Anthony Juniper v. Melvin Davis

74 F.4th 196

Opinion

PUBLISHED

UNITED STATES COURT OF APPEALS FOR THE FOURTH CIRCUIT

No. 21-9

ANTHONY BERNARD JUNIPER, Petitioner - Appellant,

v. MELVIN C. DAVIS, Warden, Wallens Ridge State Prison, Respondent - Appellee.

Appeal from the United States District Court for the Eastern District of Virginia, at Richmond. John A. Gibney, Jr., Senior District Judge. (3:11-cv-00746-JAG)

Argued: March 7, 2023 Decided: July 19, 2023

Before DIAZ, Chief Judge, and GREGORY and WYNN, Circuit Judges.

Affirmed by published opinion. Judge Wynn wrote the opinion, in which Chief Judge Diaz and Judge Gregory joined.

ARGUED: Arianna Helene Zoghi, VIRGINIA CAPITAL REPRESENTATION RESOURCE CENTER, Charlottesville, Virginia, for Appellant. Kevin Michael Gallagher, OFFICE OF THE ATTORNEY GENERAL OF VIRGINIA, Richmond, Virginia, for Appellee. ON BRIEF: Robert Lee, Elizabeth Peiffer, VIRGINIA CAPITAL REPRESENTATION RESOURCE CENTER, Charlottesville, Virginia, for Appellant. Jason S. Miyares, Attorney General, Charles H. Slemp, III, Chief Deputy Attorney General, Andrew N. Ferguson, Solicitor General, Erika L. Maley, Principal Deputy Solicitor General, M. Jordan Minot, Assistant Solicitor General, OFFICE OF THE ATTORNEY GENERAL OF VIRGINIA, Richmond, Virginia, for Appellee.

2

WYNN, Circuit Judge:

This habeas case returns to us after a remand for an evidentiary hearing. Petitioner Anthony Juniper argues that the Commonwealth of Virginia’s 1 suppression of evidence and knowing presentation of false or misleading testimony were cumulatively material under Brady v. Maryland and Napue v. Illinois so as to entitle him to habeas relief. The district court rejected Juniper’s contention. We affirm.

I.

Juniper’s habeas proceedings have followed a long and winding path, thanks in large part to the Commonwealth’s refusal to disclose potentially exculpatory evidence to the defense. Our decision today marks the culmination of more than a decade of proceedings in federal court, and nearly two decades of post-conviction proceedings. We begin by summarizing this history.

Keshia Stephens, her younger brother Rueben Harrison, III, and two of her daughters, four-year-old Nykia Stephens and two-year-old Shearyia Stephens, were shot to death in Keshia’s apartment in Norfolk, Virginia, on January 16, 2004. 2 Juniper v. Commonwealth, 626 S.E.2d 383, 393–94 (Va. 2006). The police began trying to contact Juniper—Keshia’s “on-again, off-again” boyfriend, id. at 394—“the day after the

1

Respondent in this case is Melvin Davis in his official capacity as Warden of Wallens Ridge State Prison. For simplicity, we refer to Respondent and the prosecution at trial as the Commonwealth.

2

The names of a number of individuals in this case, including the victims, have been spelled in different ways over the course of the proceedings. In this opinion, we employ the spellings that we and other courts have used previously. In doing so, we of course mean no disrespect to the individuals involved, to the extent the spellings used herein are incorrect.

3

murders,” Juniper v. Davis, No. 3:11CV746, 2021 WL 3722335, at *12 (E.D. Va. Aug. 23, 2021). On January 23, having had no luck in locating Juniper, police “obtained murder warrants, and placed him on TV as wanted for murder.” Id. (internal quotation marks omitted). Juniper turned himself in three days later. Id.

The Commonwealth tried Juniper for the murders in January 2005. As we explain in detail below, “the prosecution presented a mountain of testimonial and forensic evidence of [Juniper]’s guilt” at trial. Juniper v. Hamilton, 529 F. Supp. 3d 466, 477 (E.D. Va. 2021). The jury convicted Juniper of the four murders, and the court sentenced him to death. 3 Id. at 475. On July 1, 2021, however, his sentence was commuted to life without parole after Virginia abolished the death penalty by legislation. 4 Juniper appealed his convictions and sentences, but the Supreme Court of Virginia affirmed, and the United States Supreme Court denied certiorari. See Juniper, 626 S.E.2d at 393, cert. denied, Juniper v. Virginia, 549 U.S. 960 (2006). He also unsuccessfully pursued post-conviction relief in state court, and the United States Supreme Court again denied certiorari. See Juniper v. Warden of Sussex I State Prison, 707 S.E.2d 290, 296

3

The jury also convicted Juniper of “statutory burglary while armed with a deadly weapon[] and four counts of use of a firearm in the commission of a felony.” Juniper, 626 S.E.2d at 393.

4

2021 Special Session I Va. Acts 826, 851 cl. 3 (providing that “any person under a sentence of death imposed for an offense committed prior to July 1, 2021, but who has not been executed by July 1, 2021, shall have his sentence changed to life imprisonment, and such person who was 18 years of age or older at the time of the offense shall not be eligible for (i) parole, (ii) any good conduct allowance or any earned sentence credits . . . , or (iii) conditional release”). Juniper was over the age of 18 at the time of the murders.

4

(Va.) (en banc), cert. denied, Juniper v. Kelly, 565 U.S. 1082 (2011).

A few months before the Supreme Court of Virginia denied Juniper’s habeas petition, he learned about key evidence that he claims the Commonwealth did not disclose: statements by Keshia’s neighbors, Wendy Roberts and her son Jason Roberts. Their statements, if accepted by the jury, would have foreclosed the prosecution’s timeline of the murders and pointed to an alternative suspect. We detail those statements below.

Juniper apparently learned of the Robertses’ statements during the October 2010 federal jury trial of Robert Ford, the lead police investigator on Juniper’s case. Juniper v. Pearson, No. 3:11-CV-00746, 2013 WL 1333513, at *11 (E.D. Va. Mar. 29, 2013). Ford was charged with, and convicted of, “taking bribes from drug defendants in exchange for falsely representing to judges and prosecutors that those defendants had cooperated in homicide investigations.” Juniper v. Zook, 876 F.3d 551, 559 (4th Cir. 2017); see United States v. Ford, 470 F. App’x 146, 146 (4th Cir. 2012) (per curiam) (affirming Ford’s convictions for conspiracy to commit extortion under color of official right, extortion under color of official right, and making false statements). The Government’s evidence at Ford’s trial included investigation notes that mentioned Wendy. Juniper, 876 F.3d at 559. Juniper’s counsel only learned about Jason when they “fortuitously found him at his mother’s house while trying to find [her].” Juniper, 2013 WL 1333513, at *11 (internal quotation marks omitted).

Juniper twice sought discovery in the Supreme Court of Virginia related to police interviews of Wendy and Jason. Id. at *12. But the Commonwealth opposed the motions, and the Supreme Court of Virginia denied them. Id. Juniper also filed a second habeas

5

petition in state court based on the evidence from Wendy and Jason, but the Supreme Court of Virginia rejected the petition as untimely in October 2011. Id.

The Commonwealth scheduled Juniper’s execution for November 10, 2011, and the United States Supreme Court denied his application for a stay on November 9. Juniper v. Kelly, 565 U.S. 1031 (2011). But that same day, Juniper successfully obtained a stay from the United States District Court for the Eastern District of Virginia. He then filed a § 2254 petition in January 2012, raising numerous exhausted and unexhausted claims.

Among Juniper’s unexhausted claims was one brought under Brady related to the Roberts evidence. Brady v. Maryland, 373 U.S. 83, 87 (1963) (holding that the prosecution violates due process when it suppresses material, favorable evidence). And he again “sought extensive discovery related to the information revealed in the withheld investigative notes,” which the Commonwealth again opposed. Juniper, 876 F.3d at 560. “Recognizing the difficulty of assessing [Juniper]’s Brady claim without knowing what evidence was actually in the Commonwealth’s possession at the time of Juniper’s prosecution,” however, the district court ordered the Commonwealth “to produce several categories of evidence, including any record of statements that the Roberts[es] made to police and the photo line-up that police showed to Wendy Roberts during their investigation.” Juniper, 2013 WL 1333513, at *12.

As the district court put it, this order “brought to an end the Commonwealth’s long, senseless resistance to the disclosure of its information about the Roberts[es].” Id. at *12 n.6. And the district court concluded “that the prosecution likely withheld exculpatory evidence, in clear violation of its legal duties.” Id. at *14; accord id. at *15 (“The events

6

leading up to this point . . . show the Commonwealth’s entrenched resistance to transparency in this criminal prosecution and subsequent post-conviction proceedings.”).

Nevertheless, in 2013, having reviewed the evidence in question, the district court granted the Commonwealth’s motion to dismiss Juniper’s § 2254 petition. Id. at *1. As to the Roberts evidence, the district court concluded that even though Juniper could satisfy two Brady requirements at the motion-to-dismiss stage 5—that the evidence was suppressed and that it was favorable—he could not satisfy the third, namely, that the evidence was material. Id. at *15–16. The court reached this conclusion because it found that the Roberts evidence could not “unsettle certain basic facts about the murders,” including forensic evidence, “Juniper’s conduct and statements after the killings,” and phone records. Id. at *17. Nevertheless, the district court granted a certificate of appealability on the Roberts Brady claim. 6 Juniper v. Davis, 737 F.3d 288, 289 (4th Cir. 2013) (citing Final Order at 1, Juniper v. Pearson, No. 3:11-cv-00746 (E.D. Va. May 20, 2013)).

In 2017, we reviewed that claim and concluded that “the district court abused its

5

Later, considering a more fleshed-out record at summary judgment, the district court concluded that Juniper was entitled to summary judgment as to the favorability of the Roberts evidence and the suppression of the photo line-up shown to Wendy, but that material disputes of fact remained about “whether Juniper’s defense team knew about the Robertses’ statements.” Juniper, 529 F. Supp. 3d at 493; see id. at 491–92 & n.45, 494.

6

The district court also granted a certificate of appealability on whether Juniper was entitled to the appointment of independent counsel pursuant to Martinez v. Ryan, 566 U.S. 1 (2012). Juniper v. Davis, 737 F.3d 288, 289 (4th Cir. 2013) (citing Final Order at 1, Juniper v. Pearson, No. 3:11-cv-00746 (E.D. Va. May 20, 2013)). We concluded that he was and remanded for further proceedings. Id. at 290. On remand, the district court rejected Juniper’s Martinez claims, a decision that we recently affirmed. Juniper v. Zook, 117 F. Supp. 3d 780 (E.D. Va. 2015), aff’d sub nom. Juniper v. Davis, No. 16-2, 2023 WL 3050984 (4th Cir. Apr. 24, 2023).

7

discretion in dismissing [the] claim without holding an evidentiary hearing because it failed to assess the plausibility of that claim through the proper legal lens.” Juniper, 876 F.3d at 556. Specifically, we held that “[t]he district court did not apply the proper legal standard in determining whether [Juniper] alleged or established sufficient facts regarding materiality to warrant an evidentiary hearing.” Id. at 567. We explained that “[a] petitioner who has diligently pursued his habeas corpus claim in state court is entitled to an evidentiary hearing in federal court, on facts not previously developed in the state court proceedings, if the facts alleged would entitle him to relief, and if he satisfies one of the six factors enumerated by the Supreme Court in Townsend v. Sain.” 7 Id. at 563 (quoting Conaway v. Polk, 453 F.3d 567, 582 (4th Cir. 2006)) (citing Townsend v. Sain, 372 U.S. 293, 313 (1963)). After concluding that Juniper had diligently pursued the claim and that he could satisfy at least one of the Townsend factors, we evaluated whether the “facts alleged would entitle him to relief.” Id.; see id. at 564.

We emphasized that the applicable standard was that of Federal Rule of Civil Procedure 12(b)(6)—whether, construing the facts in the light most favorable to Juniper, drawing all reasonable inferences in his favor, and resolving all credibility determinations in his favor, his petition stated a claim to relief that was plausible on its face. Id. at 564

7

The Townsend factors are: “(1) the merits of the factual dispute were not resolved in the state hearing; (2) the state factual determination is not fairly supported by the record as a whole; (3) the fact-finding procedure employed by the state court was not adequate to afford a full and fair hearing; (4) there is a substantial allegation of newly discovered evidence; (5) the material facts were not adequately developed at the state-court hearing; or (6) for any reason it appears that the state trier of fact did not afford the habeas applicant a full and fair fact hearing.” Townsend v. Sain, 372 U.S. 293, 313 (1963).

8

(first citing Wolfe v. Johnson, 565 F.3d 140, 169 (4th Cir. 2009); and then citing United States ex rel. Oberg v. Penn. Higher Educ. Assist. Agency, 745 F.3d 131, 136 (4th Cir. 2014)); accord id. at 569. Applying that standard, we concluded that Juniper had raised a plausible claim, and therefore was entitled to an evidentiary hearing. 8 Id. at 571–72. In doing so, we stated that “we [did] not believe the inculpatory value of the [forensic evidence] [to be] so great as to preclude the Roberts evidence, if found to be sufficiently credible during an evidentiary hearing, from ‘putting the whole case in such a different light as to undermine confidence in the verdict.’” Id. at 571 (alterations adopted) (emphasis added) (quoting Kyles v. Whitley, 514 U.S. 419, 435 (1995)). We also noted that the district court could consider whether the circumstances of the case “warrant[ed] authorizing additional discovery.” Id. at 572 n.9.

On remand, the district court permitted the parties to engage in further discovery, during which Juniper learned of additional evidence that he alleged was also Brady or Napue material. Juniper, 529 F. Supp. 3d at 476, 481 (citing Napue v. Illinois, 360 U.S. 264 (1959) (holding that the prosecution violates due process when it knowingly offers or fails to correct false or misleading testimony)). Because Juniper had “diligently pursued” exculpatory evidence, but “only received the new evidence in post-remand discovery,” the district court allowed him to file an amended habeas petition. Id. at 484; see id. at 476. The August 2019 amended petition included nineteen claims alleging (1) violations of Brady

8

However, at the present stage of the proceedings—after the district court rendered a decision on the merits following an evidentiary hearing—Juniper no longer receives the benefit of the favorable 12(b)(6) standard.

9

or Napue and (2) ineffective assistance of counsel. Id. at 476, 481. “The heart of [Juniper’s amended] petition attack[ed] the credibility of the [prosecution’s] witnesses . . . . He d[id] not challenge the validity or admissibility of the [forensic] evidence.” Id. at 481.

In June 2020, the parties jointly filed more than 800 stipulations of fact. They then each moved for summary judgment. Id. at 475. In March 2021, the district court granted in part and denied in part each motion. Of Juniper’s nineteen claims, the district court granted summary judgment to the Commonwealth on seven, concluding that six were barred by the mandate rule 9 and one lacked merit. Id. at 484–85, 518–19. The court also granted summary judgment to one party or the other on some prongs of some of the remaining claims, but because material disputes of fact remained on at least one prong, concluded summary judgment was not appropriate for the full claim. See J.A. 1095–98 (district court order summarizing its summary-judgment rulings as to each claim). 10 This process left twelve outstanding claims: nine under Brady, two under Napue, and one for ineffective assistance of counsel. See id.; Juniper, 529 F. Supp. 3d at 481–82.

The district court held a five-day evidentiary hearing in June and July 2021, after which it rejected Juniper’s remaining claims. Regarding the Brady and Napue claims, “[a]t

9

The mandate rule is the “doctrine that the mandate of a higher court is ‘controlling as to matters within its compass.’” Juniper, 529 F. Supp. 3d at 482 (quoting United States v. Bell, 5 F.3d 64, 66 (4th Cir. 1993)). With some exceptions, the mandate rule generally “prohibits” a district court “from altering ‘rulings impliedly made by the appellate court’” and “from reconsidering issues ‘the parties failed to raise on appeal.’” Id. (quoting S. Atl. Ltd. P’ship of Tenn. v. Riese, 356 F.3d 576, 584 (4th Cir. 2004)). Applied here, the district court declined to “reconsider[] its previous dismissal of” certain of Juniper’s claims “that he did not appeal and [this Court] did not remand.” Id. at 483.

10

Citations to the “J.A.” refer to the Joint Appendix filed by the parties in this appeal.

10

the hearing, the Court assessed only materiality” and “assumed, without deciding, that Juniper had satisfied all other elements of his claims.” 11 Juniper, 2021 WL 3722335, at *1. The court again emphasized that “[t]he conduct of the Commonwealth at trial, and the Warden during these proceedings, leaves much to be desired. The better course is always to disclose evidence, even if, as here, its materiality is questionable.” Id. at *24. Nevertheless, because the court concluded that Juniper could not show materiality—either of any individual Brady or Napue claim or of all of them cumulatively—the court rejected the habeas petition and denied a certificate of appealability. Id. at *30.

Juniper timely appealed and sought a certificate of appealability on a single issue:

“whether the suppression of evidence and knowing presentation of false and misleading testimony were cumulatively material” under a combined Brady and Napue analysis. Opening Br. at 1 (emphasis added). This Court granted the certificate of appealability, received full briefing, and heard oral argument.

II.

The sole question before us on appeal is whether Juniper can show that the evidence he cites as inappropriately suppressed under Brady and the testimony he cites as inappropriately offered under Napue were cumulatively material. We conclude that Juniper

11

The court clarified that it was not “disturb[ing]” its summary-judgment rulings granting summary judgment on the other prongs of some of the Brady and Napue claims. Juniper, 2021 WL 3722335, at *1 n.6. Rather, because disputes of material fact had remained as to those prongs on some of the claims, the court opted to refer to “assuming” that the elements other than materiality were satisfied for all claims “[f]or simplicity’s sake.” Id.

11

cannot make this showing, and therefore affirm.

A.

We begin with the applicable legal standards.

“We review de novo the district court’s denial of [Juniper]’s 28 U.S.C. § 2254 petition.” Porter v. White, 23 F.4th 322, 326 (4th Cir. 2022). “In doing so, ‘we review the district court’s legal conclusions de novo and findings of fact for clear error.’” Id. (quoting Wolfe v. Clarke, 691 F.3d 410, 423 (4th Cir. 2012) (some internal quotation marks omitted)).

In this appeal, Juniper presses the cumulative materiality of Brady and Napue claims that were procedurally defaulted. See Juniper, 2021 WL 3722335, at *22. Procedural default occurs when “the state courts [have] denied [the petitioner’s] claims by relying on a well-established state procedural rule,” Martinez v. Ryan, 566 U.S. 1, 7 (2012), as well as “when a habeas petitioner fails to exhaust available state remedies and the court to which he would be required to present his claims in order to meet the exhaustion requirement would now find the claims procedurally barred,” Mahdi v. Stirling, 20 F.4th 846, 892 (4th Cir. 2021) (alterations accepted) (quoting Breard v. Pruett, 134 F.3d 615, 619 (4th Cir. 1998)), cert. denied, 143 S. Ct. 582 (2023). In a typical case of procedural default, “we will not review [the claim’s] merits . . . unless the petitioner establishes both cause for the default and prejudice.” Plymail v. Mirandy, 8 F.4th 308, 316 (4th Cir. 2021), as amended (Sept. 7, 2021). This is because, by statute, the petitioner must show that he “has exhausted the remedies available in the courts of the State.” 28 U.S.C. § 2254(b)(1)(A).

