UNITED STATES v. GREEN (2024)

OPINION OF THE COURT

Few guarantees are more central to our judicial system than that of a fair trial by jury. It is “the most fundamental of all freedoms”—“the great bulwark of [our] civil and political liberties.” Estes v. Texas, 381 U.S. 532, 540, 85 S.Ct. 1628, 14 L.Ed.2d 543 (1965); 2 J. Story, Commentaries on the Constitution of the United States 541 (4th ed. 1873). To that end, Rule 14 of the Federal Rules of Criminal Procedure aims “to promote economy and efficiency and to avoid a multiplicity of trials” only so long as “these objectives can be achieved without substantial prejudice to the right of the defendants to a fair trial.” Bruton v. United States, 391 U.S. 123, 131 n.6, 88 S.Ct. 1620, 20 L.Ed.2d 476 (1968) (internal quotation marks omitted). So when competing narratives call for a jury to convict one defendant in order to acquit another, we require separate juries to avoid substantial prejudice to any single defendant. Cornelius Green was not afforded such a trial. We will therefore reverse the denial of his motion to sever, vacate his conviction and sentence, and remand.

I.

A.

Though this appeal arises from three competing stories, a good deal of information remains clear and undisputed. Green was a member of the Infamous Ryders motorcycle club, along with co-defendant William Murphy and Steven Wong. Another man, Ishmael Snowell, was invited into the club; he declined membership, and allegedly began gossiping about Wong.

Ostensibly to exchange money and sort out their fractious relationship, Wong and Snowell met at the Infamous Ryders clubhouse on July 22, 2020, with Green and Murphy also present. Snowell and Murphy began a fight-club-style altercation with Green refereeing. During the fight, Wong looked at Snowell's phone and discovered photographs of money. Wong also began attacking Snowell and asked where the money was; Snowell eventually claimed that it was at his aunt's house in Reading, Pennsylvania. Wong ordered Snowell into the back of Murphy's car. Murphy got in the driver's seat, Green got in the back with a gun, and they began driving toward Reading to retrieve the money. Wong, who rode his motorcycle alongside the car, remained in contact over the phone. Wong peeled away eventually, but he continued to stay in touch.

At this point, stories begin to diverge. The three men in Murphy's car stopped at a Wawa gas station and Green went in to buy water. Testimony differed about whether Murphy or Snowell could have left the car while Green was inside or whether a motorist stopped and offered Murphy and Snowell help (which they rejected). In any event, neither Murphy nor Snowell made any attempt to flee the car; once Green returned, he took over driving and they continued their journey to Snowell's aunt's house.

According to Snowell, all three men entered his aunt's house upon arrival and unsuccessfully searched for the money Wong had seen in Snowell's photos. Wong reacted angrily over the phone when he learned of the missing money, and Snowell recalls Wong ordering Green to shoot him. Confronted with that threat, Snowell attempted to escape from Green and Murphy. But before he could leave the property entirely, Green engaged him in a fight outside. Snowell recalls Green attempting to put him back in the car, but he resisted long enough for neighbors to call the police. Green and Murphy then left the scene.

Murphy testified at trial, and told a slightly different story. Murphy said that, while stopped at Wawa, he never had possession of Green's gun. Once in Reading, he sat in the car for about five minutes while Green and Snowell entered the house. At that point, he went inside and saw Green brandishing a gun at Snowell. He intervened, pleading for Green to put the gun down. While Murphy argued with Green, Snowell walked out of the house and refused to cooperate with them any longer. And so after a chaotic series of events, Murphy left the scene with Green on Wong's orders before police arrived.

B.

The grand jury indicted Green, Murphy, and Wong for kidnapping Snowell and holding him to commit a robbery in violation of 18 U.S.C § 1201(a)(1). All three defendants pleaded not guilty; Green moved to sever his trial from both Murphy and Wong. The District Court denied the motion without prejudice, reasoning that Green had not borne his burden to justify severance under Rule 8(b) or Rule 14 of the Federal Rules of Criminal Procedure. Wong's trial was nonetheless severed from that of his co-defendants because of a last-minute change in attorneys. Wong was eventually acquitted, in a trial at which Murphy testified.

At the beginning of Green and Murphy's joint trial, each party previewed their respective theories of the case in opening statements to the jury. The Government would seek to prove that Wong had ordered both Green and Murphy to kidnap and rob Snowell; Murphy, who acknowledged there was a kidnapping, would seek to prove that he had been coerced by Wong and Green to commit the kidnapping; and Green would argue that there was no kidnapping.

