omitted)).