Here, the Supreme Court of Virginia found Juniper’s claims based on the Roberts

12

evidence to be untimely and thus procedurally defaulted. Juniper, 2013 WL 1333513, at *12. And Juniper’s claims based on evidence he received for the first time even later than the Roberts evidence are similarly procedurally defaulted because the state court would dismiss them as untimely. See Juniper, 529 F. Supp. 3d at 488 (“[A] Virginia court would find [Juniper’s] claims barred by the statute of limitations.” (citing Va. Code § 8.01- 654.1)). 12 The Commonwealth also properly raised the “affirmative defense” of procedural default “before the district court.” Torrence v. Lewis, 60 F.4th 209, 213 n.2 (4th Cir. 2023).

Thus, in order for a federal court to review Juniper’s claims, he normally would be required to establish cause and prejudice to overcome the procedural default. But “[a]n application for a writ of habeas corpus may be denied on the merits, notwithstanding the failure of the applicant to exhaust the remedies available in the courts of the State.” 28

12

The district court cited a statute, Va. Code § 8.01-654.1, that provided deadlines specifically for habeas petitions filed by prisoners on death row. The Virginia General Assembly repealed that statute when it abolished the death penalty. See 2021 Special Session I Va. Acts 826, 851 cl. 2. However, another Virginia statute provides that a habeas petition “attacking a criminal conviction or sentence shall be filed within two years from the date of final judgment in the trial court or within one year from either final disposition of the direct appeal in state court or the time for filing such appeal has expired, whichever is later.” Va. Code § 8.01-654(A)(2). Those deadlines have long since passed.

To be sure, the Supreme Court of Virginia has indicated that the statutory deadlines provided by Va. Code § 8.01-654 are subject to tolling under Va. Code § 8.01-229(D) for purposes of Brady claims, because Brady claims by their nature involve obstruction by the government. Hicks v. Dir., Dep’t of Corr., 768 S.E.2d 415, 419 (Va. 2015) (“[I]n a claim for habeas corpus relief based on a Brady violation, the failure to disclose exculpatory evidence qualifies as obstruction by the defendant that prevents the filing of the claim for purposes of Code § 8.01-229(D).”). However, neither party has pointed to this tolling provision or argued that Juniper’s claims are not procedurally defaulted. In the absence of any such argument, and because the “rule limiting our review of procedurally defaulted claims ‘is not a jurisdictional one’” and we intend to skip straight to the merits analysis anyway, we will assume that Juniper’s claims are procedurally defaulted. Plymail, 8 F.4th at 316 (quoting Yeatts v. Angelone, 166 F.3d 255, 260 (4th Cir. 1999)).

13

U.S.C. § 2254(b)(2). Further, a showing of materiality under Brady would establish prejudice for purposes of this analysis, while a showing of suppression would establish cause. Juniper, 876 F.3d at 564–65 n.6 (citing Banks v. Dretke, 540 U.S. 668, 691 (2004)). “Thus, if [Juniper] cannot overcome the procedural bar, he necessarily fails on the merits, and vice versa.” Juniper, 2013 WL 1333513, at *14. Accordingly, we will follow the district court’s lead by “bypass[ing] the cause-and-prejudice analysis and proceed[ing] to the merits of Juniper’s claims.” Juniper, 2021 WL 3722335, at *23 n.37.

B.

Under Brady v. Maryland, “the suppression by the prosecution of evidence favorable to an accused upon request violates due process where the evidence is material either to guilt or to punishment.” 373 U.S. at 87. To demonstrate a Brady violation, a petitioner must “show that the undisclosed evidence was (1) favorable to him either because it is exculpatory, or because it is impeaching; (2) material to the defense . . . ; and (3)” suppressed. United States v. Wilson, 624 F.3d 640, 661 (4th Cir. 2010). Like the district court, we will assume that the evidence Juniper cites was favorable and suppressed and will focus only on materiality.

Evidence is material under Brady “if there is a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different.” Kyles, 514 U.S. at 433–34 (emphasis added) (quoting United States v. Bagley, 473 U.S. 667, 682 (1985)). “A ‘reasonable probability’ is a probability sufficient to undermine confidence in the outcome.” Bagley, 473 U.S. at 682. The petitioner “does not have to show by a preponderance of the evidence that disclosure of the evidence would

14

have resulted in acquittal.” United States v. Ellis, 121 F.3d 908, 915–16 (4th Cir. 1997). Nor must he “demonstrate that after discounting the inculpatory evidence in light of the undisclosed evidence, there would not have been enough left to convict,” because “[t]he possibility of an acquittal on a criminal charge does not imply an insufficient evidentiary basis to convict.” Kyles, 514 U.S. at 434–35.

Rather, the question is “whether in [the suppressed evidence’s] absence [the defendant] received a fair trial, understood as a trial resulting in a verdict worthy of confidence.” Id. at 434. That is, a petitioner demonstrates a Brady violation “by showing that the favorable evidence could reasonably be taken to put the whole case in such a different light as to undermine confidence in the verdict.” Id. at 435. So, while “whether evidence withheld is material is not a sufficiency of evidence test . . . . [,] we do not ignore other evidence presented at trial in determining our confidence in the outcome. Instead, we evaluate the whole case, taking into account the effect that the suppressed evidence, had it been disclosed, would have had on the evidence considered at trial.” Ellis, 121 F.3d at 918 (footnote omitted).

Additionally, we must weigh the cumulative materiality of the allegedly suppressed, favorable evidence. Bowman v. Stirling, 45 F.4th 740, 754 (4th Cir. 2022). “In order to determine ‘the aggregate effect’ of the withheld evidence, the court must both ‘add to the weight of the evidence on the defense side all of the undisclosed exculpatory evidence’ and ‘subtract from the weight of the evidence on the prosecution’s side the force and effect of all the undisclosed impeachment evidence.’” Juniper, 876 F.3d at 568 (alterations adopted) (quoting Smith v. Sec’y, Dep’t of Corr., 572 F.3d 1327, 1347 (11th Cir. 2009)). And “[o]nce

15

the evidence on the scales is adjusted to take into account the combined force and effect of the undisclosed evidence favorable to the defense, the standard that is applied is . . . whether what is left on both sides of the scale . . . creates a reasonable probability that a jury would acquit, and a reasonable probability is one sufficient to undermine our confidence in the guilty verdict.” Id. at 569 (quoting Smith, 572 F.3d at 1347); accord Monroe v. Angelone, 323 F.3d 286, 302 (4th Cir. 2003) (“In assessing the issue of materiality, we must evaluate the importance of the Commonwealth’s suppression of the Habeas Evidence. To do so, we first assess the Commonwealth’s evidence that [the petitioner] committed [the crime]. We then weigh against this evidence the strength of [the petitioner]’s defense. Finally, we consider whether the Habeas Evidence, had it been disclosed and used effectively, is likely to have affected the verdict . . . . In other words, we examine whether the Commonwealth’s suppression of the Habeas Evidence was material to the fairness of [the petitioner]’s trial.” (citation omitted)).

One subset of Brady claims is a Napue claim. “Under the Supreme Court’s decision in Napue v. Illinois, the government may not knowingly use false evidence, including false testimony, to obtain a tainted conviction or allow it to go uncorrected when it appears. False testimony includes both perjury and evidence that, though not itself factually inaccurate, creates a false impression of facts which are known not to be true.” Burr v. Jackson, 19 F.4th 395, 410 (4th Cir. 2021) (alterations adopted) (citations and internal quotation marks omitted), cert. denied, 143 S. Ct. 151 (2022). A Napue claim thus “requires a showing of the [(1)] falsity and [(2)] materiality of testimony and [(3)] the prosecutor’s knowledge of its falsity.” Basden v. Lee, 290 F.3d 602, 614 (4th Cir. 2002).

16

Again, like the district court, we will assume that Juniper could establish falsity and knowledge, and focus only on the materiality prong.

Materiality under Napue is easier for a petitioner to satisfy than it is for other types of Brady claims. 13 United States v. Cargill, 17 F. App’x 214, 227 (4th Cir. 2001) (per curiam) (unpublished but orally argued) (noting that the Napue standard “is less onerous than the Brady one”); accord Ellis, 121 F.3d at 915 & nn.5–7 (explaining the difference between the Brady and Napue standards). While a petitioner advancing a regular Brady claim must show a reasonable probability of a different result had he been aware of the suppressed evidence, a petitioner can show materiality for Napue purposes “if there is any reasonable likelihood that the false testimony could have affected the judgment of the jury.” 14 United States v. Agurs, 427 U.S. 97, 103 (1976) (emphasis added), holding

13

The Tenth Circuit has concluded that the Napue and Brady standards “ultimately mandate the same inquiry” “for all practical purposes” because a petitioner who satisfies the Napue inquiry “will still have to meet the harmless error standard of Brecht v. Abrahamson, which the Supreme Court has held is met by the [Brady] test.” Douglas v. Workman, 560 F.3d 1156, 1173 n.12 (10th Cir. 2009) (citation omitted) (first citing Brecht v. Abrahamson, 507 U.S. 619 (1993); and then citing Kyles, 514 U.S. at 435–36). We are not so sure. Certainly, in Kyles v. Whitley, the Supreme Court concluded that once a Brady error has been established, “it cannot subsequently be found harmless under Brecht.” Kyles, 514 U.S. at 436. But that does not mean that a Napue error would not also automatically satisfy Brecht, a question that Kyles explicitly left open. Id. at 433 n.7. Our sister circuits are split on this issue. See Haskell v. Superintendent Greene SCI, 866 F.3d 139, 150 (3d Cir. 2017) (collecting cases and discussing the circuit split). We need not resolve this matter—which the parties have not briefed—today because we do not find any meritorious Napue claims here.

14

In a 2016 per curiam decision, Wearry v. Cain, the Supreme Court may have implied that it views the two standards as equivalent. There, it cited the Napue standard when discussing a Brady claim. Wearry v. Cain, 577 U.S. 385, 392 (2016) (per curiam). If it did intend to equate the two standards, it is unclear whether this should be understood to make the Napue standard more difficult to satisfy or make the Brady standard easier.

17

modified by Bagley, 473 U.S. 667; accord Daniels v. Lee, 316 F.3d 477, 493 (4th Cir. 2003); United States v. Arias, 217 F.3d 841, at *6 (4th Cir. 2000) (per curiam) (orally argued but unpublished table decision). As with other Brady claims, the materiality of multiple Napue claims should be assessed cumulatively. See Arias, 217 F.3d 841, at *6.

Juniper sought, and this Court granted, a certificate of appealability only on the cumulative materiality of the alleged Brady and Napue material in this case. This Court has not previously resolved which standard—Brady, Napue, or some mix of the two—applies when assessing the cumulative materiality of Napue and “run-of-the-mill Brady claims” together. Juniper, 2021 WL 3722335, at *28. The district court and our prior, unpublished decision in United States v. Arias avoided the question by concluding that the result would be the same under either standard. See Arias, 217 F.3d 841, at *6; Juniper, 2021 WL 3722335, at *28. Because the question is squarely presented by this case, and because the

Compare Brown v. Louisiana, 143 S. Ct. 886, 887 (2023) (Jackson, J., dissenting from the denial of certiorari) (arguably taking the latter approach), with Prystash v. Davis, 854 F.3d 830, 838 n.6 (5th Cir. 2017) (rejecting the latter approach). Wearry was a per curiam, unargued decision “without full briefing” that made only a passing reference to the standards in question and did not provide clear guidance on this point. Wearry, 577 U.S. at 392 (Alito, J., dissenting). “That procedural posture is not one in which we would expect the Supreme Court to change the standard for an important issue like Brady materiality.” Prystash, 854 F.3d at 838 n.6. And in post-Wearry decisions, the Supreme Court and this Court have continued to describe the Brady standard in the traditional way. E.g., Turner v. United States, 582 U.S. 313, 324 (2017); Bowman, 45 F.4th at 752. Accordingly, we continue to understand the Brady and Napue standards as distinct. See United States v. Ausby, 916 F.3d 1089, 1093 n.1 (D.C. Cir. 2019) (noting Wearry’s discussion of the standards, but concluding that the two standards are distinct); cf. United States v. Dodge, 963 F.3d 379, 384 (4th Cir. 2020) (adhering to “prior Fourth Circuit precedent that could be read as being in tension with intervening Supreme Court reasoning” where there was “no directly applicable Supreme Court holding”).

18

parties extensively briefed it below, we will address it.

Two sister circuits have addressed this question: the Second and the Ninth. The Ninth Circuit explained the tension in determining which standard to apply:

Although we must analyze Brady and Napue violations “collectively,” the difference in the materiality standards poses an analytical challenge. The Napue and Brady errors cannot all be collectively analyzed under Napue’s “reasonable likelihood” standard, as that would overweight the Brady violations. On the other hand, they cannot be considered in two separate groups, as that would fail to capture their combined effect on our confidence in the jury’s decision.

Jackson v. Brown, 513 F.3d 1057, 1076 (9th Cir. 2008). The Ninth Circuit’s solution was a two-step process: the Court should “first consider the Napue violations collectively” under the Napue standard, and if they “are not material standing alone,” then, second, the court should “consider all of the Napue and Brady violations collectively” under the Brady standard. Id.

For its part, the Second Circuit has noted that “undisclosed Brady material [may]

undermine[] the credibility of specific evidence that the government otherwise knew or should have known to be false.” United States v. Vozzella, 124 F.3d 389, 392 (2d Cir. 1997). In other words, Brady material could cast light on the falsity of Napue material. In such situations, the Second Circuit held, the Napue standard should apply. Id. In Arias, this Court interpreted the Second Circuit’s holding as being limited to where “the undisclosed evidence was not only related to the false testimony, but it actually demonstrated the falsity of the testimony.” Arias, 217 F.3d 841, at *6.

In briefing below, the parties converged on an analysis combining the Second and Ninth Circuits’ approaches. Under this proposal, the Court would apply the Napue standard

19

to a cumulative analysis of the Napue claims, plus any Brady claims that showed the falsity of the testimony; and then the Court would apply the Brady standard to a cumulative analysis of all the Brady and Napue claims.

We find this approach—which combines the insights of the Second and Ninth Circuits and is consistent with Arias—to make good sense, and we adopt it. Thus, we hold that a court considering the cumulative materiality of Brady and Napue claims must (1) evaluate, pursuant to the Napue standard, the cumulative materiality of the Napue evidence and any Brady evidence tending to show the falsity of the testimony at issue in the Napue claims; and (2) evaluate, pursuant to the Brady standard, the cumulative materiality of all of the Napue and Brady evidence. To be clear, however, the analysis need not take place in this order. For example, if it is evident that the petitioner would succeed at step (2), there is no need to jump through the hoop of the step (1) analysis as well. But to deny habeas relief, of course, both steps will need to be evaluated.

C.

Before considering cumulative materiality, we must determine exactly what it is we are accumulating. Certainly, we must consider the cumulative impact of the evidence presented in the Brady and Napue claims before us, combined with the evidence at trial. E.g., Monroe, 323 F.3d at 302. But may we also consider other evidence, such as evidence presented in Brady claims that we are not otherwise considering because they are barred by the mandate rule, or evidence that was not suppressed and not introduced at trial but may shed light on evidence that was suppressed?

The district court answered this question in the affirmative, noting that the

20

materiality analysis “considers what effect suppression ‘might have had on the preparation or presentation of [the petitioner’]s case,’” and that these categories of evidence may be relevant for that purpose. Juniper, 2021 WL 3722335, at *16 n.30 (quoting Bagley, 473 U.S. at 683).

By contrast, at least one sister circuit (the Eleventh) has limited the cumulative-

materiality analysis only to evidence that survived the other prongs of the Brady test and was not procedurally barred, albeit without explicitly considering the Bagley rule cited by the district court. See Smith, 572 F.3d at 1337, 1342. We have suggested the same in unpublished authority. See United States v. George, 466 F. App’x 304, 309 n.1 (4th Cir. 2012) (per curiam).

While such a rule is easy to apply, it does not account for the full cumulative-

materiality analysis we are instructed to undertake. In considering the cumulative materiality of the Brady claims before it, a “‘court may consider directly any adverse effect that the prosecutor’s failure to respond might have had on the preparation or presentation of the defendant’s case’ and ‘should assess the possibility that such effect might have occurred in light of the totality of the circumstances and with an awareness of the difficulty of reconstructing in a post-trial proceeding the course that the defense and the trial would have taken’ had full disclosure been made.” Moseley v. Branker, 550 F.3d 312, 318 (4th Cir. 2008) (quoting Bagley, 473 U.S. at 683); cf. Agurs, 427 U.S. at 112 (holding that materiality “must be evaluated in the context of the entire record”).

We thus agree with the district court that we must consider whether there is evidence of which the defendant was aware at trial, but chose not to pursue or present, that he may

21

have treated differently had he been aware of the suppressed evidence. Similarly, we agree with the district court that evidence underlying Brady claims that are barred by the mandate rule may be considered in the cumulative-materiality analysis to the extent that evidence may have affected the “preparation or presentation of the defendant’s case.” 15 Bagley, 473 U.S. at 683.

In sum, in conducting the cumulative-materiality analysis, we may consider how non-Brady evidence bears on the preparation or presentation of Juniper’s case.

III.

The applicable legal principles having been established, we turn to the question before us: whether the evidence cited in Juniper’s remaining Brady and Napue claims is cumulatively material. We begin with a detailed recitation of the evidence at trial, the evidence at issue in the remaining claims, and other evidence we will consider in light of our holdings above. We then turn to the cumulative-materiality analysis, assuming for purposes of this analysis that Juniper can satisfy the other prongs of his remaining Brady and Napue claims. We conclude that the alleged Brady and Napue evidence, while

15

Here, the relevant claims barred by the mandate rule are further complicated by the fact that those claims were rejected by the state Supreme Court. Under § 2254, “a determination of a factual issue made by a State court shall be presumed to be correct,” and a federal court may not grant habeas relief on a claim rejected on the merits by the state court unless the state-court decision “was contrary to, or involved an unreasonable application of, clearly established Federal law” or “was based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding.” 28 U.S.C. § 2254(d), (e). So, even if the claims at issue were not barred by the mandate rule, Juniper would face a steep climb to demonstrate individual Brady violations as to claims previously adjudicated in state court. Nevertheless, in evaluating cumulative materiality, we may consider what effect this allegedly suppressed evidence may have had on the defense’s preparation and presentation of its case.

22

substantial, cannot satisfy the combined Brady/Napue cumulative-materiality test laid out above.

A.

The evidence at trial was as follows. 16 i.

In January 2004, Keshia lived with her children in a second-floor apartment in Norfolk, Virginia. The apartment was accessible via a staircase at the back of the building that led directly to her door. Hers was one of two second-floor apartments in the two-story building, each accessed with its own staircase. The building also included three downstairs apartments.

Keshia’s entryway led directly into a living room with an attached kitchen, through which the dining room was accessible. Beyond the living room lay a hallway, from which several other rooms branched off. First, on the right, was a hallway bathroom with a tub. The bathroom was followed by the children’s bedroom on the left and Keshia’s bedroom on the right. Keshia’s bedroom contained a bed to the left upon entering, a chest of drawers along the right-hand wall, and a dresser along the far wall, beyond the bed. A smaller bathroom with a shower was accessible to the right, past the chest of drawers. We have provided the crime-scene diagram that the Commonwealth introduced at trial, which

16

The parties’ Joint Appendix on appeal contains only excerpts of the trial transcript. The full trial transcript also does not appear in electronic form on the district court docket; rather, the district court received the transcript in paper form. Accordingly, we rely on the district court’s description of those portions of the trial transcript that have not been provided to us.

23

depicts the layout of Keshia’s apartment, in the Appendix to this opinion.