Concerned by these openings, Green renewed his motion to sever at the end of the first day of trial. He argued that he faced two prosecutions: one by the Government, which alleged he had cooperated in a joint kidnapping effort, and another by Murphy, who alleged that Green had forced him to participate in a kidnapping. Entertaining Green's renewed motion, the District Court questioned Murphy's counsel about whether Murphy would claim that Green had coerced him:

[COUNSEL FOR MURPHY]: My understanding is that there will be more testimony where Ishmael Snowell will say there was a gun in the car and that that gun was handled by [Green], and that's his testimony. But for [Murphy's] purposes, the force to begin the engagement in any robbery or kidnapping came from Mr. Wong, and anything that Mr. Murphy did after that was sheer survival.

THE COURT: So you're not going to advance a defense that the coercion about which you spoke in your opening came from, directly or indirectly, from Mr. Green?

[COUNSEL FOR MURPHY]: No, he's going to testify to what happened, and that it came from Wong, but he is going to testify to what happened at the house.

App. 159 (emphasis added). With Murphy's mixed proffer in mind, the Court “[did]n't see the kind of conflict that would cause [it] to sever” Green's trial. App. 159. But Green's counsel asked for clarification: would Murphy testify that Green was on the phone with Wong while inside Snowell's aunt's house, and that Green had threatened Snowell with a gun? Murphy's counsel deflected, noting that he expected Snowell to testify to the fact “that Wong was on the phone with Green and Green had a gun out.” App. 159. He said nothing about his client's testimony, and the District Court again denied Green's motion without prejudice.

Trial proceeded. The Government called Snowell to testify the next day, among other witnesses. He recalled being driven to his aunt's house by Murphy at Wong's behest, with Green seated alongside him in the backseat armed with a gun. Snowell testified that Wong had ordered “them” to shoot him when the trio could not locate the money pictured on his phone. And he remembered escaping from the house before he could be harmed.

At the end of the Government's case-in-chief, Green renewed his severance motion for a final time. And while Murphy's counsel reiterated that Murphy would testify that Wong was the “major force behind” the kidnapping, counsel was clear that Murphy would also testify about what he saw inside Snowell's house. App. 250. Specifically, he proffered that Murphy's testimony would include “Mr. Green being directed to look for stuff and to, potentially, shoot Mr. Snowell.” App. 250. The District Court again denied the motion. It read at length from our relevant precedent and concluded, “I do not find these [defenses] to be mutually[ ]antagonistic ․ It's clear to me that the jury could do what this particular case requires. They are able to assess the guilt or innocence of each Defendant, on an individual and independent basis.” App. 250–51.

Murphy then testified in his own defense. Consistent with his proffer, he noted that Wong prompted a fistfight between him and Snowell, that Green was present to referee, that Wong pointed a gun at Murphy and told him to drive Green and Snowell to Snowell's aunt's house, and that Murphy eventually saw Green threaten to shoot and kill Snowell. In particular, he recalled stepping between Green and Snowell and pleading with Green to spare the latter's life. Murphy also testified that the Federal Bureau of Investigation and the Bureau of Alcohol, Tobacco, Firearms and Explosives interviewed him as part of an investigation of Green and Wong. He noted that agents said they could help him, “as long as [he] help[ed] [them] take down Green and Wong,” because they “really want[ed] them for other crimes that they've committed.” App. 261, 269.

While the Government used its closing to argue for the conviction of both Murphy and Green, it acknowledged that Murphy's testimony painted Green as fully involved in the kidnapping. Green's closing emphasized that Snowell was dishonest, and that Murphy was lying about coercion. For example, Green pointed to the fact that Murphy and Snowell made no attempt to escape when he left the car and went inside Wawa. Murphy's closing reiterated his coercion theme, and referenced Green several times. Specifically, he alleged that “Green [was] driving,” that “Green ha[d] a gun,” and that Murphy “was forced” to participate in the kidnapping because “they were threatening me.” App. 293 (emphasis added). And the Government further drove that point home on rebuttal: “Mr. Green wants you to believe that the kidnapping didn't occur, [Murphy] wants you to believe that it did occur, and he was justified in doing so. Obviously, both of those theories cannot be true.” App. 294.