At 12:44 PM on January 16, 2004, someone placed a 911 call reporting possible shots fired at an upstairs apartment in Keshia’s building. 17 Officer Thomas Atkinson responded to the call around 12:50 PM and walked up a set of stairs toward a second-story apartment, which was not Keshia’s. As he was about to knock on the door, its occupant— Chaunte Hodge—approached from the parking lot and asked if she could help him. After speaking with her, Atkinson went downstairs and spoke with Hodge’s elderly mother-in- law, Frances Frazier, who lived in the apartment below Keshia’s. Hodge and Frazier each denied making the 911 call, with Hodge explaining that she had not been home. Atkinson asked both women if they had heard shots fired. Based on his conversations with them, and neither hearing nor seeing anything suspicious, Atkinson believed the call to have been a false report. In light of this testimony, defense counsel “attempted to establish that no gunshots had been fired at the time” that Atkinson responded to the 12:44 PM 911 call. J.A. 1002. We will see that, if true, that would fly in the face of the prosecution’s case against Juniper.

Officer Atkinson stood in the parking lot of the apartment building for fifteen to twenty minutes, chatting with a second officer, Paul Liverman, who had arrived as Atkinson’s backup. Both officers left the apartment complex around 1:15 PM. Atkinson testified that he “had no idea” that there was another stairwell around the back of the building or another second-floor apartment, namely, Keshia’s. J.A. 74, 76. Liverman also

17

Although it is not clear from the record before us that the jury was made aware of this fact, the caller is identified in the 911 call log as a woman.

24

testified that he was unaware of the second upstairs apartment and that he saw no disturbances at the scene.

Police received a second 911 call sometime later, 18 to which Officer W.G. Snyder and another officer responded around 2:20 PM. This time, the officers located and climbed the correct stairway to Keshia’s apartment. Officer Snyder testified that the “whole center part of the [front] door was completely knocked . . . inward into the apartment, and wooden debris from the door was lying inside the apartment.” Juniper, 529 F. Supp. 3d at 479 (second alteration in original). The officers entered the apartment, at which point they discovered four-year-old “Nykia’s body lying across Rueben on the bed” in Keshia’s bedroom, and two-year-old “Shearyia’s body lying across Keshia’s body on the floor beside the bed.” Id. They “received no response” from the victims. Id.

After satisfying themselves that there was nobody else in the apartment, the officers were about to exit when they noticed a cigarette butt lying on top of the debris from the door. They had not noticed it upon first entering the apartment; Officer Snyder testified that, “[d]ue to the nature of the scene,” his focus at that moment was on whether they were about to confront someone in the apartment, and therefore he was “really not paying attention” to what was on the floor. J.A. 91. After exiting, the officers posted themselves outside the apartment to protect the scene until additional officers arrived.

“Brian Nichols, the paramedic who responded to Officer Snyder’s call, and who testified as an expert in emergency response, got to the apartment at about 2:38 [PM].”

18

It is not clear whether the precise time of this call was provided at trial. According to the 911 call log, this call came at 2:12 PM and was placed by Melinda Bowser.

25

Juniper, 2021 WL 3722335, at *11. “He examined the bodies and pronounced all four dead based on his findings that none of them were breathing and they all had lividity, or ‘pooling of blood in the skin,’ in their extremities.” Id. He also noted “dark, dried blood” on Keshia’s shirt. Id. “He testified that, based on his training, the dried blood on Keshia and the extent of the lividity in each victim indicated ‘that this was not an incident that occurred within the last hour.’” Id. This testimony thus placed the time of death before 1:40 PM. However, the medical examiners and autopsy reports did not provide a time of death.

The medical evidence showed that all four victims “had been shot, and Keshia had been stabbed in the abdomen.” Id. at *1. “The stab wound did not fatally wound Keshia, but tore through the muscle of her abdominal wall. There was a great deal of blood accompanying the wound, . . . which led the medical examiner performing the autopsy to conclude that the stab wound was probably the first injury inflicted on Keshia.” Juniper, 529 F. Supp. 3d at 477 (quoting Juniper, 626 S.E.2d at 394). Keshia had also been “shot three times,” including twice in the abdomen, “and grazed by a fourth bullet.” Id. (quoting Juniper, 626 S.E.2d at 394). “Shearyia was shot four times while in her mother’s arms,” including once in the head. Id. (quoting Juniper, 626 S.E.2d at 394). Keshia’s body was located on the floor of her bedroom, between the bed and the dresser, with Shearyia lying on top of her. Their bodies were not visible from the bedroom’s doorway.

Rueben was shot three times in his legs and torso, striking his pelvic bone and breaking his femur. Id. “The medical examiner testified that the broken bones would have caused excruciating pain and immediately disabled Rueben.” Id. (quoting Juniper, 626 S.E.2d at 394). Finally, Nykia was shot once in the head. Id. at 478. “The medical examiner

26

testified that the bullet’s path was consistent with Nykia ducking her head and body toward the shooter prior to being shot.” Id. (quoting Juniper, 626 S.E.2d at 394). Her body was found lying across her uncle’s on the bed.

The two children were naked when police found them, and the medical examiner who performed their autopsies “testified that Shearyia had what appeared ‘to be ink marks or magic marker marks on her skin,’ including ‘on her nose and cheek.’” Juniper, 2021 WL 3722335, at *4 n.14; see id. at *28. The prosecution also introduced photographs of Shearyia showing “magic marker ink on the side of her face” and on her back. Juniper, 626 S.E.2d at 415. These details would become relevant to the prosecution’s theory of the case.

Investigators quickly zeroed in on Juniper as a suspect in the homicides. Juniper was Keshia’s “on-again, off-again” boyfriend, and police received various reports on the day of and days immediately following the murders that Juniper was abusive toward Keshia, as well as rumors that he had confessed. Juniper, 529 F. Supp. 3d at 478 (quoting Juniper, 626 S.E.2d at 394). At trial, “[o]ne of Keshia’s best friends testified that she learned about Juniper’s and Keshia’s relationship about a year and a half before Keshia’s murder,” and “[s]he described Juniper as ‘real jealous, insecure, [and] abusive mentally and physically.’” Juniper, 2021 WL 3722335, at *11 (third alteration in original).

The most important eyewitness at Juniper’s trial was his friend of several years, Renee Rashid. Rashid testified that on the morning of January 16, 2004, she received a call from Juniper asking for a ride to Keshia’s apartment “to get his shit.” J.A. 118. The request was not unusual—Rashid gave Juniper rides around town “whenever [she] could,” and Juniper “seemed normal” on the phone call. J.A. 107, 155. Rashid explained that she was

27

not sure about the precise timing of events that day, but agreed that the call from Juniper took place around 9:20 AM.

Juniper lived with his mother, Gwendolyn Rogers. So, upon agreeing to drive him to Keshia’s apartment, Rashid drove to Rogers’s home, arriving around 9:40 AM. Juniper requested that Rashid pick up cigarettes for him, so she drove to a corner store a couple of blocks away to buy the cigarettes. Rashid then returned to Rogers’s house, where she went inside and encountered Juniper, Rogers, and Juniper’s friend, Keon Murray, who was lying on the couch. Rashid stated that she did not notice any bulges on Juniper’s clothing suggesting that he had a gun or a box of ammunition on him. Rashid and Juniper soon departed, and Rashid drove them to Keshia’s apartment, arriving between 10:15 and 10:20 AM.

Juniper and Rashid entered Keshia’s apartment, where Rashid saw each of the four victims alive. Rueben was lying on the couch in the living room, seemingly asleep. Rashid spoke to Keshia—whom she had known through Juniper for about a year—while Juniper “went straight to” Keshia’s bedroom. J.A. 117. During their conversation, Keshia told Rashid that she was about to give the children a bath. Juniper then called Rashid to come join him, and she did, assisting him in disconnecting a DVD player from a television in Keshia’s bedroom. She then went to the hallway bathroom, where the two children were standing in the bathtub. Shearyia had “magic marker writing” on her face, which Rashid took to have been the work of Nykia. J.A. 122. Rashid testified that she gently teased the children about the ink: “I said little mama, you wrote on your little sister’s face. I said you

28

shouldn’t have been writing on your sister’s face, and they were laughing.” Id.

While talking to the two girls in the bathroom, Rashid overheard sounds “like items being moved around” in the bedroom, after which Keshia walked toward the bedroom, asking Rashid as she passed by what Juniper was doing. J.A. 122. Rashid told her she didn’t know. Rashid then heard Keshia speaking to Juniper, repeatedly making comments such as, “There’s nobody but you,” and “I told you I’m not seeing anybody but you,” in a “reassuring voice.” J.A. 123. Rashid could also hear Juniper speaking, but could not understand what he was saying because he was “mumbling real low.” J.A. 124. Rashid then heard Keshia’s “voice start[ing] to rise,” so she “assume[d] that [Juniper] was accusing Keshia of being with someone else.” Id.

Afraid an argument was about to break out, Rashid walked to the door of the bedroom, where she saw Keshia sitting on the end of the bed and Juniper leaning toward her. Rashid told Juniper she was “ready to go” and encouraged him to leave. J.A. 125. Juniper handed Rashid the DVD player and followed her to the door of the apartment. They passed Rueben, who was still lying on the couch. Rashid exited the apartment first and, hearing the door shut behind her, assumed Juniper was still following her as she began to descend the staircase. But as she was doing so, Rashid heard Juniper call out Keshia’s name, followed by a “loud boom” that she described as “sound[ing] like the door being kicked in.” J.A. 127–28.

Frightened, Rashid hurried to her car without stopping to look behind her. She got into the car, which was parked on the street in front of Keshia’s building, and rolled down the window on the side nearest Keshia’s apartment. She could hear Keshia crying and

29

repeating her statement that she was not seeing anyone but Juniper. After “it got quiet,” Rashid honked her horn and shouted to alert Juniper that she wanted to leave. J.A. 129. Juniper yelled at Rashid to “[g]o ahead,” so she began to drive slowly away. J.A. 130. As she was driving, she heard “four booms rapidly in succession”—covering a span of about two seconds—from the direction of Keshia’s apartment. Id. She testified that the booms “sounded like gunshots,” though she admitted she “wasn’t sure they were gunshots at first.” J.A. 130–31. She “was scared,” so she left and drove back to Rogers’s house “because [she] couldn’t get [Juniper] out of the apartment.” J.A. 131, 174. Rashid was not sure how long she had spent at Keshia’s apartment.

Juniper’s longtime good friend, Keon Murray, also testified at trial, and provided his view of what occurred after Rashid returned to Rogers’s home from Keshia’s apartment. Murray explained that he awoke at Rogers’s house on the morning of January 16 to a “commotion,” with Rashid and Rogers “arguing.” J.A. 201. Rashid was crying and seemed to be “in shock,” and started to tell Murray “what had happened.” J.A. 201–02. For her part, Rashid confirmed that Murray was still at the house when she returned and testified that she told Rogers that Juniper was not leaving Keshia’s apartment, though she did not describe “the other things that [she] had seen and heard.” J.A. 131–32.

Rashid stated that Rogers “picked up the phone” to call Keshia’s apartment and made a phone call, but nobody answered. J.A. 132. Phone records, however, showed an incoming call to Rogers’s house from Keshia’s apartment at 11:34 AM, lasting three minutes and ten seconds. Murray explained that Juniper called Rogers’s house from Keshia’s apartment, and that Murray was the one who spoke to Juniper. He stated that he

30

knew Juniper was calling from Keshia’s apartment because of caller ID. Murray testified that Juniper told him, “[t]hey gone,” and that Keshia’s apartment was surrounded. J.A. 212. Juniper also stated that he “killed them,” which Murray understood to mean the people in the apartment. J.A. 212. Murray stated that he did not believe Juniper when he heard his confession because Juniper “wasn’t that kind of person” and “loved these children.” J.A. 228.

Further testimony came from Juniper’s acquaintance, Tyrone Mings, who was good friends with Murray. Mings lived with his girlfriend, Melinda Bowser, 19 about a block away from Keshia. He knew Keshia through Bowser, because the two women worked together.

Mings testified that he got up early on January 16. Sometime that morning, he was at home when he received a call from “Keon”—presumably Keon Murray. J.A. 243. According to Mings, Murray “told me wanted [sic] to check on somebody at Keshia’s house or whatever that he heard some shots.” Id. Mings’s testimony about this phone call was not corroborated at trial, either by phone records or by testimony from Murray (who was not asked about it). Mings explained that he also wanted to go to Keshia’s apartment to pick up a check for Bowser from the women’s shared workplace.

Mings testified that it “took [him] a while,” but that he walked over to Keshia’s apartment “eventually.” Id. As he mounted the apartment stairs, Mings observed that “[t]he [front] door was kicked in” and had “a big hole” in it. J.A. 247. He began calling for Keshia

19

Bowser did not testify at trial.

31

and Juniper. He stepped over the broken-in door and entered the apartment, where he saw Juniper pacing to and from the living room window with “a[n automatic] gun in his hand, [and] some white stuff on his face.” J.A. 249. Mings explained that he felt “kind of shook up” upon encountering Juniper in this way. Id.

Mings asked Juniper what was going on and where Keshia was. Juniper directed Mings to the bedroom, so Mings went to the doorway of that room, where he saw “a dude and a baby laying on the bed.” J.A. 251. He did not see Keshia, so he called her name and “was saying yo to the baby and the dude that was on the bed,” but did not receive a response. Id. He returned to the living room and told Juniper he had not seen Keshia. Juniper responded that she was “in between the bed and the dresser.” Id. Mings returned to the bedroom “and called [Keshia’s] name” again, still receiving no response. J.A. 252. Mings explained that he left Keshia’s apartment at that point.

Upon returning to his apartment, Mings told Bowser what he had seen, and the pair decided to call the police. There is no record evidence that they actually called police at that time. But Mings’s testimony was merely that they decided to call the police; he did not state that they in fact did so at that moment.

Meanwhile, back at Rogers’s house, Rogers asked Rashid to pick up Juniper’s cousin, John Jones, Jr. 20 Rashid “was under the impression” that Rogers hoped Jones could get Juniper out of Keshia’s apartment. J.A. 133. Because Rashid did not know where Jones lived, Murray went with her. They drove to Jones’s apartment and picked up Jones and his

20

Jones did not testify at trial.

32

girlfriend’s infant daughter. Rashid then drove Murray, Jones, and the baby to Keshia’s apartment. Murray and Jones went to look for Juniper, while Rashid stayed in the car with the child.

Murray testified that he only walked “to the back of the car,” while Jones “walked to the sidewalk that [led] to the stairway to [Keshia’s] apartment.” J.A. 214. Murray and Rashid both explained that Jones called to Juniper and that Juniper came out of the apartment. Murray testified that, when Juniper exited the apartment, Murray saw that he was carrying a black and chrome automatic weapon in his hand. Juniper, Murray, and Jones got into Rashid’s car, with Juniper sitting in the front passenger seat. Rashid then drove off.

At the same time, Mings had begun to walk back toward Keshia’s apartment “to check on what was going on down there.” J.A. 254. He testified that as he was walking, he saw Juniper, Murray, and Jones leave the apartment and get into a car occupied by “a female,” which then departed. J.A. 256. For his part, Murray testified that, when the car arrived at Keshia’s apartment, he saw “Stickman Ty” “standing” “[u]p the street . . . [t]alking on the phone.” J.A. 220, 230. This appears to be a reference to Mings; Mings’s nickname was Ty, and Murray explained that “stickman” meant “friend.” J.A. 229.

After witnessing Juniper leave Keshia’s apartment, Mings returned home, and Bowser called the police. The prosecution argued that this was the 12:44 PM call evidenced in the trial record. Standing in front of their own apartment up the street from Keshia’s, Mings and Bowser watched the police arrive and then depart. This aligns with the accounts given by Officers Atkinson and Liverman, who responded to the 12:44 PM 911 call but

33

left within half an hour, believing the call to have been a false report.

Rashid testified that meanwhile, in the car, Juniper was “jittery” and “breathing real hard,” and “wasn’t the same person that [she] dropped off” earlier that morning. J.A. 137. “He was looking in the mirror of the sun visor” with “his eyes . . . wide open,” and he kept saying things like “[t]hey’re behind us.” J.A. 138, 181. Rashid testified that Jones told Juniper that there was nobody behind them, to stop looking in the mirror, and to “[w]ipe [his] face.” J.A. 138. Later in the car ride, she saw a black gun in Juniper’s lap. The gun’s barrel was pointed toward Rashid, and she “felt threatened.” J.A. 185. Murray stated that Juniper “look[ed] nervous,” “[l]ike he was in shock,” and that he had a “powdery substance on his face” that Murray thought was cocaine. J.A. 216, 218. Murray also testified that Juniper “pulled [the car’s sun visor] down and . . . was looking in it.” J.A. 218.

Rashid and Murray contradicted each other in terms of who got out of the car where.

But ultimately, Rashid explained that once she had dropped off Juniper, Murray, Jones, and the baby, she returned home and called Rogers “as soon as [she] walked in the door.” J.A. 188. She said that she told Rogers she hoped Juniper and Keshia “hadn’t been fighting,” and that if they had been, she would “never take him over there again.” Id. Phone records showed a call from Rashid’s house to Rogers’s house at 1:10 PM. Rashid testified that she did not learn about the murders until three days later, on January 19.

Mings testified that, having witnessed the police leave Keshia’s apartment complex soon after arriving, he “figure[d] they didn’t go up” to Keshia’s apartment. J.A. 257. Accordingly, he and Bowser decided to call the police again. The prosecution contended that this was the second 911 call evidenced in the record, to which Officer Snyder and his

34

partner responded. This time, “when [Mings and Bowser] saw [the police] show up” to Keshia’s apartment, they decided to “walk[] down there with them.” J.A. 258. Mings testified that he told the police that he had seen victims inside the apartment, but did not tell them about having seen Juniper there.

The police collected forensic evidence from the scene. First were five cigarette or cigar butts: the cigarette butt that Officer Snyder spotted on top of the debris from the broken-down front door, plus three cigarette butts and a cigar butt located in an ashtray on the bed in Keshia’s bedroom. “Juniper could not ‘be eliminated as a source’” for the DNA on the cigarette butt located on top of the door debris, and “the likelihood of that DNA matching someone other than Juniper was ‘on[]e in greater than six billion.’” Juniper, 2021 WL 3722335, at *3 n.11 (alteration in original). As for the butts in the ashtray, “Keshia’s DNA was on the cigar butt and one of the cigarette butts,” and “[t]he DNA profiles on the other two cigarette butts did not match each other or any other known DNA.” 21 Id. at *3.

Next, officers located a knife blade and handle on the floor of Keshia’s bedroom.

The Commonwealth’s tool-mark expert “concluded that the knife blade and handle ‘at one time . . . were one piece,’” and “[t]he medical examiner who conducted Keshia’s autopsy testified that the knife blade was consistent with Keshia’s stab wound.” Id. at *2 (first

21

The sources of these DNA profiles were identified in 2010 as two men who are not otherwise mentioned in the record, except to clarify that one of them was Keshia’s brother. The district court excluded testimony from the two men at the evidentiary hearing because their identities were not known by anyone, including the Commonwealth, at the time of trial.

35

alteration in original).

The knife blade was about three inches long and had a serrated edge; it “[l]ooked like possibly a steak knife blade.” Id. (alteration in original). The blade “had ‘a very thin film of blood’ on it,” and “[t]he DNA on the blade came from a single source: Keshia.” Id. The Commonwealth also found DNA on the knife handle, which “came from a single source: Juniper.” Id. at *3. “The DNA expert testified that the DNA on the knife handle ‘was very strong’; she collected ‘a hundred times more’ DNA from the knife handle than she did from the knife blade.” Id. Further, “[t]he DNA on the knife handle was ‘an identical match’” to Juniper, “and the likelihood of the DNA on the knife handle matching someone other than Juniper was ‘one in greater than six billion.’” Id. at *3 n.9.