The District Court instructed the jury on the elements of kidnapping and provided joint-trial instructions consistent with our decision in United States v. Voigt, 89 F.3d 1050, 1096 (3d Cir. 1996). It took the jury three hours to convict Green and acquit Murphy. The District Court went on to sentence Green to 132 months' (eleven years') imprisonment. His sentence final, Green timely appealed.

II.1

Green's burden is a heavy one. As an initial matter, we review a district court's decision to deny a motion to sever for abuse of discretion. United States v. Boscia, 573 F.2d 827, 832 (3d Cir. 1978). We “determine from the record, as it existed when the motion was made, what trial developments were then reasonably foreseeable, and in that light decide whether the district court abused its discretion in denying the severance motion.” United States v. McGlory, 968 F.2d 309, 340 (3d Cir. 1992). And for us to reach such a conclusion, we must be convinced that the District Court's “decision rests upon a clearly erroneous finding of fact, an errant conclusion of law or an improper application of law to fact.” NLRB v. Frazier, 966 F.2d 812, 815 (3d Cir. 1992) (citation omitted).

Second, Federal Rule of Criminal Procedure 14 allows for the severance of joint trials only “[i]f the joinder of ․ defendants ․ or a consolidation for trial appears to prejudice a defendant.” Our default procedure is thus to conduct a joint trial of defendants who are indicted together, “because joint trials ‘promote efficiency and serve the interests of justice by avoiding ․ inconsistent verdicts.’ ” United States v. Urban, 404 F.3d 754, 775 (3d Cir. 2005) (quoting Zafiro v. United States, 506 U.S. 534, 537, 113 S.Ct. 933, 122 L.Ed.2d 317 (1993)).

But the Supreme Court teaches that severance under Rule 14 should be granted “only if there is a serious risk that a joint trial would compromise a specific trial right of one of the defendants, or prevent the jury from making a reliable judgment about guilt or innocence.” Zafiro, 506 U.S. at 539, 113 S.Ct. 933. The right to a fair trial is one such trial right. See, e.g., United States v. Agurs, 427 U.S. 97, 107, 96 S.Ct. 2392, 49 L.Ed.2d 342 (1976); United States v. Gatto, 995 F.2d 449, 455 (3d Cir. 1993); see also United States v. Lane, 474 U.S. 438, 446 n.8, 106 S.Ct. 725, 88 L.Ed.2d 814 (1986) (misjoinder can “deny a defendant his Fifth Amendment right to a fair trial”). But only where prejudice from the violation of a right is “clear and substantial” does Rule 14 require severance. McGlory, 968 F.2d at 340 (quoting United States v. Eufrasio, 935 F.2d 553, 568 (3d Cir. 1991)).

III.

Given the high standards outlined above, motions to sever predicated on prejudicial joint trials are frequently made and infrequently granted. See Voigt, 89 F.3d at 1094–95; see also United States v. McGill, 815 F.3d 846, 926 (D.C. Cir. 2016) (reasoning that a “conflicting defense” is rarely a ground for reversing an order denying severance); United States v. Shively, 715 F.2d 260, 267 (7th Cir. 1983) (“Severance is argued in almost every case where there are multiple defendants, and appellate courts give the argument short shrift, regarding it as a matter within the discretion of the trial judge.”). But the mere rarity of the argument's success does not eviscerate the Fifth Amendment's guarantee of a fair trial. The risk of an unfair trial may occur in three scenarios: (1) “when evidence that the jury should not consider against a defendant and that would not be admissible if a defendant were tried alone is admitted against a codefendant;” Zafiro, 506 U.S. at 539, 113 S.Ct. 933; (2) when “essential exculpatory evidence that would be available to a defendant tried alone were unavailable in a joint trial;” id., and (3) when defendants assert mutually antagonistic defenses, Voigt, 89 F.3d at 1094. With reference to the third category, we have long held that a mutually antagonistic defense exists when “acquittal of one codefendant would necessarily call for the conviction of the other.” Id. But before making that assessment, we address several specific considerations relevant to our analysis.

First, Voigt does not stand for the proposition that defenses are mutually antagonistic only when they are true mirror images of each other (that is, where Defendant A may say that Defendant B committed a crime alone, and Defendant B may say that Defendant A committed the crime alone). Rather, we emphasize that a defense must “call for” the conviction of the other. See id. A defendant is not required to show that a jury would be left with no other option but to convict him if a jury believes his co-defendant. After all, a jury may always solve that problem by issuing inconsistent verdicts between two defendants in a joint trial. Harris v. Rivera, 454 U.S. 339, 345, 102 S.Ct. 460, 70 L.Ed.2d 530 (1981). So we pay special attention to what a given defense “calls for” from the jury.