“The police also recovered a fingerprint on the right side of the knife blade toward the base of the knife, where the blade would have connected to the handle,” which “matched Juniper’s right thumb.” Id. at *2. While the Commonwealth’s fingerprint expert “acknowledged that it was ‘possible’ that a thumbprint left on a knife could remain intact on the knife if someone later used the knife to stab somebody,” the expert also explained that “‘[f]ingerprints are very fragile. They can be wiped away [or] smeared very easily just by the constant handling or the shuffling, touching of other objects against it.’” Id. (alterations in original). Juniper presented no evidence that he had ever used the knife innocently before.

While the gun used to commit the murders was never recovered, the police did recover ammunition from the scene. From the floor of the bathroom connected to Keshia’s bedroom, they recovered an “empty PMC, Caliber 9mm Luger ammunition box,” J.A. 314,

36

“and two loose unfired cartridges,” Juniper, 2021 WL 3722335, at *3. They further “recovered eight cartridge casings and two bullets from on and around the bed” in Keshia’s bedroom, “and medical examiners later recovered four bullets from the victims’ bodies during the autopsies.” Juniper, 2021 WL 3722335, at *3. The unfired cartridges, cartridge casings, and bullets were all nine-millimeter caliber, and the “type of cartridge case . . . would be fired [from] a nine-millimeter Luger semiautomatic pistol.” J.A. 99. According to the Commonwealth’s firearms expert, “[a]ll the cartridge casings were fired from the same firearm” and all six bullets were also fired from a single firearm, but without “a firearm to compare the cartridge casings and bullets to,” she could not “testify that the cartridge casings and bullets had been fired from the same firearm.” Juniper, 2021 WL 3722335, at *3 (emphasis added).

The police did not identify any prints of value—that is, quality prints that would be identifiable “to a particular individual,” J.A. 94—“from the cartridge casings or bullets recovered from on and around the bed, or from the ammunition box or loose cartridges recovered from the master bathroom,” Juniper, 2021 WL 3722335, at *3. They did recover one print of value from the cartridge tray inside the ammunition box, but they “were unable to identify it to any particular individual,” including Juniper. J.A. 94.

Juniper was incarcerated awaiting trial. A fellow inmate, Ernest Smith, testified that he knew Keshia. He stated that on October 13, 2004, a few months after he had met Juniper, the two were playing chess when he asked Juniper whether he had committed the murders. Juniper confessed that he had, and told Smith that he killed the children “because he didn’t want to leave any witnesses.” J.A. 273. He also told Smith that “he couldn’t sleep because

37

he kept having nightmares” and “was seeing [Keshia] in his dream[s].” Id. Smith did not ask further details about the crimes because he “didn’t want to hear no more.” J.A. 285.

Finally, the crime scene video did not show blood outside of Keshia’s bedroom, nor was there any evidence that Keshia sought medical attention for her stab wound or even tried to use a towel or other bandage to stanch the flow of blood. As discussed below, this is highly suggestive circumstantial evidence that the stabbing and shootings occurred in very close temporal proximity.

ii.

With all of the foregoing evidence, the Commonwealth was able to present a coherent narrative at trial. Juniper and Keshia were in a romantic relationship, and he was prone to jealousy and was physically abusive. He had Rashid bring him to Keshia’s apartment on the morning of the murders, arriving by 10:20 AM. Sometime between 10:20 and 11:34 AM, Juniper and Keshia argued about whether she was seeing anyone else, and Rashid heard what sounded like Juniper kicking in the front door, followed by gunshots. Juniper shot all four victims in Keshia’s bedroom, and also stabbed Keshia before shooting her. He then remained in the apartment for some time, using cocaine. At 11:34 AM, Juniper called his mother’s house from Keshia’s apartment and spoke to Murray, to whom he confessed. Murray, in turn, called Mings, who ventured to the apartment, saw Juniper there with a gun, and saw two of the victims deceased. Around the same time, Rashid and Murray picked up Jones and then went to retrieve Juniper from Keshia’s apartment. Mings saw them leave, at which point Bowser called the police at 12:44 PM.

The police arrived but did not go to the correct apartment, and left around 1:15 PM,

38

believing the call to have been a false report. Simultaneously, Rashid returned home and called Juniper’s mother, having dropped off Juniper and the other people in her car. Mings and Bowser saw the police leave and decided to place a second 911 call. This time, the police discovered the kicked-in door and the bodies. A paramedic pronounced them dead around 2:40 PM, and concluded that they had been deceased for at least an hour.

In sum, the essential elements of Rashid, Mings, and Murray’s testimony aligned with each other and with the phone calls in the records; forensic evidence supported that Juniper stabbed Keshia—who was likely shot very shortly thereafter, given the lack of blood outside her bedroom or evidence that she had sought to stop the bleeding—and that he was in the apartment, smoking a cigarette, after its front door was kicked in; and Murray and Smith both testified that he had confessed to them.

iii.

Juniper did not present any witnesses in his defense at the guilt phase. Instead, defense counsel primarily sought to damage the Commonwealth’s case by undermining the credibility of the Commonwealth’s witnesses.

Rashid admitted that, although she learned within days that Keshia and her family had been murdered, she waited until January 29 to speak with police—and only did so after speaking to a lawyer, who brought her to the police station and joined her for the interview. She said that she delayed going to the police because she was afraid of Juniper and “didn’t know what to do.” J.A. 187. Additionally, she conceded that she spoke to police for some time before her tape-recorded statement was taken. When asked whether the police “g[a]ve [her] any details” about the crime, scene, or victims, Rashid testified that she did not recall.

39

J.A. 189.

Defense counsel cross-examined Rashid about discrepancies between her trial testimony and her January 29 statement to police. For example, defense counsel read from her statement to police that she first learned about the murders on January 17—the day after they occurred—rather than January 19, as she stated at trial. Rashid’s response was that she did not remember when she first heard the news.

More importantly, Rashid also admitted that during the January 29 conversation with police, she never mentioned Murray’s presence with her on the trip to pick up Jones and then Juniper. Instead, she told investigators that Jones just “showed up” at Rogers’s house. J.A. 176, 320. Rashid explained that “when [the investigator] was asking [her] questions the first time, [she] initially spoke too soon and [she] was afraid to recant [her] answer because [she] was afraid he would think [she] was lying.” J.A. 176–77. She returned to the police station with her attorney on February 3 and gave another taped statement to rectify this omission. Finally, Rashid acknowledged that she had an immunity agreement with the prosecution as to any charges for misdemeanor accessory after the fact. 22 Further, the jury had numerous reasons to question the credibility of Mings, Murray, and Smith. Ming’s account of the phone call from Murray was not corroborated at trial. And Mings acknowledged that he did not tell the officers who arrived in response to the second 911 call that he had seen Juniper in Keshia’s apartment—and indeed, that he did

22

Her immunity agreement did not cover any felony offenses.

40

not implicate Juniper at all until “a couple months” later. J.A. 258. He testified that he omitted Juniper from his initial description because he “feared for [his] safety,” since he thought Juniper—whom he only knew “a little”—“might consider [him] a witness against him.” J.A. 258, 263–64. Defense counsel sought to discredit this explanation by noting that Mings’s failure to implicate Juniper on the day of the murders kept Juniper on the street, as an ostensible threat, for longer. Counsel also questioned Mings as to why, if he saw Juniper “standing between [Mings] and the front door with a gun in his hand . . . [a]nd cocaine all over his face,” and Mings “thought [Juniper] was out of his mind,” Mings had remained in the apartment long enough to go to the back bedroom “not once, but twice” and did not initially tell the police about Juniper’s supposed presence. J.A. 265.

The jury also heard that Mings had prior felony convictions, and Mings agreed that he was wearing “jail attire” during his testimony and explained he was “currently serving a sentence” for “a probation violation on [his] previous convictions.” J.A. 240, 260. Defense counsel drew out that Mings had spoken to his lawyer about his testimony in this case and that the judge who had sentenced him for the probation violation was aware of his cooperation, such that Mings knew he was “saving [him]self as much as four years in the penitentiary” “by agreeing to testify for the Commonwealth.” J.A. 262.

Like Mings, Murray was wearing a “[j]ail jumpsuit” during his testimony, explained he was in custody for “[a] felony violation,” and conceded that he also had prior felony and misdemeanor convictions. J.A. 194. Additionally, during the prosecution’s preliminary questions, he repeatedly claimed that he did not remember anything about the day of the murders before saying, “Your Honor? I mean, I can’t do this shit.” J.A. 199. Shortly

41

thereafter, he asked if he had to testify. And when asked about what Rashid told him when she first returned to Rogers’s house from Keshia’s apartment, he tried to “plead the Fifth.” J.A. 202. At that point, the court excused the jury, and the prosecution granted Murray immunity for any possible charges of misdemeanor accessory after the fact or misprision of a felony. The jury then returned to the courtroom, and Murray’s testimony continued without further incident. The jury later learned about the immunity agreement through Murray’s testimony.

Murray conceded that he did not speak with police about the case at all until “five or six[] weeks” after the murders, 23 even though he supposedly heard a confession from Juniper on the day of the murders and found out later that night that the family had in fact been killed. J.A. 224. And he conceded that he only spoke to the police when “they caught up with” him. Id. Indeed, police called him and arranged an appointment, which he failed to keep the first time. Murray explained his delay in speaking to the police as arising out of fear.

Murray further conceded that he knew his “cooperation could be brought to the attention of” the court that had recently sentenced him for a probation violation. J.A. 226. But he testified he had not received any “promises” from the Commonwealth “in exchange for” his testimony, nor any threats of prosecution by the police investigators. J.A. 195. Rather, somewhat confusingly, he alleged that Juniper’s investigator, Wayne Kennedy, told him he could get charged with being an accessory to murder if he did not testify, and

23

It was actually only three weeks.

42

said that was why he testified. When defense counsel asked whether Murray had told Kennedy “that the police had threatened . . . to charge [him] if [he] didn’t cooperate,” Murray said he could not recall. J.A. 232.

Murray also conceded that he had told Kennedy some lies, such as that he was not at Rogers’s house on the day of the murders and did not go with Rashid to Keshia’s apartment. He further admitted that he had told Kennedy “that when the police questioned [him], [he] gave the police” that same information, even though “none of that was true.” J.A. 232–33. That is, the jury heard that Murray told Kennedy that he did not go to Keshia’s apartment, and that he told Kennedy that he had told police the same thing. When defense counsel pressed him on how the jury was to know “when [he was] telling the truth and when [he was] not,” Murray said he did not know. J.A. 233.

Smith also testified while wearing “[j]ail clothes” and explained that he contacted the police about a week and a half after hearing Juniper’s jailhouse confession, after “contemplating” whether he should do anything with the information. J.A. 274, 286. He acknowledged that he had a number of previous felony convictions, including several for which he was then serving a five-year sentence, but said he had not spoken with his attorneys about seeking a lower sentence based on his cooperation. However, he admitted he knew his “lawyers [could] bring to the attention of the sentencing judge the fact that [he had] cooperated in this case,” and that he was not opposed to that happening. J.A. 277.

Defense counsel drew out some of the implausible aspects of Smith’s story. When counsel asked Smith to name “every subject [he] [could] possibly think of that [he] and [Juniper] talked about” before October 13, 2004, Smith stated that they had only ever

43

spoken about chess. J.A. 282. Yet, when Smith asked “[o]ut of nowhere” on that date whether Juniper had committed the murders, Juniper immediately responded that he had. J.A. 284.

Defense counsel also argued in closing that the cigarette butt with Juniper’s DNA that was found on top of the door shards could have been kicked there by the officers responding to the second 911 call, who conceded that they were not paying attention to what was on the floor when entering the apartment.

Finally, the jury heard a number of weaknesses in the Commonwealth’s case, including some contradictions in the eyewitnesses’ accounts; that two people who lived in Keshia’s apartment building apparently had not heard anything suspicious as of 12:44, leading police to believe the 12:44 PM 911 call was a hoax; that Keshia was stabbed before being shot, leaving little time for the stabbing in Rashid’s account; that Rashid did not see any bulges on Juniper’s clothing suggestive of a gun or box of ammunition when he left his mother’s house that morning; that Rashid was not certain at first that what she had heard were gunshots; that the eyewitnesses were not certain of the timing of events; that both Mings and Murray identified the gun held by Juniper as automatic, when the cartridge casings located at the scene were the type that would be fired from a semiautomatic weapon; that DNA from two unknown individuals was found on cigarette butts in an ashtray on Keshia’s bed; that the gun used in the murders was never found; and that a fingerprint on the tray inside the ammunition box came from someone other than Juniper.

Despite defense counsel’s efforts to undermine the Commonwealth’s case, the jury

44

returned verdicts of guilty as to the four murders.

B.

Juniper’s remaining Brady and Napue claims provide numerous pieces of additional evidence that we assume for purposes of this opinion were suppressed and favorable (Brady) or known by the prosecution to be false or misleading (Napue). Specifically, that evidence includes (1) statements from Keshia’s neighbors, Wendy and Jason Roberts; (2) statements from two other neighbors, Chaunte Hodge and Patrick Danso; (3) conflicting statements from John Jones, Juniper’s cousin; (4) conflicting and prior statements from Tyrone Mings, who testified at trial, and his girlfriend, Melinda Bowser; (5) conflicting and prior statements from Juniper’s friend, Keon Murray, who testified at trial; (6) evidence that other inmates, in addition to Ernest Smith, told investigators that Juniper confessed to them, and that those inmates did so falsely in an attempt to secure benefits for themselves; and (7) evidence that conflicts with Officer Atkinson’s trial testimony. We consider each category of evidence in turn, noting why we do not believe it to be individually material. 24 i.

Wendy Roberts and her son Jason Roberts lived in the apartment building next door to Keshia’s building. On the day of the murders, Investigator Dennis Jones spoke with Wendy and Jason. He typed a one-page summary of their statements, which the parties have stipulated accurately reflected what they told him, except as to one point noted below.

24

Juniper did not appeal any individual-materiality determinations. However, we find it helpful to our cumulative-materiality analysis to explain the weaknesses in the individual materiality of each category of evidence.

45

Wendy told Investigator Jones that she saw a woman arguing with a man twice on the day of the murders: once around 9:30 or 10:00 AM, and later around 12:30 or 12:45 PM. During the second confrontation, the man told the woman that “she had not better [sic] be there when he returned.” J.A. 443. While Wendy did not identify either individual by name, her physical descriptions of them match Keshia but not Juniper. Wendy knew Keshia to be her neighbor and also recognized the man, who she thought was an ex-boyfriend because he came around “all hours of the day and night” beeping the horn for Keshia, and came by every morning “wanting to see the kids.” Id. Investigator Jones’s summary states that Jason was not home at the time of the arguments, but that he had seen the people in question “many times,” though the parties did not stipulate that Jason in fact said this. Id.

Wendy also said that at about 1:30 PM, she heard “three or four bangs” that she thought were firecrackers. Id. Jason, who was home by that time, also stated that he heard the bangs, which he thought were the sound of “someone hammering.” Id.

Later that night, Wendy spoke with Investigator William Conway. Investigator Conway wrote in his notes only that Wendy’s statement to him “was consistent with that of the interview given to [Investigator] Jones.” J.A. 394. The parties stipulated that this was true.

Wendy also spoke to Investigators Conway and Robert Ford the next day, at which time they showed her a six-person photo lineup that included Juniper. Wendy selected someone other than Juniper, although she said she was “not 100 percent sure.” J.A. 397. Jason was present for this interview, but there is no indication in the record of what, if

46

anything, he said.

Wendy and Jason’s statements to investigators immediately following the murders, if credited, would be significant. According to the prosecution’s timeline of the crimes, Keshia had already been killed by 12:30, when Wendy allegedly saw her arguing with a man who was not Juniper. And Wendy and Jason’s statements indicate that they heard gunshot-like noises at 1:30 PM, again placing the time of the murders later than in the prosecution’s theory, and at a time when there is no evidence that Juniper was at Keshia’s apartment.

In 2011—after Juniper belatedly and fortuitously learned of the existence of these contemporaneous statements—Wendy and Jason each signed an affidavit describing what they saw on the day of the murders. Their accounts in their affidavits differed in significant ways from their statements to Investigator Jones.

In her 2011 affidavit, Wendy explained that she heard (but did not see) a man and woman arguing in Keshia’s apartment the night before the murders. She made no mention of seeing Keshia on the day of the murders. Rather, she noted that she heard someone honking a car horn repeatedly that morning. Sometime between 1:00 and 2:30 PM, she “heard a series of loud pops” that she assumed were from a “truck backfiring.” J.A. 438. For his part, Jason stated in his affidavit that he was “home all day” on January 16, which contradicts Investigator Jones’s notes. J.A. 441. His affidavit also stated that he “heard gunshots” in the afternoon, which he recognized “because of [his] experience in Emergency Response.” Id. This account again contrasted with his statement to Investigator

47

Jones, in which he reported that the noises he heard sounded like someone hammering.

Wendy and Jason’s affidavits also each added a crucial detail absent from Investigator Jones’s notes: that immediately after hearing the popping noises, they saw a suspicious man leave in a car. Wendy explained that she “saw a man come out from the stairs leading up to” Keshia’s apartment and “walk[] quickly to a car.” J.A. 439. The man saw her watching and asked what she was “looking at.” Id. He got into the car and drove away, alone. Jason asserted that, when he heard the gunshots, he looked out of the window and saw a man running to a car parked outside the apartment building.

The affidavits also both recorded that “a lot of police officers arrived” “[w]ithin five minutes” of the popping sounds. J.A. 439; accord J.A. 441. But the undisputed evidence shows that “a lot of officers” showed up for the first time around 2:40 PM, at which point the victims had been dead for at least an hour and cannot have been shot mere minutes earlier.

Wendy and Jason each testified at the 2021 evidentiary hearing, again giving accounts that differed from Investigator Jones’s 2004 notes and from their 2011 affidavits. Indeed, Juniper’s counsel conceded below that “[a]t th[e evidentiary] hearing,” the Robertses’ testimony did not “establish a different timeframe than the Commonwealth offered” at trial. J.A. 1423.

Wendy testified that the night before the murders, she heard “very loud arguing”

between a man and woman coming from Keshia’s apartment. J.A. 1253. Early on the morning of the murders, she was standing outside when she saw a man “laying on the horn”

48

of his car and screaming that he wanted to see his kids. 25 J.A. 1256. He returned “after 10:30” but “before noon,” went to Keshia’s apartment, came back down, and then yelled up at Keshia’s window that she “better not be [t]here” when he got back. J.A. 1258, 1262, 1264. He departed, then returned a third time “shortly thereafter”; Wendy’s testimony seems to indicate this was before 1:00 PM. J.A. 1264. Wendy saw the man approach Keshia’s apartment, and then heard “four shots, which [she] thought were firecrackers.” J.A. 1261. After hearing the shots, she waited a few minutes to go outside. When she did, she saw the same man running across the lawn toward his car. He asked her what she was “looking at.” J.A. 1262. Later, police arrived. Her testimony did not include any mention of seeing Keshia that morning.

When the Commonwealth asked Wendy about the discrepancies between her statements to police at the time of the murders and her testimony at the hearing, she conceded that the police notes “say something very different” and that “obviously after it happened it would have been fresher in [her] mind.” J.A. 1280. She explained that, nevertheless, “I feel that what I am testifying today is what I remember.” Id.

Wendy also said that Jason was awake for all three of the man’s visits and that she and Jason talked about the man’s comments to Keshia. And she described the man as “thin[]” and “not a big man.” J.A. 1268.