In the same vein, some language from our precedents suggests that judicial economy may justify joint trials so long as a jury can compartmentalize evidence against particular defendants, even where prejudice exists. McGlory, 968 F.2d at 341; see also United States v. Jackson, 649 F.2d 967, 973 (3d Cir. 1981). And we do not discount that concern; severance of joint trials will often result in the expense of more judicial time and resources. But we have observed a steady decline of criminal trials in our circuit, as in others. See generally Robert J. Conrad, Jr. & Katy L. Clements, The Vanishing Criminal Jury Trial: From Trial Judges to Sentencing Judges, 86 Geo. Wash. L. Rev. 99 (2018) (examining decline of federal jury trials and corresponding contributing factors). In fact, over the last twenty years the number of criminal jury trials has fallen by nearly fifty percent across the federal system—from 2,751 trials in the year preceding March 2004 to just 1,526 trials in the year preceding March 2024. United States Courts, Table D-4–U.S. District Courts–Criminal Federal Judicial Caseload Statistics (Mar. 31, 2004) (2004 Data), https://perma.cc/8LJ5-G9RQ; United States Courts, Table D-4–U.S. District Courts–Criminal Federal Judicial Caseload Statistics (Mar. 31, 2024) (2024 Data), https://perma.cc/39DE-663S. And now, just one out of every fifty federal criminal defendants will proceed to a jury trial. 2024 Data, https://perma.cc/39DE-663S.

To be sure, it has ever been the case that no defendant should “be deprived of a fair trial because it is easier or more economical.” Boscia, 573 F.2d at 833. Our judicial system does not aim to resolve cases as quickly or inexpensively as possible, to the detriment of a criminal defendant.² To be sure, certain severed trials might require more judicial time and energy to adjudicate than would a single joint trial. But even that eventuality is not assured; in some cases, severed trials might actually conserve court resources. See Zafiro, 506 U.S. at 544–45, 113 S.Ct. 933 (Stevens, J., concurring) (declining to endorse “unqualified” preference for joint trials, and reasoning that trying certain multidefendant cases separately is “not only more reliable, but also more efficient and manageable than some of the mammoth conspiracy cases which the Government often elects to prosecute”).

So, it is true that judicial economy remains a relevant consideration for courts in determining whether to sever the joint trials of defendants who raise allegedly antagonistic defenses. But to the extent that our earlier decisions rest on the notion that there is a “substantial” public interest in economically resolving cases in joint trials, we are not convinced that concerns of judicial economy should weigh as heavily in our severance analysis as they have in prior decades.

IV.

Taking account of our standards for discerning mutually antagonistic defenses and for assessing judicial economy, we begin by considering whether Green has demonstrated that his trial was improperly joined to Murphy's in light of mutually antagonistic defenses. We next review several evidentiary byproducts of the joint trial, and whether they speak to the propriety of joinder. For the interdependent reasons set forth below, we conclude that Green's joinder was improper.

A.

We start with Green and Murphy's alleged mutually antagonistic defenses. While Murphy had yet to testify before the District Court denied each of Green's motions to sever, it was nonetheless aware of the outlines of Murphy's testimony following several proffers from his attorney. While those proffers involved Murphy placing the heft of the blame on Wong for coercing him to participate in a kidnapping scheme, they also clearly indicated Murphy would testify that a kidnapping did occur and that Green was coercive and violent on the day of the alleged kidnapping.

Our review of these proffers is cabined by the mutually antagonistic defense doctrine and its corresponding abuse-of-discretion standard. We may review two categories of information: that which was actually known to the District Court at the time it considered a motion to sever, and that which was reasonably foreseeable. McGlory, 968 F.2d at 340. So, a district court's decision not to sever—even if a complete trial record reveals incompatible defenses—is no abuse of discretion if the District Court does not know and cannot foresee the mutually antagonistic nature of the parties' arguments. Id.

Given that context, assessing future testimony is inherently difficult. A lawyer's proffer of a client's expected testimony, by its very nature, cannot resolve all ambiguities about what the witness will actually say on the stand. A pretrial hearing, on the other hand, permits “relevant testimony [to be] elicited in full,” United States v. Blunt, 930 F.3d 119, 128 n.6 (3d Cir. 2019), and eliminates guesswork. That benefits both courts and litigants, as a proffer may overstate or understate the extent to which one witness's testimony might conflict with another's. We thus reiterate the advantages of conducting a pretrial hearing in situations where defendants may present mutually antagonistic defenses. But when one is not held, a proffer is as close a substitute as possible, and frames our consideration of reasonably foreseeable testimony.