Jason, by contrast, testified at the evidentiary hearing that he “slept in” on the day of the murders and woke up to popping sounds that he “construed . . . to be gunshots.” J.A.

25

There is no dispute that Nykia and Shearyia’s father was in prison at the time.

49

1209. He ran to the window and saw a “large,” “stocky” man run from Keshia’s apartment building to a car. J.A. 1209, 1214. He then went outside to check on his mother. He explained that this all happened “around the time elementary school [got] out,” which Wendy had indicated in her 2011 affidavit was at 2:30. J.A. 1209. He testified that the police arrived “within the next hour or two.” J.A. 1217.

When asked about Investigator Jones’s note that he had not been home earlier that day, Jason testified that he was in fact home all day. He explained that he did not remember telling police that he was not there, but that he did tell police multiple times about the man he saw running away—which was not recorded in any police notes in the record before us. He also stated that he did not remember telling police that he heard hammering, but rather that he had told them that he heard popping noises. (The parties stipulated that Jason “told Investigator Jones . . . that he thought the bangs sounded like hammering.” J.A. 931.) When confronted with his affidavit, which said the police arrived within five minutes rather than hours, he said the affidavit was correct.

After hearing the Robertses’ 2021 live testimony and reviewing the notes of their 2004 statements and their 2011 affidavits, the district court found them not to be credible. Juniper, 2021 WL 3722335, at *25. Typically, we review the district court’s factual findings for clear error, see Fed. R. Civ. P. 52(a)(6); United States v. Heyer, 740 F.3d 284, 292 (4th Cir. 2014), and “give ‘even greater deference to the trial court’s findings’” related to witness credibility, Heyer, 740 F.3d at 292 (quoting Anderson v. City of Bessemer City, 470 U.S. 564, 575 (1985)); accord Teleguz v. Zook, 806 F.3d 803, 811 (4th Cir. 2015) (“Credibility determinations are ‘deserving of the highest degree of appellate deference.’”

50

(quoting Evergreen Int’l, S.A. v. Norfolk Dredging Co., 531 F.3d 302, 308 (4th Cir. 2008))).

Of course, in the Brady context, evaluating credibility often requires a temporal mismatch. The question is what effect the suppressed evidence would have had at the time of trial, which means the credibility of witnesses must be evaluated for how the jury would have perceived them at that time. See Ellis, 121 F.3d at 918; accord U.S. ex rel. Hampton v. Leibach, 347 F.3d 219, 255 (7th Cir. 2003) (noting, in the ineffective-assistance-of- counsel context, that while “[y]ears of drug abuse may well have taken their toll on [a witness]’s faculties” by the time of an evidentiary hearing, “the district court had to look backward and consider what effect [his] testimony might have had” at the time of trial, when he was “not yet a heroin user”).

Nevertheless, credibility is usually best evaluated by the witness’s testimony on the stand. E.g., Williams v. Ryan, 623 F.3d 1258, 1266 (9th Cir. 2010) (rejecting district court’s conclusion “on the basis of written statements alone that [certain witnesses] were inherently unbelievable,” and concluding that the court should have held “an in-court evidentiary hearing in order to determine whether [the witnesses] would have been able to provide material evidence that may have changed the result” of the trial); cf. Juniper, 876 F.3d at 572 (noting that the district court would need to evaluate the Robertses’ credibility during the evidentiary hearing on remand). And in Brady cases, the unfortunate fact is that such testimony might not be heard until many years later, thanks to the government’s suppression of relevant evidence.

To be sure, there may be instances where credibility is simply not able to be evaluated live, and must instead rest on the cold record—for example, if the witness has

51

had a catastrophic medical incident in the interim between the event witnessed and the time of the evidentiary hearing, affecting their memory or ability to express themselves. The unfortunate reality of the passage of time is that some witnesses who would have been found credible at the time of trial may no longer be, just as some witnesses may have died.

But absent any indication that a witness’s credibility has significantly shifted over time, a district court may rely on their live testimony in making its determination of how credible the jury would have found them at the time of trial. This evaluation could include an assessment of the witness’s demeanor, their willingness (or lack thereof) to admit that they no longer remember the facts in question, and, to some degree, the consistency of their story across time. After all, a witness who confidently testifies to a story twenty years after the fact that differs from the story they told originally may not be evincing solely an expected lapse in memory; a district court might rightfully conclude, for example, that the witness is the sort of person who confidently makes statements that the witness believes to be accurate, but which are actually false—and that the witness would have similarly lacked credibility at the time of trial.

Here, Juniper argues that the district court impermissibly discredited the Robertses’

statements because of “natural memory lapses” that occurred over the seventeen years between the murders and the evidentiary hearing. Opening Br. at 60; accord Oral Arg. at 6:11–19 (arguing that the district court improperly “conflated memory with credibility”), available at https://www.ca4.uscourts.gov/OAarchive/mp3/21-0009-20230307.mp3. Instead, Juniper argues, the district court should have considered the alleged consistency of the Robertses’ testimonies in their statements “on the day of the murders and the day

52

after” and “in their state habeas affidavits seven years later.” 26 Opening Br. at 61. Of course, “[w]hen we remanded this matter for an evidentiary hearing—at [Juniper]’s express request—we made plain that the district court could, and indeed, might need to, make credibility determinations” based on their live testimony. Teleguz, 806 F.3d at 811; see Juniper, 876 F.3d at 571–72.

Further, the district court’s credibility determination did not rely on “natural memory lapses.” In fact, as Juniper himself acknowledges, the court explicitly recognized that details fade from memory over time. See Opening Br. at 61–62. And nothing in the court’s analysis indicates that it held it against Wendy or Jason when they were forthright that they did not remember certain details. See J.A. 1218, 1247, 1264 (district court reassuring Wendy and Jason that their inability to recall a detail was all right).

Rather, during the evidentiary hearing, the district court noted to counsel that Wendy and Jason were “not exactly witnesses who inspire confidence in what they say” and had a “limited” ability to testify accurately. J.A. 1422, 1433. The court further noted that it “really didn’t find [Wendy and Jason] to be the most credible witnesses.” J.A. 1423.

26

Even if we were to accept Juniper’s invitation to focus on the Robertses’

testimonies in their contemporaneous statements in 2004 and the state habeas affidavits in 2011, rather than on their live 2021 testimony, we note several inconsistencies between those statements that cast doubt on their credibility. Most significantly, Wendy mentioned in 2004—but not 2011—that she saw Keshia arguing with a man around 12:30 PM on the day of the murders. And Wendy and Jason claimed in 2011—but not in 2004, according to the investigator’s notes—that they saw a man fleeing from the direction of Keshia’s apartment immediately after hearing popping noises, and just minutes before a swarm of police descended on the scene. This would place the shooting around 2:30 PM—which contradicts both Wendy’s 2004 statement that it occurred around 1:30 PM and the forensic evidence indicating that the victims had been killed no later than 1:40 PM (or at the very least, far enough before 2:40 PM that blood had time to dry on Keshia’s shirt).

53

In its written opinion, the court found Wendy and Jason “sincere, but not credible.” Juniper, 2021 WL 3722335, at *25. The court concluded from “[w]atching them testify . . . that they just did not know what they were talking about.” Id. And it found, based on both their “shifting stories” and their “demeanor in testifying,” that “no reasonable juror would have believed their timeline of events over the coherent and consistent timeline and accounts Rashid, Murray, and Mings gave.” Id. While “shifting stories” could be the result of natural memory lapses, in context, we understand the district court to be referring to the kind of witnesses Wendy and Jason were—that is, those who would make strong assertions seven and then seventeen years after the fact that contradicted their earlier accounts.

The district court’s adverse credibility finding is supported by the record. To take one example, when Wendy was asked where she moved after living next door to Keshia in Norfolk, she said that she “came to New York.” J.A. 1295. When asked the follow-up question of whether she “live[d] anywhere else in Virginia before” moving to New York, she asserted, “No. I did not.” Id. When pressed whether she had ever lived in Newport News, Virginia, however, she said she had—and then explained she lived there between Norfolk and New York. While this error could conceivably be explained as a lapse in memory—albeit a significant one—it is suggestive of a lack of credibility, that is, a failure to ensure the accuracy of one’s statements.

Similarly, when Wendy was pressed at the evidentiary hearing about the significant discrepancies between her contemporaneous statements to police and her 2021 hearing testimony, she acknowledged that the statements were “very different,” and stated in explanation only that she “[felt] that what [she was] testifying [at the hearing] [was] what

54

[she] remember[ed].” J.A. 1280. In other words, even while conceding that the events “would have been fresher in [her] mind” on the days immediately surrounding the murders, Wendy asserted that on the day of the hearing, she remembered events very differently. Id. That could be interpreted as a sign of candor (albeit candor only brought out on cross- examination). But it could just as easily be understood as demonstrating a tendency to provide firm testimony regarding events that Wendy had no reason to suppose she was remembering accurately—and in fact, where she was confronted with evidence that her memory was not accurate. And that is suggestive of a lack of credibility as a witness.

Additionally, there is evidence that both Wendy and Jason had medical or mental limitations at the time of the murders that could well have affected their abilities to testify. And there is no indication that those limitations worsened in the intervening time before their testimony at the evidentiary hearing such that their testimony at the hearing was somehow less reliable than it would have been at the time of trial.

Accordingly, we see no clear error in the district court’s credibility determination, and must defer to it here. We thus give little weight to Wendy and Jason’s statements in analyzing the Brady claims before us.

ii.

The statements of two other neighbors who spoke to police on the day of the murders are also relevant. Both neighbors told police that they saw men whose descriptions do not match Juniper near Keshia’s apartment on the morning of the murders.

Chaunte Hodge, who lived in Keshia’s building, told police that, between 10:30 and

55

11:00 AM, she saw a man get out of the passenger side of a green car 27 containing two other men and walk toward the stairs to Keshia’s apartment. About five minutes later, she saw him get back into the car and drive away.

Patrick Danso lived in a nearby apartment building, the parking lot for which was “directly next to the staircase leading to [Keshia’s] apartment.” J.A. 963. Danso told police that around 11:30 AM, a man approached him and asked Danso if he was waiting for someone. Danso was afraid of the man. The man then walked “between [Keshia’s] apartment and the fence behind it” and returned a few minutes later with another man. Id. The second man told the first to “[j]ust forget about it[,] it ain’t worth it.” J.A. 964.

But there is no further corroboration of these accounts in the record. There is no indication of who these men were. And, as the district court found, even if the jury believed these accounts to be accurate, it “does not necessarily undermine the prosecution’s timeline. Each witness said only that they saw someone go towards Keshia’s apartment; not that they saw someone go into Keshia’s apartment.” Juniper, 2021 WL 3722335, at *26. Nor does this testimony directly undermine the varied evidence against Juniper in this case, including the consistent timelines from the prosecution witnesses and the forensic evidence. As such, these sightings can have little weight in the cumulative-materiality analysis.

We recognize that it is not Juniper’s fault that he was unable to follow these leads at the time of the investigation, because the Commonwealth suppressed them. But we can

27

Rashid’s car was brown.

56

only evaluate the record before us. And these accounts demonstrate only that Juniper could have argued to the jury that other viable suspects existed and that the police’s failure to investigate these leads shows shoddy investigative work. In light of the record as a whole, there is little reason to suppose the jury would have found such arguments persuasive. 28 iii.

John Jones, Juniper’s cousin who did not testify at trial, made several statements to police in the aftermath of the murders. Some of his statements conflict with the testimony of Rashid, Murray, or Mings, albeit largely in minor ways, such as the color of Rashid’s car and Juniper’s gun and what type of jacket Juniper was wearing.

More significantly, Jones’s story shifted over time, and in ways that Juniper could

28

Juniper contends that testimony from Rashid’s 2019 deposition points to other individuals having been in the apartment between Rashid’s visit and the murders. Specifically, in 2019, Rashid stated that she saw an ashtray on the floor of Keshia’s living room. But the only ashtray in the crime scene photos is the one on the bed, which could suggest that the apartment had a single ashtray that was moved from one room to the other after Rashid saw it on the floor. Further, Juniper contends that “Rashid testified at her deposition that the ashtray was empty.” Opening Br. at 65. Because the ashtray on the bed contained several butts with DNA from Keshia and other contributors who were not Juniper, Juniper contends that this means that other individuals were in the apartment after Rashid left. And that, he argues, means that her testimony of hearing the gunshots that killed the victims cannot be accurate.

There are several problems with this argument. First, Rashid’s testimony on this point was not suppressed—it didn’t come up until 2019—and there is no reason to suppose that, had the defense been aware of any of the suppressed evidence in this case, that would have prompted them to ask Rashid at trial about whether she saw an ashtray at the apartment. Second, this deposition testimony comes fifteen years after the murders, and therefore is of questionable reliability, particularly given that it pertains to such a minor point. And finally, Juniper significantly overstates the testimony. When asked whether the ashtray in the living room contained any cigarette butts, Rashid actually stated ambivalently: “I don’t know. I don’t know. . . . I think—let me see. I don’t think I saw cigarette butts. I don’t think I saw anything in the ashtray. I don’t know.” J.A. 886.

57

have used—had he been aware of the statements at the time of trial—to argue that the changes were in response to police pressure.

“Soon after the murders, police wanted to question” Jones, but were “[u]nable to find him.” Juniper, 529 F. Supp. 3d at 494. So, on January 20, “the police informed ‘news stations’ that Jones was wanted for questioning about the Stephens family murders, as well as for unrelated outstanding warrants for his arrest.” Id. at 494–95. Jones “turned himself in that evening,” and Investigators Ford and Brian Wray “questioned him soon after.” Id. at 495. During that conversation, Jones stated that he had seen Juniper on the day of the murders when Juniper came to Jones’s apartment sometime between noon and 2:00 PM. But “Jones maintained that . . . he had not been to Keshia’s apartment that day and . . . had no personal knowledge of anything to do with the murders.” Id. When interviewed again on January 22, “Jones continued to deny going to [Keshia’s apartment] and said he didn’t have anything to say.” J.A. 408.

Later that night, the jail where Jones was detained on unrelated charges recorded telephone calls between Jones and his sister, mother, girlfriend, and girlfriend’s mother. On the calls, Jones’s girlfriend reported to him that Investigator Ford had threatened her with problems she and Jones could face in the media, and charges Jones could face, if Jones did not provide the information he had about the murders. Indeed, the parties stipulated that, at some point during the investigation and before these recorded telephone calls, “Investigator Ford made multiple comments to Jones’s girlfriend . . . about problems she and Jones could be facing,” and that these comments “motivated” Jones’s girlfriend and his other family members “to press Jones into cooperating with Ford in developing

58

evidence implicating Juniper” in the murders. J.A. 955.

Consistent with this stipulation, the calls reflect that the four women begged Jones to tell investigators what he knew. They argued that it was obvious that Juniper did not care about him if he was going to let Jones take the fall for the murders, since at the time, Jones’s face was the only one appearing in the media associated with the crimes. 29 The pressure apparently worked; immediately after the phone calls from his family, Jones called Investigator Ford and stated he wanted to speak to investigators and “tell the truth.” J.A. 408.

Investigators brought Jones back to the station that same night, where Jones conceded that a woman had driven him and his girlfriend’s baby to Keshia’s apartment on the date of the murders, that he saw from the apartment steps “that the door was busted in,” and that Juniper emerged from the apartment holding a gun and with cocaine on his face. J.A. 409. He explained that he had previously denied going to Keshia’s apartment because he “didn’t want to be a part of it.” J.A. 792. However, he affirmatively omitted Murray from his account, stating that nobody else had joined him, the unnamed woman, and his girlfriend’s baby on the drive to Keshia’s apartment. And the parties stipulated that Investigator Ford showed Jones a picture of Keshia’s broken-in door during the interrogation.

On February 5—two days after they took Rashid’s revised statement that included Murray—investigators brought Jones in yet again. “In accordance with his practice,

Juniper was not named in the media until the next day, January 23. Juniper, 2021

29

WL 3722335, at *12.

59

Investigator Wray either showed Rashid’s February 3, 2004, statement to Jones or just held it up and told Jones that he had a statement from the driver saying there was someone else in the car.” J.A. 953. Confronted with this information, Jones admitted that Murray was also in the car. He said he had not previously mentioned Murray because he “didn’t feel nobody else should be involved.” J.A. 873.

Had Juniper been aware of Jones’s shifting statements to police and of the recordings of calls between Jones, his girlfriend, and other family members, he could have opted to call Jones as a witness, or sought to introduce some of these statements in another way. This would have allowed him to further question whether Murray was involved— given that, as the jury heard, Rashid had also omitted Murray at first—and to argue that police had used inappropriate pressure tactics to obtain evidence against Juniper.

But calling Jones or introducing his statements would have been a very risky move for the defense. Overall, Jones’s statements to the police were consistent with the other witnesses’ accounts of the day in question and damning for Juniper. He reported seeing Juniper exiting Keshia’s apartment holding a gun, and told police that “when he saw on the news about the homicide [sic], . . . he knew that [Juniper] had done it.” J.A. 411.

As the district court put it, “[h]ad Jones testified, his testimony would largely have reinforced the prosecution’s case. The jury would have heard the account of yet another person who placed Juniper in Keshia’s apartment on the day of the murders, with the door broken in, and holding a gun.” Juniper, 529 F. Supp. 3d at 497; accord J.A. 1460 (district court stating at the evidentiary hearing that “to put evidence on about the Jones claim[,] the rest of the stuff Jones presented comes in, and that is fatal to Mr. Juniper”); Juniper, 2013

60

WL 1333513, at *20 (“Far from altering the outcome of the case, Jones’ testimony would have likely compounded the evidence of Juniper’s guilt, for that would have made four different individuals—Rashid, Murray, Mings, and now Jones—who personally saw Juniper at the crime scene with a gun.”).

Juniper argues that, “[p]rovided concealed evidence about prior inconsistent statements and police practices, . . . jurors reasonably could have viewed at least some of these consistencies as a hallmark of police coaching and coordination of witnesses.” Opening Br. at 52. Maybe. But Juniper’s counsel would have been taking quite a risk to introduce Jones’s statements in the hope that the limited evidence of police coaching they suggest would overcome the damning evidence Jones provided.

Indeed, even Jones’s first statement—when he denied having gone to Keshia’s apartment, and instead indicated merely that Juniper had come to his apartment after visiting Keshia—provided several incriminating details. He stated that Juniper was wearing “a holster” when he saw him that day; that he “was acting real nervous” and “like the police were coming”; that even though Juniper “didn’t tell him that he assaulted” Keshia, “he always does”; and that Juniper’s mother told Jones that Juniper committed the murders. J.A. 401–02, 406–07.

We also note that, because of Juniper’s choices below, we are limited in our ability to interpret how much pressure Investigator Ford actually put on Jones’s girlfriend. Juniper opted not to call Investigator Ford during the evidentiary hearing, and failed to serve Jones’s girlfriend in order to call her, even after the court granted him extra time to do so.

61

(Jones himself died in 2008.)

iv.

Juniper also contends that he could have used Brady material—Mings and Bowser’s statements to police—to impeach Mings’s testimony at trial in a number of ways. As a refresher, Mings was an acquaintance of Juniper’s and was good friends with Murray. He lived with his girlfriend, Bowser, about a block from Keshia, who had worked with Bowser.