In view of Murphy's various proffers, we will consider seriatim the District Court's decisions to deny Green's motion to sever. He first raised his motion in pretrial papers, and the District Court denied it without prejudice. That was the correct decision, and we do not disturb it here; the District Court and Green were unaware of the contours of Murphy's duress defense until trial began, and Murphy had made no proffers that would shed light on his intentions. Indeed, even Green conceded the possibility that his motion was premature at the time he first made it.

Our review of the District Court's denials of Green's motions to sever becomes more complicated as the record progresses. Still, based on the information before it, we conclude that Green's motion after the first day of trial did not warrant severance. At the time, the District Court did not know whether Murphy would even testify or what the Government's case-in-chief would show. Without that information, the District Court was placed in the difficult position of predicting the content and scope of witness testimony based solely on the parties' opening statements and Murphy's single proffer. And given those unanswered questions, we cannot fault the District Court for erring on the side of our default preference for joint trials.

But we must still consider Green's final motion at the end of the Government's case-in-chief. Several pieces of testimony had crystalized the events in question. Snowell, testifying for the Government, claimed that Murphy had driven him and Green to his aunt's house, that Green had been armed with a gun, that Green threatened him with the gun, and that he managed to escape from his aunt's house after the trio (Green, Murphy, and Wong by cell phone) was unable to find the money they sought. In other words, Snowell described a kidnapping in which only Green had personally threatened him with a firearm—consistent with Murphy's counsel's earlier proffer regarding Snowell's expected testimony.

When Green made his third severance motion, Murphy's counsel was called upon once more to proffer what his client would testify. Counsel noted that Murphy would identify Wong as the “major force” behind the kidnapping, but also say that Green was in phone contact with Wong, possessed a gun, and was prepared to shoot Snowell. So the testimony Snowell had just provided that he had never been free to leave and that a kidnapping had, in fact, occurred, would be corroborated by Murphy, directly rebutting Green's defense. Moreover, Snowell's testimony was consistent with much of the testimony Murphy was projected to give in support of his own coercion defense. So the testimony Snowell had just provided was consistent with the testimony Murphy was projected to give.

True, the District Court had not yet heard Murphy's testimony. It therefore lacked complete information about what Murphy would say about Green. But courts must also consider what trial evidence is reasonably foreseeable and assess its impact on allegedly irreconcilable defenses. See Blunt, 930 F.3d at 125 (concluding that co-defendant's motion papers, in which she admitted she would offer evidence admissible only through her own testimony, showed that testimony necessitating severance would come in at trial). Here, Murphy's counsel's series of proffers rendered foreseeable enough Murphy's intention to testify to events implicating Green. While no proffer can replace hindsight, Murphy was consistent in advising that his testimony would include at least some information tending to prove that Green had coerced him into a kidnapping. That expected testimony runs straight into the District Court's earlier question: “So you're not going to advance a defense that the coercion about which you spoke ․ came from, directly or indirectly, from Mr. Green?” App. 159. Murphy's proffers, while cloaked in a focus on Wong, previewed exactly the sort of defense the District Court was concerned about. They thus provide a strong reason to sever Green's trial from Murphy's.

A further justification for severance can be found in a point implicit in our prior decisions, but that we have not discussed at any length: that for multiple reasons, defenses are less likely to be mutually antagonistic where they arise as part of a conspiracy, as opposed to part of a joint trial of several defendants charged with separate substantive crimes. That is because conspiracies are unique; unlike substantive offenses, they “require[ ] an agreement—and in particular an agreement to do an unlawful act—between or among two or more separate persons,” and do not depend on whether the substantive offense is actually committed. Ziglar v. Abbasi, 582 U.S. 120, 153, 137 S.Ct. 1843, 198 L.Ed.2d 290 (2017); Salinas v. United States, 522 U.S. 52, 65, 118 S.Ct. 469, 139 L.Ed.2d 352 (1997). Therefore, when a co-conspirator's defense implicates a defendant in committing a substantive offense, that evidence is less likely to be unduly prejudicial.