First, Mings testified at trial that he went to Keshia’s apartment because of a call from Murray and to pick up a check for Bowser. But on January 16, 2004, Mings told police he went to Keshia’s apartment because Bowser told him to let Keshia know that she was about to come over—he made no mention of Murray calling him. For her part, Bowser told police that she sent Mings to Keshia’s apartment after calling and getting no answer. It was not until May 12 that Mings mentioned a call from Murray. But even then, he said he went to Keshia’s apartment because Murray told him that Juniper and Keshia were “fussing”—he did not say, as he did at trial, that Murray spoke of shots fired. J.A. 471, 483. Had Juniper been aware of these earlier statements, he could have used them to impeach Mings’s testimony. However, the impeachment value would have been limited because the jury already had reasons to doubt Mings’s story: the call from Murray to Mings was otherwise unsupported in the record—Murray did not testify to the call, nor did the prosecution introduce call records pointing to such a call—and Mings conceded at trial that he had another reason for going to the apartment, namely, to get Bowser’s check.

Second, on January 16, Mings told police that he had gone to Keshia’s apartment

62

once; saw the door was kicked in and left; and then returned, at which point he entered the apartment and discovered the two victims on the bed. Bowser’s January 16 statement to police indicated she accompanied Mings on the second trip, though she stayed outside. Undated notes from the prosecutor seem to suggest that Bowser’s brother came with Mings on his first trip. On May 12—consistent with his testimony at trial—Mings reported only one visit, by himself, in which he saw the two victims on the bed and was told by Juniper that Keshia was on the floor between the bed and dresser, but did not see her himself. Again, Juniper could have used these shifting statements to impeach Mings’s credibility. But because he already provided many reasons for the jury to question Mings’s credibility, we doubt these relatively minor discrepancies would have made any difference.

Third, at trial, Mings did not provide times for events and said he was not paying attention to the clock on the date of the murders. But in various statements to police, Mings did provide specific times. Had Juniper sought to impeach Mings’s testimony with this discrepancy, however, he would have faced a double-edged sword. The timeline Mings provided in May—when he suggested that he first went to Keshia’s apartment around 10:00 AM—did not match up with the prosecution’s theory at trial, but the one he provided on the day of the murders did.

Fourth, while Mings did not testify to having seen anyone near Keshia’s apartment, in a January 16 statement he indicated that he saw a man sitting in a silver vehicle. He stated that he asked the man if he had seen anyone go up to Keshia’s apartment, and the man said he had not. Mings was not asked at trial whether he saw anyone outside Keshia’s apartment, so the impeachment value of this alleged discrepancy would have been

63

negligible. At most, Juniper could have used it to argue that this was another lead the police failed to pursue. But nothing in Mings’s statement suggested the man had anything to do with Keshia or was doing anything suspicious, so such an attack would have been of no help to Juniper’s case.

Fifth, on January 16, Mings affirmatively told police that he did not observe anyone alive at Keshia’s apartment, or any weapons or drugs:

Q: Did you observe anybody else [other than the victims] inside [Keshia’s apartment]?

A: No, sir.

Q: Did you observe any kind of weapons in the [apartment]?

A: No, sir.

Q: Did you observe any kind of narcotics, drug paraphernalia, or anything like that inside the [apartment]?

A: No, sir.

J.A. 448. He also omitted any mention of seeing Juniper leave the apartment later. In his May 12 statement and at trial, by contrast, Mings claimed he saw Juniper in the apartment—holding a gun and with a white substance on his face—and that later he saw Juniper leave the apartment.

During questioning by the Commonwealth at trial, Mings did not state that he affirmatively denied having seen anyone in the apartment in his January 16 statement to police, but he did concede that he omitted Juniper from his account:

Q: When you talked to the police verbally on January 16 . . . , did you tell them at that time that you had seen . . . Juniper[] inside [the apartment]?

A: No.

...

Q: Later did there come a point in time when the police came to talk to you again

64

about this?

A: Yes.

Q: Was it in the next week or had some time passed?

A: Yeah, a couple months.

Q: And at that time did you indicate the information . . . that you told the jury today about seeing the defendant in there?

A: Yes.

J.A. 258–59. Accordingly, the jury was at least partially aware of the discrepancy between Mings’s January 16 statement and his trial testimony.

Juniper argues that there is a meaningful distinction between an omission and an affirmative statement. But he does not explain why that distinction matters in this case. In context, we see little reason the jury would have given weight to this distinction—the difference between the omission and the affirmative statement could easily be chalked up to the specific questions asked. That is, neither side asked Mings at trial whether he had affirmatively denied seeing Juniper on January 16. And in any event, it appears obvious to us that the salient point is the one that was brought out at trial: that Mings did not tell police about having seen Juniper in the apartment until months after he was first interviewed.

Finally, at trial, defense counsel elicited that the judge who had sentenced Mings for a recent probation violation knew of his cooperation in this case. But evidence obtained during post-remand discovery casts further light on the sequence of events: Bowser told the prosecutor about Mings’s issues with probation; Mings got a favorable probation- violation sentence; and then Mings implicated Juniper. Further, Mings reached out to a judge on the eve of Juniper’s trial to request leniency in his ongoing court troubles, noting that he was “assi[s]ting the [C]ommonwealth in a very big case.” J.A. 311. This evidence would have provided additional avenues for impeachment of Mings’s testimony, but again,

65

the value of such impeachment is reduced by the fact that the jury knew Mings could benefit from his testimony.

Mings died in 2020, before the 2021 evidentiary hearing. Juniper’s counsel opted not to call Bowser at the hearing, despite having properly served her.

v.

Post-remand discovery also uncovered various information related to Murray that Juniper argues he could have used to (1) suggest coaching by police and (2) impeach Murray.

First, investigators initially interviewed Murray on February 6 after learning from Rashid and Jones on February 3 and 5, respectively, that Murray had been in the car with them. In his taped statement, Murray—who had stayed at Juniper’s house the night before the murders—reported that Rogers (Juniper’s mother) and Rashid woke him up on the morning of the murders, telling him that Rashid had heard Juniper kick in Keshia’s door and then heard gunshots. He also recounted a story akin to his testimony at trial regarding the trip with Rashid to Keshia’s apartment. Those statements, made just weeks after the murders and generally aligning with those of other witnesses, are not helpful to Juniper’s case.

But evidence developed post-remand could be used to argue that investigators influenced these harmful statements. Investigators’ notes from the February 6 interview indicate that Murray “at first denied having any knowledge of what happened” regarding the murders and said he did not pick Juniper up from Keshia’s apartment. J.A. 436–37. After investigators “advised [Murray] that he was not telling the truth” and “showed” him

66

the contradictory evidence, however, he agreed and conceded he was in the car with Rashid and Jones. J.A. 437, 954. The parties’ stipulated facts also note that “Investigator Wray told Murray what other witnesses had said” and that “investigators showed Murray photographs of alleged suspects and evidence” before he gave a recorded statement. J.A. 989.

It is hard to gauge the weight the jury would have placed on this witness-influence argument, however, because Juniper’s counsel did not take the opportunity at the evidentiary hearing to question Murray about it, despite the fact that Murray testified. So all we know is that investigators provided Murray with information about other witnesses’ statements and the evidence in the case, after which Murray ceased denying that he had knowledge about the murders and instead implicated Juniper. That is consistent with coaching, but it is also consistent with a witness changing his tune once he realized the police knew more than he initially thought.

Juniper’s strongest argument for the former interpretation is that Murray provided information to police in multiple homicide investigations. Investigator Conway’s 2019 deposition testimony indicates that, around the time of the murder investigation, he considered Murray a “jailhouse snitch.” J.A. 822. Another investigator, John Kowalski, testified at the evidentiary hearing that Murray was “very well-known” to the police force and provided him with information on several cases in the early 2000s in exchange for assistance on Murray’s own cases. J.A. 1349. The parties stipulated that investigators “knew that Murray had provided information to police in cases prior to Juniper’s case, including in a homicide investigation.” J.A. 992. (Murray, however, claimed at the

67

evidentiary hearing that before Juniper’s trial, he had never provided information to the police department in another case.)

So, armed with this evidence, Juniper could have argued to the jury that police shaped the testimony of Murray, a known informant, in a mutually beneficial arrangement to gain damning evidence against Juniper by offering Murray assistance with his ongoing court troubles. For reasons we outline in detail below, we do not believe that would have been a persuasive argument. Further, some evidence developed on remand cuts against Juniper’s contention on this point: Officer Kowalski noted that he found Murray’s information “valuable” and “viewed [it] as being accurate,” and that he had “made it very clear to [Murray] early on when he began providing information” that “once he cried wolf one time,” Kowalski “would never be able to trust anything that [Murray] told [him] again.” J.A. 1351. In other words, there is evidence that Murray was a valuable informant because his information was accurate, and that Murray had incentives to continue telling the truth.

Second, Juniper could have used the suppressed evidence to impeach Murray’s trial testimony by pointing to discrepancies between his statements to police and his testimony at trial. Several of these details were relatively minor and therefore would have provided minimal impeachment value. For example, in Murray’s February 6 taped statement, he stated that he stayed in the car and only “hopped out” when Juniper exited the apartment, at which point he got into the back seat of the car and let Juniper sit in the front. J.A. 456. At trial, by contrast, Murray said he got out of the car when Jones did and walked to the back of the car. Moreover, at trial, Murray said Juniper’s gun was automatic; in his

68

statement, he said it was semiautomatic. The February 6 taped statement also included some details about the drive from Keshia’s apartment that he did not include at trial, albeit details that did not look good for Juniper, such as that Juniper was holding his “gun in the air” and saying he was “not going to jail” while “looking all crazy.” J.A. 457. Those discrepancies would provide no help to Juniper because, even if he used them to impeach Murray’s credibility, he would do so at the cost of bringing in additional near- contemporaneous incriminating details.

More significantly, Murray’s February 6 statement did not mention a phone call from Juniper in which Juniper confessed to him. Instead, Murray said that Rashid mentioned gunshots when she came back to Rogers’s house after leaving Juniper at Keshia’s apartment. On June 15, police interviewed Murray in the local jail’s booking office regarding a different pending case. When they asked him about Juniper, Murray made statements that contradicted aspects of both his February 6 statement and his trial testimony. For the first time, he described the phone call from Juniper in which Juniper confessed. But he added that Juniper gave instructions about his funeral “because he was not going to be taken alive” and stated that Juniper’s mother spoke to Juniper before handing the phone to Murray. J.A. 899. Neither detail was included at trial.

There may well have been impeachment value to noting that Murray did not provide his account of the phone-call confession until June, especially when combined with the fact that Murray was an informant who knew he stood to gain from providing information. Nevertheless, it is not obvious that Juniper would have opted to introduce this evidence. After all, the confession narrative was already questionable based on the evidence at trial

69

given Murray’s credibility issues and that Rashid did not testify that Murray spoke on the phone with Juniper. Further, even though his February 6 statement omitted the confession, it included a different narrative implicating Juniper: that Rashid told Murray about the gunshots when she returned from Keshia’s apartment.

Finally, Murray’s trial testimony also forms the basis of one of Juniper’s remaining Napue claims. Juniper contends that Murray’s implication at trial that he had no personal motive for testifying against Juniper was misleading. The parties have stipulated that, in fact, when Murray spoke with police on February 6, he “wanted assistance from the police regarding his [probation] violation in exchange for information”; that when Murray spoke with police on June 15, he asked what the police “could do with his probation violation”; and that he only said “that Juniper had confessed to him” after asking for help with his probation violation (albeit, help that the investigator told Murray he could not promise would be forthcoming). J.A. 988, 990–91. We explain below why this is insufficient to support a Napue claim in light of the other trial evidence.

vi.

Juniper also points to Brady evidence that, “[i]n addition to Smith, . . . investigators met with at least three other inmates who also claimed that Juniper had confessed to them or that they were present when Juniper stated his intention to commit the murders.” J.A. 1004. He argues that this evidence “demonstrate[s] there was a culture of snitching at the jail and inmates”—including Smith—“easily could concoct stories to obtain a benefit.” Opening Br. at 43.

The parties stipulated that “[t]he prosecution was aware that jail inmates were trying

70

to provide false information against Juniper in hopes of obtaining a benefit.” J.A. 1005. None of the three inmates involved testified at trial, which counsel for the Commonwealth represented at the evidentiary hearing was because “the Commonwealth didn’t deem them credible.” J.A. 1159.

One of the inmates in question provided a 2019 affidavit and testimony at the 2021 evidentiary hearing to the effect that in 2004, he had sent a letter to the Commonwealth’s attorney seeking to provide false information implicating Juniper in order to benefit himself, and that many inmates at the local jail wanted to do the same. Juniper failed to serve the other two inmates to appear at the evidentiary hearing, even after the court recessed for a week to give him a chance to do so.

In our view, the weight of this evidence is minimal. Just because other inmates also sought to implicate Juniper—and assuming that all three of them were doing so falsely— does not mean Smith was lying. In any event, the jury could assess Smith’s credibility themselves, and already had plenty of reason to disbelieve him: his prior felonies; the fact he was an inmate snitching about another inmate’s alleged jailhouse confession; the fact that he knew Keshia; the “inherently incredible” story—in Juniper’s counsel’s words, Opening Br. at 46—of Juniper confessing to a quadruple murder, including of two children, to someone he barely knew, over a casual game of chess; and the fact that Smith stood to gain from his testimony.

vii.

Finally, Juniper’s other remaining Napue claim rests on testimony given by Officer

71

Atkinson 30—the officer who responded to the 12:44 PM 911 call, went up the wrong stairs to an apartment belonging to Hodge (not Keshia), and spoke to Hodge and Frazier. He testified at trial that when he spoke with Hodge, he “had no idea” that the building had another upstairs apartment, that he did not “remember asking [Hodge] directly if there was a second apartment upstairs,” and that he did not know that a second stairwell led up to Keshia’s apartment “until hours later.” J.A. 74, 76, 78. This testimony was consistent with a two-page handwritten report Atkinson produced after responding to the 911 call, in which “he wrote that he ‘never saw the stairs which [led] to the apartment where it turned out the homicide had occurred.’” Juniper, 529 F. Supp. 3d at 510 (alteration in original).

However, a statement Hodge gave around 5:30 PM on the day of the murders partially contradicted Atkinson’s account. Id. “Much, but not all, of her description matched Officer Atkinson’s report.” Id. But Hodge also “said that Officer Atkinson ‘asked [her] if there was another apartment upstairs,’ and she ‘told him around on the other side.’” Id. (alteration in original) (emphasis omitted).

The parties stipulated that, “[i]n the course of responding to the 12:44 [PM] 911 call on January 16, 2004, Officer Atkinson learned that there were two upstairs apartments” at the building, and that he “learned the location of the apartment that was the potential[] subject of the 911 call and where its entrance was located.” J.A. 999, 1001. They also stipulated that “Officer Atkinson’s handwritten report provides evidence that despite being

30

Juniper’s opening brief does not raise the Atkinson Napue claim at all, and therefore he has waived it. Grayson O Co. v. Agadir Int’l LLC, 856 F.3d 307, 316 (4th Cir. 2017). However, given the cumulative nature of the analysis, and for completeness, we nevertheless address it.

72

made aware of the location of the apartment that was the subject of the 911 call, he concluded no further investigation was necessary.” J.A. 1003. And they stipulated that “[t]here is no evidence in the record that Officer Atkinson and Officer Liverman reported they saw the door of the victims’ apartment.” Id.

Atkinson and Hodge testified at the evidentiary hearing. Atkinson stated that, at the time he responded to the 911 call, he believed “that the entire upstairs was one apartment.” J.A. 1183. The district court found that he “presented as a credible witness.” Juniper, 2021 WL 3722335, at *26 n.40. However, it is well established that parties are bound by their stipulations, except in “certain limited conditions” that the Commonwealth has not argued apply here. Sprint Nextel Corp. v. Wireless Buybacks Holdings, LLC, 938 F.3d 113, 123 (4th Cir. 2019); accord Christian Legal Soc’y Chapter of the Univ. of Ca., Hastings Coll. of the L. v. Martinez, 561 U.S. 661, 676–77 (2010).

Hodge could not remember speaking to any officers outside her apartment on January 16, even after reading her statement from that day. Nor could she remember the content of the statement she gave to the police later that day at the police station, though she testified that she answered the police’s questions truthfully.

We explain below why the Atkinson evidence is of limited value and insufficient to support a Napue claim.

C.

Several other pieces of evidence are relevant, even though they were not presented at trial and do not constitute Brady or Napue material. They include (1) testimony from Frances Frazier, who lived in the apartment below Keshia; (2) testimony from Kevin

73

Waterman, another neighbor of Keshia’s; (3) evidence tending to impeach Ernest Smith; (4) the call logs from two 911 calls on the day of the murders; and (5) certain stipulations entered into by the parties below. This evidence is not Brady or Napue material because it was known by Juniper at the time of trial, not raised by Juniper as a possible claim, or else barred by the mandate rule. Nevertheless, as explained above, we may consider this evidence in the cumulative-materiality analysis to the extent it may have affected the “preparation or presentation of [Juniper]’s case.” Bagley, 473 U.S. at 683.

First, Frazier, who lived in the apartment below Keshia’s, told the police around 1:00 PM that she had been home all day and had heard people moving furniture in the morning, but had not heard any gunshots or any other “unusual noises.” Juniper, 2013 WL 1333513, at *25; see Juniper, 2021 WL 3722335, at *16 n.30. These specific statements were not presented at trial, although the jury did hear Atkinson’s testimony that he asked Frazier whether she had heard gunshots and that, “as a result of that discussion[,] [he] did not believe there was any further reason to investigate” the apartment building. J.A. 80. The jury could therefore infer that Frazier told Atkinson that she had not heard any gunshots.

Further, the parties stipulated that Frazier told officers around 3:00 PM that she had still not heard gunshots as of that time, even though the shootings had undisputedly taken place by then. Because there were certainly gunshots in the apartment above Frazier’s by 3:00 PM, Frazier’s statement that she had not heard any as of that time undermines the value of her statement that she had not heard any as of 1:00 PM, either. To be sure, the parties also stipulated that the record contains no evidence about Frazier’s whereabouts

74

between 1:00 and 3:00 PM that day. Still, it is worth asking why officers would bother to note that Frazier told them at 3:00 that she had not heard gunshots during the preceding two hours if she had, in fact, been absent from the premises during that time.

The Supreme Court of Virginia rejected a Brady claim based on Frazier’s statements, Juniper, 707 S.E.2d at 299, and the district court did the same, finding it barred by the mandate rule, 31 Juniper, 529 F. Supp. 3d at 485–86. Frazier passed away years ago, so she was not available to testify at the evidentiary hearing.

Nevertheless, we may consider whether Frazier’s evidence would have impacted the preparation or presentation of Juniper’s defense. It is possible that, had Juniper known of Frazier’s statements to police, he would have followed up with her to clarify whether she was home between 1:00 and 3:00 PM. 32 We simply do not know what she would have told him, though again, we think it would be odd for her to affirmatively tell officers on the day of the murders that she had not heard any gunshots during that two-hour window if she was not home to hear anything. It is also possible Juniper would have chosen to call Frazier as a witness. But, depending on the results of his (potential) investigation of her,

31

Juniper argues that the district court erred in applying the mandate rule, but fails to develop that argument. We need not wade into whether Juniper has waived the argument, see Grayson O Co., 856 F.3d at 316, or whether the mandate rule applies (as the Commonwealth urges), because we conclude that we may consider the evidence for the purpose of evaluating its impact on the preparation or presentation of Juniper’s defense, and our conclusion would be no different if we considered the evidence more broadly.

32

It is not obvious that he would have done so in a timely enough manner to get clear answers from Frazier. Juniper’s investigator, Kennedy, did not speak to Waterman until December 31, 2004—nearly a year after the murders. Nor did he attempt to speak to Hodge until that same date. There is no reason to suppose that he would have treated the evidence from Frazier any differently, and it is impossible to know what Frazier would have remembered nearly a year after the fact.

75

that may well have been risky: if she had testified that she was at home between 1:00 and 3:00 PM, then the force of her testimony that she had not heard shots before 1:00 would have been diminished, because again, a number of shots undisputedly were fired in her apartment building before 3:00 that day (seemingly by 1:40, and certainly before 2:40). It is not obvious that Juniper would have taken that risk when he could already point jurors to the fact that Atkinson had concluded, after speaking with Frazier, that the 911 call was a false report. So we think the impact of Frazier’s evidence on the preparation or presentation of Juniper’s defense would have been minimal.

Second, Kevin Waterman lived on the block adjacent to Keshia’s. On the day of the murders, he told police that, between 1:30 and 2:00 PM, he heard several “loud shots sounding similar to a nail gun”—specifically, “three quick shots, and then one shortly thereafter.” J.A. 391. In a taped statement later that afternoon, Waterman stated that he heard four “loud pops” at 1:32 PM, which he thought was “hammering or a nail gun” from construction that had been taking place at “the same apartment building.” J.A. 798. At the 2021 evidentiary hearing, he explained that he heard “three to four loud pops” between 1:30 and 2:00 PM but “didn’t think anything of it because there was construction going on . . . on the corner,” so he “figured it was like an air gun.” J.A. 1190. But when pressed, he indicated that he was certain what he heard was gunshots.

Waterman’s statements to police are not Brady material because Juniper was undisputedly aware of them at the time of trial. His counsel opted not to call Waterman as a witness because no other evidence of which they were aware suggested that the shots that killed the victims occurred so late in the day. Juniper, 2013 WL 1333513, at *30 (citing

76

Juniper, 707 S.E.2d at 304); see J.A. 312–13 (Juniper’s investigator’s notes about meeting with Waterman); J.A. 339 (Waterman affidavit noting he spoke to Juniper’s investigator before trial); J.A. 380 (defense counsel’s affidavit explaining why they did not call Waterman). Of course, that is because the Commonwealth did not provide Juniper with the Robertses’ statements. Waterman’s statements to police accorded with part of Wendy’s 2004 statement: that she heard three or four bangs around 1:30 PM.

Juniper would therefore certainly have viewed Waterman’s statements in a different light in preparing and presenting his defense had he been aware of the Roberts evidence: Waterman’s statements meant that another neighbor had heard four gunshot-like sounds around 1:30 PM. He may have even chosen to call Wendy, Jason, and Waterman to make this point at trial. But Waterman’s testimony would not have changed Wendy and Jason’s basic credibility as witnesses—in rendering its adverse credibility determination against Wendy and Jason, the district court was certainly aware that their initial statements were partially corroborated by Waterman’s.

Instead, Waterman’s testimony would only have provided support that Wendy’s account of hearing three or four bangs around 1:30 was accurate. But the Commonwealth could have easily undermined the import of that testimony: none of the three earwitnesses (Wendy, Jason, or Waterman) identified the sounds as gunshots in their reports to police— indeed, Waterman and Jason in his very first statement to the police both identified the sounds as construction-related (hammering or a nail gun); according to Waterman, there was in fact construction going on nearby that could explain the sounds; each earwitness spoke of only hearing at most four shots, when there were undisputedly at least eight shots

77

fired; and a 911 call was placed at 12:44 PM, reporting gunshots at Keshia’s apartment, which as we discuss below is difficult to explain if the shooting occurred at 1:30.

To be sure, Juniper could have retorted that some of the same critiques could be made of Rashid’s testimony: she heard only four “booms,” which she “wasn’t sure . . . were gunshots at first.” J.A. 130–31. But there is a ready explanation for why Rashid only heard four of the shots—she was driving away and may have only heard the first few shots before she was out of range to hear the remainder. The same cannot be said of Wendy, Jason, and Waterman, who heard the noises from in or just outside their apartments. And, of course, Rashid’s testimony fit in with the substantial other evidence at trial in a way the other three earwitnesses’ did not. To accept that the shootings occurred at 1:30, the jury would have had to discount all the other eyewitness testimony; Juniper’s DNA on the knife and the cigarette butt found on the door shards; and the 12:44 PM 911 call. Accordingly, had Juniper opted to call Wendy, Jason, and Waterman at trial, the impact of their testimony—while certainly providing some evidence for an alternative timeline—would not have been substantial.

Third is evidence concerning Ernest Smith. Evidence produced during discovery suggests promises from the prosecution to Smith to seek a sentencing benefit for Smith in exchange for his testimony, which he later received. This evidence could have been used to impeach Smith. The district court did not consider a claim based on this evidence,

78

finding it precluded by the mandate rule. 33 Juniper, 529 F. Supp. 3d at 485. But again, we may consider how this evidence would have impacted the preparation or presentation of Juniper’s defense. In our view, that impact would have been quite minimal; Juniper could have sought to impeach Smith a bit more than he did, but he was already able to impeach his testimony in numerous ways.

Additionally, Smith testified at the evidentiary hearing that Keshia was a “[f]riend of the family,” that he first met her in 1997 or 1998, and that he “[s]omewhat” thought of her as a sister. J.A. 1379. He said that he saw her “[o]ften” and had babysat her oldest child (who was not one of the victims in this case). J.A. 1380. The prosecution was aware that Smith had known Keshia for years and had babysat her oldest child. But at trial, Smith said merely that he knew Keshia; he was not asked to elaborate on their relationship. Nevertheless, the defense could have cross-examined him on this point to elucidate how well he knew Keshia. Indeed, the defense was not caught off guard by Smith’s admission at trial that he knew Keshia; the parties stipulated that defense counsel’s “handwritten notes of an interview with Juniper” two months before trial “state[d] that Smith knew the victims.” J.A. 1005.

Further, Smith provided details about his relationship with Keshia in a 2007 affidavit, well before the Supreme Court of Virginia issued its 2011 final ruling on Juniper’s habeas petition. That creates a potential statute-of-limitations issue for

33

As with the Frazier evidence, Juniper passingly argues that this evidence should not have been barred by the mandate rule. The Commonwealth disagrees. But, again, we need not address this issue because we consider the evidence regardless.

79

considering any claim based on this evidence. Juniper, 529 F. Supp. 3d at 513 n.91. Nevertheless, for completeness, we consider whether this evidence would have impacted the preparation or presentation of Juniper’s defense. But because the defense was aware before trial that Smith knew Keshia, and could have followed up on that point, we believe any such impact to be negligible.

Fourth, the Joint Appendix includes the call logs from the two 911 calls discussed at trial. Juniper has consistently alleged that these logs were not provided to his trial counsel, though he has never raised a Brady claim on that basis in federal court 34—perhaps because they largely support the prosecution’s narrative. In any event, the district court admitted the logs into evidence in the proceedings below, so we review them here.

The first log states that a female caller phoned at 12:44 PM to report that “about an hour ago she heard a gunshot” in the upstairs apartment, and that she was “worried about her neighbor” because “she usually hears noise coming from the [apartment] and she has not heard anything.” J.A. 333 (capitalization altered). The caller described Keshia’s apartment as “upstair[s] in the corner near the bushes” and stated that she did not “want to

34

In its 2011 opinion, the Supreme Court of Virginia noted that Juniper “assert[ed]

that the 911 telephone call log would have shown that the first call took place at a pay phone about a half-mile away from the apartment instead of in Bowser’s apartment, as Mings had testified.” Juniper, 707 S.E.2d at 298. But it rejected a Brady claim based on the call log because the record “demonstrate[d] that Mings told . . . Kennedy, [Juniper]’s investigator, that he and Bowser walked to a nearby convenience store and used a pay phone to make the initial call.” Id. at 299. Juniper has not pursued that claim in federal court. Additionally, the district court rejected an ineffective-assistance-of-counsel claim based on counsel’s failure to elicit the inconsistency between the call log and Mings’s testimony, which “would have been relevant only to whether Mings was credible.” Juniper, 2013 WL 1333513, at *35. Juniper did not appeal that conclusion.

80

be involved.” Id. That is consistent with an anonymous call from Bowser trying to get police to check on Keshia’s apartment in light of what Mings had reported to her after his visit to the apartment, and is also consistent with Mings having learned about the shootings sometime shortly after 11:34 AM (when Juniper allegedly confessed to Murray, who then allegedly called Mings and told him there had been gunshots fired at Keshia’s apartment)— i.e., “about an hour” before the 12:44 PM 911 call. Consistent with the officers’ testimony at trial, the log indicates that the officers were dispatched at 12:45, arrived at the scene between 12:46 and 12:50, and left at 1:12 PM. The call log also indicates that the officers reported to dispatch that Frazier said she “did not hear any gun shots.” J.A. 333.

The second 911 log records that Bowser called from her apartment at 2:12 PM and stated that “her [friend’s] upstairs [apartment] door [was] busted in” and that while “they could hear a TV going in [the] background,” she was not sure where her friend was. J.A. 335 (capitalization altered). This time, Bowser provided her own name and address, and also provided Keshia’s name. The log indicates that Bowser stated that she would “be waiting outside to show” the police where to go, which accords with Mings’s trial testimony that he and Bowser approached the police when they saw them return to the apartment. Id. According to the log, police were dispatched between 2:27 and 2:30, and the fire department was also dispatched at 2:32 PM. At 2:30, the log records that “on call back [the caller] said she thinks the man inside is dead.” Id. It is not clear whether this means Bowser called 911 back, or vice versa, 35 but it supports that Bowser knew there was at least

35

In Mings’s May 12, 2004, taped statement to investigators, he stated that “the

81

one deceased victim in the apartment. The log notes that police reported four possible victims inside the apartment at 2:37 PM, and that numerous officers responded to the scene shortly thereafter.

Finally, the parties stipulated that Murray and Juniper were at Juniper’s mother’s house the night before the murders and that, according to Juniper’s trial counsel’s handwritten notes, Juniper told counsel that he and Rashid went to Keshia’s apartment on the morning of the murders. They also stipulated that “Officer Atkinson concluded based on his investigation that there had been false 911 calls made about [Keshia’s] building in the past.” J.A. 1002.

IV.

The legal principles and numerous facts having been set out, we turn to an application of principles to facts. Our analysis proceeds in the two steps we established above: first, applying the Napue standard, we evaluate the cumulative materiality of the Napue evidence and any Brady evidence tending to show the falsity of the testimony at issue in the Napue claims; then, applying the Brady standard, we evaluate the cumulative materiality of all of the Napue and Brady evidence. We conclude that Juniper is not entitled to relief at either stage of this analysis.

A.

At the first step, we determine whether there is “any reasonable likelihood” that the

second time” he and Bowser called police “they just took a long time to come, so [he and Bowser] called them a third time.” J.A. 477. This is arguably consistent with the “call back” at 2:30—perhaps Bowser called at 2:12 PM, and then, when the police had not shown up by 2:30, called again to report a deceased victim in order to spur the police into action.

82

Napue material, combined with any Brady material showing its falsity, “could have affected the judgment of the jury.” Agurs, 427 U.S. at 103.

Here, Juniper’s two remaining Napue claims that survived summary judgment are that (1) “Murray’s testimony implying that he had no personal motive for testifying against Juniper” was “misleading,” and (2) “Officer Atkinson’s testimony that he did not know the existence or location of Keshia’s apartment when he responded to the 12:44 [PM] 911 call” was “false.” Juniper, 2021 WL 3722335, at *27. For purposes of this analysis, we assume the testimony in question was false or misleading and that the prosecution knew it to be so. Additionally, some Brady evidence supports the falsity or misleading nature of this testimony—namely, Murray’s statements to police demonstrating that he asked for help with his probation violation in exchange for testimony and only raised Juniper’s supposed confession to him after asking for that help, and Hodge’s statement to police on the day of the murders that she told Atkinson there was a second upstairs apartment.

In our view, however, there is no reasonable likelihood that these two pieces of false and misleading testimony, combined, could have affected the jury’s judgment.

If there was any evidence to suggest that Officer Atkinson actually investigated Keshia’s apartment around 1:00 PM and found the door to be intact at that time, that would be highly significant—it would obliterate the Commonwealth’s timeline. But there is not. At most, Hodge’s statement and the parties’ stipulations support that Atkinson learned from Hodge that there was a second upstairs apartment (the entrance of which could not be seen from Hodge’s apartment). Had the jury known that, they may well have viewed Atkinson as having conducted shoddy policework by not checking out the other upstairs apartment,

83

and then having lied about his awareness of its existence—perhaps out of embarrassment that it turned out a quadruple homicide had occurred in that apartment.

On the other hand, the jury may well have viewed Atkinson’s decision not to conduct further investigation as reasonable, given that his discussions with two women who lived in the relatively small, five-apartment building led him to believe nothing was amiss, and that there had been false 911 calls about the building in the past. Either way, Atkinson lying under oath with the prosecution’s knowledge would have undermined his credibility and that of the prosecution, and may have led the jury to view the police and prosecution more negatively. Yet Atkinson was, on the whole, a minor witness.

Evidence undermining Murray’s credibility would have been more important, as he was a more central witness. Yet the jury already had multiple compelling reasons to question Murray’s credibility: his initial claim on the stand that he did not remember the day of the murders at all; his attempt to “plead the Fifth”; the lack of testimony from Rashid about a supposed phone call from Juniper to Murray in which Juniper confessed; Murray’s prior felony convictions, including one for which he was serving an active sentence at the time of trial; Murray’s delay of several weeks before he spoke to police at all; and the fact that Murray was given immunity for potential misdemeanors related to the crime. Additional impeachment evidence would have moved the needle only slightly. See Bowman, 45 F.4th at 755 (concluding that a suppressed document’s “additional impeachment value would be slight” where the witness in question’s “memory of the events had already been impeached through testimony” at trial).

Combining the Napue claims, this evidence would have allowed Juniper to call into

84

question the policework in this case, albeit only of one officer, who was not one of the investigators; plant some doubt about the timeline, to the extent he could try to hint that perhaps Officer Atkinson did see that Keshia’s door was intact at 1:00 PM—which is dubious, at best, given it was impossible to see Keshia’s door from Hodge’s or Frazier’s apartments and the parties stipulated that there is no record evidence that Officer Atkinson reported that he saw the door; and question the veracity of Murray, who provided testimony harmful to Juniper.

But there is no reasonable likelihood that this evidence could have affected the judgment of the jury. It in no way undermines the significant forensic evidence against Juniper, Smith’s testimony of hearing a confession, or the independent testimony of Rashid and Mings. And it only slightly undermines the credibility of the police and Murray. No reasonable juror who voted to convict based on the evidence presented at trial would change their vote based on the Napue evidence presented in this case.

B.

That means that, to obtain relief, Juniper must be able to show that all of the Brady and Napue evidence is cumulatively material under the Brady standard—that is, that there is “a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different.” Kyles, 514 U.S. at 433–34. Again, we conclude the answer is no.

There is no doubt that there is a lot of evidence that we are assuming is favorable to Juniper and the Commonwealth suppressed. As the district court put it, “[m]easured in words—or pages, or even pounds—Juniper’s Brady claims could seem significant”

85

because “[h]e has identified numerous categories of evidence that could have been used to impeach parts of the prosecution’s key witnesses’ testimony.” Juniper, 529 F. Supp. 3d at 515. Nevertheless, the court ultimately concluded that, “[d]espite the large quantity of evidence Juniper points to, the vast majority of it has little quality. The potential impeachment evidence he identifies nibbles around the edges of the prosecution’s case, but it never takes a bite out of its center.” Id.

We agree. In our view, there is no reasonable probability that this evidence would have changed the jury’s verdict, in light of the very significant evidence presented at trial that the Brady and Napue material does not undermine. At bottom, “[t]his was not a thin or circumstantial case,” Bowman, 45 F.4th at 758, or one where the suppressed evidence, “[v]iewed cumulatively, . . . completely undermines the crux of the [Commonwealth]’s closing argument,” Long v. Hooks, 972 F.3d 442, 464 (4th Cir. 2020) (en banc), as amended (Aug. 26, 2020); accord Bowman, 45 F.4th at 757 (concluding that Brady evidence was not cumulatively material where it arguably undermined the testimony of two witnesses, but did not otherwise undermine significant other evidence against the petitioner). 36 That is, even once we add the weight of any new exculpatory evidence to the

36

To be sure, the suppressed evidence would partially undermine a part of the Commonwealth’s closing argument—that there was no evidence the witnesses coordinated their overlapping statements. Juniper could have used the suppressed evidence to argue that there was coordination, such as in the alleged coordinated insertion of Murray into the narrative. But that would have been, at best, an argument that the jury would have been free to reject. And the suppressed evidence does nothing to undermine the other aspects of the prosecution’s closing argument, such as the forensic evidence. The evidence thus does not “completely undermine[] the crux” of the prosecution’s closing argument. Long, 972 F.3d at 464.

86

defense side and subtract the weight of any new impeachment evidence from the prosecution’s side, there is not a reasonable probability that the jury would have reached a different result. Nor does the favorable evidence “put the whole case in such a different light as to undermine confidence in the verdict.” Kyles, 514 U.S. at 435. The prosecution’s case would have been weaker had it disclosed the evidence in question, but, here, we are confident that the result would have been the same.

i.

We begin by reviewing each category of Brady and Napue evidence and our conclusions about the weight it would have carried at trial. The categories of evidence include (1) the evidence from Wendy and Jason Roberts and other neighbors; (2) the impeachment evidence of Jones, Mings, and Murray; (3) evidence casting doubt on the investigation into Juniper; and (4) evidence that other inmates, besides Ernest Smith, attempted to implicate Juniper.

First and foremost is the evidence from Wendy and Jason Roberts, which formed Juniper’s central Brady claim, on the force of which he obtained an evidentiary hearing. At that hearing, the district court heard live testimony from Wendy and Jason, as well as from neighbors Hodge and Waterman, and at which it considered prior written statements from or about those witnesses, as well as neighbors Frazier and Danso. Each of these witnesses provided reason to doubt the neat timeline the Commonwealth put forward at trial. Frazier said she had heard no gunshots before 1:00 PM; Wendy, Jason, and Waterman said they heard gunshots around 1:30 PM; Hodge and Danso told police they saw other people approaching Keshia’s apartment on the morning of the murders; in her original statement,

87

Wendy told police that she saw Keshia alive around 12:30 or 12:45, arguing with a man other than Juniper; and in later statements and at the evidentiary hearing, Wendy and Jason said they saw a man fleeing Keshia’s apartment just after hearing the gunshots. But the district court concluded that Wendy and Jason were not credible witnesses—a determination to which we defer for the reasons discussed above. And we have explained above why the evidence from Hodge, Frazier, Danso, and Waterman would not have carried significant weight.

The second category of evidence to consider is the impeachment evidence of Jones, Mings, and Murray. We do not believe the jury would have placed much weight on most of this impeachment evidence; much went to small details, about which it is not surprising that witnesses differed from each other or over time. Further, Mings and Murray were extensively impeached at trial (Jones did not testify)—including via inconsistencies in their stories, their prior felony convictions, and the possible benefits they could receive by testifying against Juniper—and we are confident that even with only the evidence presented at trial, the jury took their testimony with a large grain of salt.

Still, portions of the impeachment evidence related to these three men could have been significant at trial. First is the fact that Rashid, Jones, and Murray all initially denied Murray’s involvement. This came out at trial only as to Rashid. But the jury may have found it significant that Jones also omitted Murray from his initial narrative, and that Murray, too, originally denied involvement.