So, grouping defendants who are part of a single conspiracy into a single trial makes intuitive sense. Because a conspirator's crime is an agreement to commit another crime, trying together all the parties to that agreement is both efficient and a better means of presenting the scope of a criminal conspiracy to a jury. United States v. Provenzano, 688 F.2d 194, 199 (3d Cir. 1982), abrogated on other grounds by Reves v. Ernst & Young, 507 U.S. 170, 179, 113 S.Ct. 1163, 122 L.Ed.2d 525 (1993). For that reason, other circuits have emphasized that the rule that “persons who are jointly indicted shall be tried together ․ applies with particular force to conspiracy cases.”³ United States v. Gonzalez, 804 F.2d 691, 694 (11th Cir. 1986) (internal citation omitted). And for our part, we “believe that good reasons support the general rule that persons charged with conspiracy should be tried together.” United States v. De Peri, 778 F.2d 963, 984 (3d Cir. 1985). “Chief among these is the conservation of public resources that would be lost if the same evidence were presented at separate trials and the decreased possibility of prejudice where the evidence against each defendant is strong.” Id. And if that were not enough, joint conspiracy trials also limit “the tactical disadvantage to the government from disclosure of its case.” Jackson, 649 F.2d at 973.

But what makes conspiracies amenable to joint trials is what can make substantive charges against individuals comparatively suited to separate trials. Unlike in a conspiracy case, the Government need not prove the existence of an agreement or any cooperative criminal activity when individuals are charged with substantive offenses. And the possibility of prejudice is heightened because a co-defendant's evidence may have little prejudicial effect on a defendant's guilt in agreeing to commit an offense, yet that same evidence may be significantly more harmful as to a defendant's guilt in committing a substantive offense. See Zafiro, 506 U.S. at 539, 113 S.Ct. 933. While separate trials heighten the risk that the Government's trial strategy will be revealed piecemeal to a number of defendants, Jackson, 649 F.2d at 973, that risk is lessened when the Government is obligated to prove up substantive charges against individual defendants.

The Government did not charge Wong, Green, and Murphy as part of a conspiracy. Whether that was the result of a lack of evidence of an agreement or for any other reason is not for us to question. But with no allegations of a conspiracy, evidence offered by Murphy that might have been harmless in the context of a conspiracy could now be very damaging. See Zafiro, 506 U.S. at 539, 113 S.Ct. 933. Add to this the mutually antagonistic defenses presented by Green and Murphy, and the difficulties with conducting a joint trial of non-coconspirators become even more evident. With no allegation that Green, Murphy, and Wong had agreed to pursue criminal ends, the risk of prejudice is heightened further; the jury might be more likely to believe that one defendant committed a substantive offense merely because his co-defendant—but not co-conspirator—committed a crime.

In sum, we recognize that “courts have consistently held that finger-pointing and blame-shifting among coconspirators do not support a finding of mutually antagonistic defenses.” Voigt, 89 F.3d at 1095. And, to be sure, there is an element of blame-shifting in Murphy and Green's versions of events. But Murphy's proffers and the lack of any conspiracy mean that their defenses go beyond finger-pointing; instead, they resemble a triangular prosecution between the Government, Green, and Murphy.

B.

Beyond Green and Murphy's dueling defenses, prejudicial joinder may exist in cases where evidence is either admitted or barred from the courtroom because of a trial's joint nature.⁴ And so we consider evidence used to inculpate Green that was admissible only through Murphy. Specifically, Murphy testified that federal agents were interested in interviewing him because they planned to “take down Green and Wong.” App. 261, 269. That prejudicial evidence was unaccompanied by any limiting instruction.⁵

Prejudicial joinder “might occur when evidence that the jury should not consider against a defendant and that would not be admissible if a defendant were tried alone is admitted against a codefendant.” Zafiro, 506 U.S. at 539, 113 S.Ct. 933. That is what happened in this case, as Murphy testified that federal agents said they could help him avoid legal jeopardy from the FBI and ATF. He claimed they made that promise “as long as [he] help[ed] them take down Green and Wong,” because they “really want[ed] them for other crimes that they've committed.”⁶ App. 269.

That evidence is doubtless prejudicial; Murphy linked Green directly with ongoing investigations by two federal law enforcement agencies, information that would be entirely inadmissible in a solo trial of Green. See Fed. R. Evid. 401 (relevance), 403 (probative value substantially outweighed by risk of unfair prejudice), 404 (other bad acts). And no curative instruction was offered following that testimony.⁷ It follows that—even apart from the mutually antagonistic character of their defenses—the joint trial also exposed Green to unwarranted prejudicial testimony. But we must focus on that which was actually known or reasonably foreseeable to the District Court at the time it considered the motion to sever, McGlory, 968 F.2d at 340, and we cannot conclude that the District Court would have reasonably foreseen this evidence. As with our preceding discussion, this evidence alone might not warrant severance under Rule 14; that said, its admission as part of Green's joint trial again highlights the risk of prejudice that was posed in denying the motion and may be considered when we assess whether “clear and substantial” prejudice resulted from that denial.