Yet to get the evidence from Jones in, as noted, Juniper would have risked having the rest of Jones’s testimony put forward at trial—testimony that, on the whole, was

88

harmful to him. As for the fact that Murray himself initially denied his involvement, we doubt that would have swayed the jury at all; after all, the jury had seen Murray make such denials with their own eyes when he first took the stand at trial.

Additionally, the ultimate significance of the fact that three people initially omitted Murray is far from clear. On the one hand, at first blush, it seems like quite a coincidence for the Commonwealth to have to explain away. But on the other, it may not be so.

Rashid’s omission of Murray was discussed at trial. She first mentioned Murray to police on February 3. Investigators then confronted Jones with Rashid’s statement, and he conceded that Murray was present. And they then spoke to Murray, pointing to the evidence from both Rashid and Jones. Each individual may have simply been waiting to see what the police actually knew before adding more details. Cf. J.A. 873 (Jones stating that he did not mention Murray at first because he “didn’t feel nobody else should be involved”).

Juniper’s argument, of course, is that all of this suggests coaching by police. That is, armed with the statements Jones and Murray made to police, combined with Rashid’s statement discussed at trial, Juniper could have argued to the jury that police planted the idea of Murray’s involvement in each of these witnesses—presumably because he was a known snitch and they thought they could use him to implicate Juniper. But why?

The police already had testimony from a much stronger witness—Rashid—

alongside evidence from Jones and Mings. Combined, those three witnesses provided a coherent timeline and placed Juniper at the scene with a gun. The only testimony that Murray added was the alleged confession—but he did not provide any evidence about that

89

until June 16, after asking for help with his probation violation. In other words, if anything, the evidence supports that Murray sought to manipulate the police by offering juicier testimony in exchange for help with his charges—not that police pressured three different witnesses to add Murray into their narratives, even though he was otherwise completely uninvolved, in the hopes that they could get Murray to provide especially damning evidence against Juniper—evidence that Murray did not provide until months after other witnesses described his involvement.

That brings us, however, to the evidence that both Mings and Murray changed their tales months after the murders to implicate Juniper after seeking or obtaining assistance with their probation violations. For Mings, this meant describing seeing Juniper at the apartment; for Murray, it meant describing the phone call in which Juniper confessed to him. The jury knew that Mings had omitted Juniper initially, but not that he had affirmatively denied that anyone was at the apartment. And the jury knew that Murray had delayed in speaking to police, but did not know that it was months after his first contact with police that he first mentioned the phone call.

Certainly, this evidence would have impeached both Mings and Murray in the eyes of the jury. But, as we have noted, both were already highly imperfect witnesses for the prosecution. And, even supposing this evidence led the jury to doubt their testimony even more—and to reject that Mings saw Juniper at the apartment or that Juniper confessed to Murray—it would not have directly undermined those aspects of Mings’s and Murray’s testimony that aligned with Rashid’s or with the phone records.

The third category of evidence to consider involves whether the Brady and Napue

90

material would have allowed Juniper to cast wide-ranging aspersions on the competence or good faith of the investigation against him. After all, he could have used the Jones evidence to argue that Investigator Ford pressured Jones to give evidence against Juniper and influenced his statements by showing him a picture of Keshia’s broken-in door (albeit, at the risk of also introducing the evidence from Jones favorable to the prosecution). And he could have used the Murray evidence to show that Investigator Wray told Murray about evidence from other witnesses and even showed him photographs of evidence, and thus to argue that the investigators coached Murray, a known snitch. Moreover, he could have used Officer Atkinson’s failure to investigate the second upstairs apartment and detectives’ failure to follow up with Wendy and the other eyewitnesses who saw men other than Juniper in the vicinity of Keshia’s apartment that morning to depict a sloppy or myopic investigation.

There are at least three problems with these arguments. First, the evidence of coaching, such as it is, goes only to Murray’s testimony (again, Jones did not testify), not to Rashid’s, Mings’s, or Smith’s. While a jury persuaded that Jones and Murray were coached could reasonably infer that police similarly influenced other witness statements, they would not necessarily reach that conclusion. They might find it significant that the record lacks evidence of such influence as to those witnesses. They might also find it notable that Rashid was accompanied by a lawyer during both of her police interrogations.

Second, while the parties have stipulated that police showed certain evidence to Murray and Jones and made “comments” to Jones’s girlfriend that “motivated” her “to press Jones into cooperating,” J.A. 955, that does not mean police shaped those witnesses’

91

entire narratives or that the comments were as drastic as Jones’s girlfriend made them sound in her out-of-court recorded statements to Jones. And despite calling Murray at the evidentiary hearing, Juniper’s counsel did not ask Murray about this alleged coaching. While he properly served Rashid; the lawyer who accompanied her during her police interviews; and Investigators Ford, Wray, and Conway, he did not call any of them at the evidentiary hearing. And he failed to properly serve Jones’s girlfriend in order to call her. In other words, Juniper had the opportunity to develop a claim of police incompetence or malfeasance, but he did not do so. He cannot deliberately opt not to develop evidence and then seek to exploit any resulting ambiguity. Cf. Teleguz, 806 F.3d at 811 (declining to allow evidence from witnesses who refused to testify at evidentiary hearing to undermine district court’s determination that witnesses who did testify were credible).

Third, the jury may well have viewed with distaste an attempt to paint the investigators in a bad light based on the limited information noted herein, which does not undermine some of the most compelling evidence against Juniper. In other words, this tactic could easily have backfired, making it appear that the defense was grasping at straws in an attempt to build a conspiracy when the evidence solidly pointed to Juniper as the culprit. In sum, we think the evidence of “coaching” could have weakened the force of Murray’s testimony, but we doubt the jury would have placed much weight on it in the grand scheme of things.

The fourth, and last piece of Brady evidence is that several other inmates, not just Smith, attempted to implicate Juniper. As noted, we find that evidence to carry little weight given the significant impeachment evidence already available regarding Smith. Finally, for

92

the reasons discussed above, we place little weight on the Napue evidence at issue here.

All of this being said, our analysis should not be read in a vacuum. Each piece of evidence discussed herein could, in another case, be highly significant. Rather, our analysis here is necessarily influenced by the “certain basic facts about the murders” that cannot reasonably be “unsettle[d]” by the Brady and Napue evidence Juniper has put forward. Juniper, 2013 WL 1333513, at *17. It is to those facts that we now turn.

ii.

The Brady evaluation is not a sufficiency-of-the-evidence test. Kyles, 514 U.S. at 434–35. But “we do not ignore other evidence presented at trial in determining our confidence in the outcome. Instead, we evaluate the whole case, taking into account the effect that the suppressed evidence, had it been disclosed, would have had on the evidence considered at trial.” Ellis, 121 F.3d at 918 (footnote omitted). So we must consider the evidence that remains undisturbed by the Brady and Napue evidence discussed above.

Rashid testified to having been an eye- and earwitness to the moments just before the murders, as well as when the shootings began. Her testimony is supported by objective facts: the front door was kicked in, as she testified to having heard happen, and the two children were naked and one had ink on her face, as she also noted. Moreover, the fact that the children were still undressed, and the ink was still present, at the time of their deaths supports that they were killed shortly after Rashid saw them: they had undressed for their bath, but had not yet been bathed to remove the ink or redressed.

Rashid’s testimony also provided evidence of motive and opportunity for Juniper to commit the crimes. She placed Juniper at Keshia’s apartment and described an increasingly

93

hostile scene in which Juniper came to retrieve a DVD player, began arguing with Keshia and accusing her of being unfaithful, kicked in the front door, and then shot Keshia and her family. Rashid’s testimony is supported by that of another witness who described Juniper as jealous and abusive toward Keshia. And when the district court asked during the evidentiary hearing why Rashid would lie, Juniper’s counsel said she was “not able to answer that question.” J.A. 1428–29. Indeed, there is no evidence whatsoever that Rashid was on anything but good terms with Juniper, frequently giving him rides around town or running errands for him, with no apparent reason to fabricate a story implicating him in the murders of four people. Of course, it is conceivable that another, unknown person also had the motive and opportunity to murder Keshia that same day, and happened to come along to do so shortly after Rashid witnessed this confrontation. But there is no evidence to support such a highly coincidental scenario.

While Mings and Murray were heavily impeached at trial—and would have been able to be further impeached with the Brady and Napue evidence—significantly, their testimony aligned with Rashid’s in key ways. See Bowman, 45 F.4th at 758 (“[T]he undisclosed evidence, at best, would have undercut [two witnesses’] reliability in the eyes of the jury. But both men’s testimony was consistent with the other evidence offered at trial.”). All three placed Juniper at Keshia’s apartment carrying a gun. Both Mings and Murray testified to having seen powder on Juniper’s face, and Rashid testified that she heard Jones tell Juniper to wipe his face. All three suggested that, in the aftermath of the murders, Juniper was acting paranoid or nervous. Mings also knew the correct locations of the bodies in the apartment, down to the fact that Keshia’s body was not visible from the

94

doorway of the bedroom. 37 A 911 call was placed at 12:44 PM, reporting gunshots at an upstairs apartment in Keshia’s building about an hour earlier. Keshia’s apartment was one of only two upstairs apartments in the building. This call was either an incredibly coincidentally timed hoax that just happened to identify the proper method, location, and (under the prosecution’s theory) rough timeline of the murders, or it came from someone with knowledge of the crimes (as the Commonwealth maintains). And if, as Juniper argues, the murders took place after the 911 call, it again was either an extremely coincidental hoax, or was placed by someone who knew gunshots would be fired at Keshia’s apartment later in the day. We think a jury would find either scenario implausible.

We suppose it is possible that the jury could conclude that someone deliberately placed the 911 call before the murders in order to muddy the timeline. But that would be an odd choice for the perpetrator, as it could well have led police to Keshia’s door, forewarning her of possible violence; led to a greater police presence in the area on that day; or, perhaps most relevantly, led police directly to the perpetrators when they were at Keshia’s apartment. It certainly seems strange to call police ahead of time and inform them

37

Juniper passingly argues that Mings was a viable alternative suspect for the murders, which would explain his knowledge of the locations of the victims’ bodies. But Juniper’s cited evidence does not support his claim that there is evidence suggesting Mings’s involvement. See United States v. Miller, 41 F.4th 302, 313 (4th Cir. 2022) (noting that, where the opening brief fails to cite the parts of the record on which it relies, the argument is waived). And the district court rejected a similar contention below, for reasons Juniper has not disputed on appeal. See Juniper, 529 F. Supp. 3d at 504 n.67.

95

of where and how you plan to commit a crime.

Additionally, to accept the deliberate-early-call theory, the jury would have to look at the crime scene here—in which Keshia was stabbed, and then she and her three family members were gunned down in a seemingly frenzied spray of bullets that struck them in various body parts—and still believe that the crimes were committed by someone coolheaded enough to plant a phone call beforehand in order to establish an alternative timeline. We think jurors would find the most likely explanation by far to be that the 12:44 PM call was placed by someone with knowledge of the crimes, which had taken place before the call was made.

Mings testified, and the call records show, that Bowser placed a second 911 call at 2:12 PM, reporting that Keshia’s door was broken in. Call records also demonstrate that a call was placed from Keshia’s apartment to Juniper’s mother’s house at 11:34 AM and from Rashid’s home to Juniper’s mother at 1:10 PM. These records align with Rashid’s testimony and, arguably, Murray’s account of the confession he heard from Juniper over the phone.

Then there is the forensic evidence. That evidence placed the knife used to stab Keshia in Juniper’s hand. While it is possible that another individual could have used a knife that Juniper had touched at some point in the past, they would have had to stab Keshia without disturbing Juniper’s fragile fingerprint on the blade, and without leaving any of their own DNA on the knife handle. That is conceivable—another assailant could have worn a glove, and the fingerprint expert stated it was possible Juniper’s thumbprint could have remained on the knife through a later stabbing by someone else. Conceivable, but not

96

likely. Juniper has pointed to no evidence he previously used the knife; it would be quite coincidental for another party to use that particular knife; and the expert explained that fingerprints are very fragile and easily wiped away.

Additionally, although the evidence showed that the stab wound occurred before Keshia was shot, leading to “a great deal of blood,” there is no evidence that Keshia sought to stem the bleeding with anything other than her bare hands or to clean the wound; nor is there evidence that any blood was found outside the bedroom. Juniper, 529 F. Supp. 3d at 477. And there is no evidence that Rueben—the other adult at the apartment—tried to stop an initial, stabbing assailant from fleeing the scene. Common sense suggests, therefore, that Keshia never left the bedroom after she was stabbed, and that she and the other victims were shot very soon thereafter. That, in turn, supports that she was stabbed and shot by the same person, or by multiple people working together. But “if anyone else was involved, [Juniper] himself has yet to say so.” Juniper, 2013 WL 1333513, at *17. And if she was stabbed and shot by the same person, the knife evidence points to Juniper as being that individual.

Finally, forensic evidence also showed that a cigarette found on top of the shards of the front door contained Juniper’s DNA. This evidence lends itself to four possibilities: (1) Juniper left the butt there while he was at the apartment after kicking in the door and committing the murders; (2) he kicked in the door, smoked a cigarette that he dropped on top of the door shards, and left before the murders occurred; (3) he returned to the apartment after the door was kicked in and dropped a cigarette butt there; or (4) a cigarette

97

butt he had previously left at the apartment was kicked on top of the door shards by someone after the door was kicked in.

Nothing in the Brady or Napue evidence suggests it was anything other than the first possibility, and we think a jury would conclude that that is what happened. The second possibility is quite remote: it would require Juniper becoming enraged to the point of kicking in the door, but not committing other violence, and then another person coincidentally coming along shortly thereafter to commit the murders. There is no evidence for the third possibility; Juniper has never suggested that he returned to the apartment after leaving with Rashid. The fourth is, we suppose, theoretically possible, and is what the defense argued in closing at trial and what Juniper argues on appeal; the officer who discovered the bodies explained he did not notice the cigarette butt when he first entered the apartment because he was focused on the possible presence of a perpetrator, meaning he or his partner could have kicked the cigarette butt onto the door shards by accident. That would, however, be yet another highly unlikely coincidence.

Putting this all together, we conclude that Juniper cannot satisfy the Brady standard.

The Brady and Napue evidence noted herein would have weakened certain aspects of the prosecution’s case, but would not have undermined the core of the evidence against Juniper. Accordingly, we conclude that the Brady and Napue evidence cannot “reasonably be taken to put the whole case in such a different light as to undermine confidence in the verdict,” and that there is not “a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different.” Kyles, 514 U.S. at 433–34, 435.

98

V.

Before closing, we pause to note—yet again—the dismal failures of the prosecution in this case.

The federal courts have spent more than a decade “baffled by the Commonwealth’s approach to discovery in this case.” Juniper, 529 F. Supp. 3d at 517. Juniper claims to have first learned of the withheld Roberts evidence by chance—during the 2010 criminal trial of the lead investigator in his case. But instead of immediately producing that evidence once Juniper specifically requested it, the Commonwealth “steadfastly opposed” its disclosure. Id. at 492 (quoting Juniper, 2013 WL 1333513, at *15). And it maintained that position for years, until finally the district court ordered it to produce the evidence in October 2012. Juniper, 2013 WL 1333513, at *12 (citing Memorandum Order, Juniper v. Pearson, No. 3:11-cv-00746 (E.D. Va. Oct. 12, 2012)). Many additional years passed before the Commonwealth ultimately turned over—again, only when ordered by the district court—what amounted to a “mountain of evidence.” Juniper, 529 F. Supp. 3d at 517; see id. at 476 (“With numerous stops and starts—and needing frequent intervention by the Court—the parties conducted discovery from May 2018 until August 2019.”).

Over the years, this Court has reprimanded the Commonwealth’s tactics as “unjustifiabl[e],” Juniper, 876 F.3d at 572, and the district court has condemned the Commonwealth’s apparent “entrenched resistance to transparency” and “long, senseless resistance to the disclosure of its information about the Roberts[es],” Juniper, 2013 WL 1333513, at *12 n.6, *15. We, like the district court, “struggle[] to see how the Commonwealth’s tactics result in the appearance, pursuit, or realization of justice,” even

99

if, at the end of this long road, we have concluded that it did not “technically violate[] Brady” because the evidence does not satisfy materiality. Juniper, 529 F. Supp. 3d at 517. Yet remarkably, Juniper’s case is not the first time this Court has found the Commonwealth’s prosecutorial methods lacking. See Juniper, 876 F.3d at 566 n.7 (collecting cases in which the Commonwealth suppressed exculpatory and impeaching evidence). As we said six years ago, “[w]e find it troubling that, notwithstanding these rebukes, officials in the Commonwealth’s Attorney’s office continue to stake out positions plainly contrary to their obligations under the Constitution.” Id.

To be sure, while Brady establishes a constitutional minimum for prosecutors, it does not encompass the entirety of prosecutorial obligations. See Cone v. Bell, 556 U.S. 449, 470 n.15 (2009) (“Although . . . Brady[] only mandates the disclosure of material evidence, the obligation to disclose evidence favorable to the defense may arise more broadly under a prosecutor’s ethical or statutory obligations.”). Prosecutors have special obligations as representatives “not of an ordinary party to a controversy, but of a sovereignty whose obligation [is] to govern impartially” and whose interest is—or ought to be—in the idea that “justice shall be done” rather than in winning in court. Berger v. United States, 295 U.S. 78, 88 (1935).

This independent ethical duty of disclosure requires more than what due process demands and is not limited by the Brady materiality standard. See ABA Comm. on Ethics & Pro. Resp., Formal Op. 09-454 at 1, 3 (2009). Rather, under the Virginia Rules of Professional Conduct, prosecutors “shall . . . make timely disclosure” to defense counsel of all information known to be favorable to the defendant, “except when disclosure is

100

precluded or modified by order of a court.” Va. Rules of Pro. Conduct 3.8(d). A timely disclosure, made as soon as reasonably practicable, allows the defendant to make effective use of the information. See ABA Formal Op. 09-454, supra, at 6. And indeed, such a disclosure was required by court order in this case as far back as April 2004, prior to trial. J.A. 62 (state trial-court order instructing “the Attorney for the Commonwealth [to] divulge and disclose to counsel for the defendant any evidence known to the Attorney for the Commonwealth which tends to exculpate the defendant”).

The dilatory tactics in this case jeopardize the public’s perception of the fairness and integrity of our criminal justice system, place an outsized burden on the court system, and demonstrate a lack of consideration for the victims’ family in this tragic case. After all, if the evidence against Juniper was as “overwhelming” as the Commonwealth maintains, Response Br. at 1, its “entrenched resistance to transparency” has dragged these habeas proceedings on far longer than they needed to be, Juniper, 2013 WL 1333513, at *15. With a modestly different record—say, one where the Brady evidence significantly undermined Rashid’s credibility and where Wendy and Jason’s testimony was deemed credible—our decision today may well have come out the other way, and would have required putting the victims’ family through a retrial nearly twenty years after these horrific murders.

We have said it before, and we repeat it again now: we will not “‘condone’ [the]

suppression of exculpatory and impeaching evidence by [the] prosecution, notwithstanding that” we have ultimately concluded that “such evidence was not material.” Juniper, 876 F.3d at 566 n.7 (quoting Muhammad v. Kelly, 575 F.3d 359, 370 (4th Cir. 2009)). “As a matter of practice”—and fundamental fairness—“the prosecution should err on the side of

101

disclosure, especially when a defendant is facing the specter of execution”—as Juniper was until July 1, 2021, after court-ordered discovery completed in this case. Muhammad, 575 F.3d at 370.

VI.

For the reasons stated, we affirm the decision of the district court.

AFFIRMED

102

APPENDIX: Crime-Scene Diagram

J.A. 1679.