The Supreme Court has separately noted that “a defendant might suffer prejudice if essential exculpatory evidence that would be available to a defendant tried alone were unavailable in a joint trial.” Zafiro, 506 U.S. at 539, 113 S.Ct. 933. Green identifies just one piece of exculpatory evidence unavailable to him at a joint trial: Murphy's testimony that a bystander asked him and Snowell—while they were alone in the car at Wawa—whether they needed help, and that they turned down the offer.⁸ That evidence, Green argues, would tend to show that neither Murphy nor Snowell thought they were in the midst of a kidnapping.

Nevertheless, Murphy offered that testimony only at Wong's subsequent severed trial. While this fact pattern is unusual, and involves information only gleaned from a co-defendant's subsequent testimony in another co-defendant's trial, it cannot alter the scope of our analysis. To determine whether severance is warranted, we look solely to what the District Court knew or could reasonably be expected to know at the time a motion to sever is made. McGlory, 968 F.2d at 340. The District Court lacked actual knowledge of Murphy's future testimony at Wong's separate trial, and it was not reasonably foreseeable that Murphy would offer new testimony about a good Samaritan at a gas station. So we cannot conclude that the District Court abused its discretion in denying severance on the grounds of unavailable exculpatory evidence.

* * *

So, we are left with just one conclusion: that the denial of Green's severance motion was improper. Green and Murphy presented mutually antagonistic defenses, and Murphy inculpated Green in a federal investigation during his testimony. But misjoinder, by itself, does not require severance; rather, we must consider whether “clear and substantial prejudice” has resulted from misjoinder. McGlory, 968 F.2d at 340. Eufrasio, 935 F.2d at 568. Only then can we determine whether the District Court abused its discretion in refusing to sever Green's trial.

V.

“Rule 14 leaves the determination of risk of prejudice and any remedy that may be necessary to the sound discretion of the district courts.” Zafiro, 506 U.S. at 541, 113 S.Ct. 933. The District Court abused its discretion in assessing the prejudice stemming from a joint trial, and the only appropriate remedy in this case is to reverse the denial of Green's motion to sever and vacate Green's conviction.

A.

First, we consider the risk of prejudice. As noted, “[a] defendant must pinpoint clear and substantial prejudice resulting in an unfair trial” in order to win reversal of a district court's denial of a motion to sever. McGlory, 968 F.2d at 340 (internal quotation marks omitted). “It is not enough to show that severance would have increased the defendant's chances of acquittal.” Id. Nor are “[m]ere allegations of prejudice ․ enough” to require vacatur of Green's conviction. Reicherter, 647 F.2d at 400.

As we noted above, severance may be required in three instances. We addressed the reasonable foreseeability of mutually antagonistic defenses above. Here, we address whether Green suffered clear and substantial prejudice on any of those three grounds. Green has made his required showing.

First, he presents far more than “mere allegations” of prejudice. Green references Murphy's testimony that the FBI and ATF told him they were targeting Green and Wong—rather than Murphy—as part of their investigation. He argues that Murphy extensively portrayed himself as Green's victim, explaining in his closing argument that “Green ha[d] a gun ․ I was forced, they were threatening me, what was I supposed to do?” App. 293. Most importantly, Murphy's “claim of innocence [was] predicated solely on” Green's guilt. Voigt, 89 F.3d at 1094. Murphy claimed he was coerced into assisting in a kidnapping. That kidnapping involved traveling in a car along with Green and Snowell. Snowell was not a kidnapper, and Murphy's defense was that he was not one either. Believing Murphy's recounting of events thus called for the jury to believe Green coerced him into kidnapping Snowell. And, to take the reverse, believing Green's story—which involved leaving Murphy and Snowell alone in the car at a gas station while he bought water, apparently free to leave—would call for the jury to disbelieve Murphy's coercion theory.

These tangled defenses are, in part, exactly why our severance doctrine asks whether juries “ ‘can reasonably be expected to compartmentalize the evidence as it relates to separate defendants in view of its volume and limited admissibility.’ ” United States v. Sandini, 888 F.2d 300, 307 (3d Cir. 1989) (quoting United States v. De Larosa, 450 F.2d 1057, 1065 (3d Cir. 1971)); see also United States v. Alexander, 982 F.2d 262, 266 (8th Cir. 1992) (holding that a defendant must show he was “prejudiced by the jury's inability to follow the trial court's instructions and to ‘compartmentalize the evidence’ as it related to the separate defendants” in order to warrant severance). Here, the trouble with compartmentalizing Murphy's testimony implicating Green is immediately apparent. Murphy's testimony about Green is linked to the jury's assessment of Green's guilt, and Green's defense is linked to the jury's assessment of Murphy's guilt as well. So, we cannot endorse the District Court's conclusion that the jury would be able to compartmentalize inculpatory evidence against Green and Murphy; to do so would vitiate our prejudicial joinder doctrine. Taking all this together, the risk of prejudice stemming from so irreconcilable a set of defenses—together with inculpatory evidence admitted through a co-defendant—is substantial, and requires a remedy.⁹

B.

We reject the argument that other measures taken by the District Court adequately addressed this prejudice. We are mindful that “Rule 14 does not require severance even if prejudice is shown; rather, it leaves the tailoring of the relief to be granted, if any, to the district court's sound discretion.” Zafiro, 506 U.S. at 538–39, 113 S.Ct. 933. To that end, we have intentionally declined to adopt a bright-line rule mandating that all cases of prejudicial joinder require severance. United States v. Balter, 91 F.3d 427, 432–33 (3d Cir. 1996). There are two remedies a district court could consider in similar circumstances: severance, or limiting instructions. See Zafiro, 506 U.S. at 539, 113 S.Ct. 933. And the choice between those remedies is tied to the risk of prejudice. “When the risk of prejudice is high, a district court is more likely to determine that separate trials are necessary, but, ․ less drastic measures, such as limiting instructions, often will suffice to cure any risk of prejudice.” Id.

Here, the risk of prejudice is quite high. True, the District Court properly instructed the jury “(1) to consider each count of the indictment and each defendant's involvement in that count separately, (2) that the burden is always on the prosecution to prove guilt beyond a reasonable doubt, (3) that its verdict as to any defendant on any count should not control [its] verdict as to any other defendant or any other count, and (4) that opening statements and closing arguments are not evidence.” Voigt, 89 F.3d at 1096 (internal quotation marks omitted). But Voigt instructions are not a panacea. In some cases, the risk of prejudice will be so high that severance is required.

For example, in Blunt we confronted similar circumstances—with one defendant advancing a duress theory and attempting to cast blame on the other—and reversed the district court's denial of a motion to sever. 930 F.3d at 121. This was despite the fact that the Blunt jury received Voigt instructions, as did the jury in this case. We reach the same conclusion here as in Blunt. In cases where mutually antagonistic defenses prejudice a defendant—as opposed to prejudicial joint trials in situations where, say, evidence that implicates a defendant is admitted for other purposes through a co-defendant—a limiting instruction will likely prove insufficient to cure prejudice. That is because mutually antagonistic defenses speak to the core of a defendant's theory of the case, and so any limiting instruction would necessarily restrict a defendant's ability to present his chosen defense. We will not endorse that result. Green has thus made his requisite showings with respect to improper joinder, prejudice, and the necessity of severance.

VI.

In the end, we take the Government at its word: Murphy contended that he had participated in a kidnapping under duress, Green contended that no kidnapping occurred, and “[o]bviously, both of those theories cannot be true.” App. 294. Because Green's joint trial was improper, and the resulting prejudice substantial, denying Green's final renewed motion to sever was an abuse of discretion warranting a new trial.

For the foregoing reasons, we will reverse the District Court's denial of Green's motion to sever, vacate his conviction and sentence, and remand for further proceedings. In light of our disposition of this case, we need not reach Green's challenge to his status as a career offender.

FOOTNOTES

2.   Some courts conduct a joint trial with two juries, another way of enforcing constitutional rights and preserving judicial economy.

3.   Proving the point, every leading case on the mutually antagonistic defense doctrine from our circuit, as well as from the Supreme Court, has arisen in the conspiracy context. Zafiro, 506 U.S. at 536, 113 S.Ct. 933 (drug conspiracy); McGlory, 968 F.2d at 314 (drug conspiracy); Voigt, 89 F.3d at 1096 (wire fraud conspiracy); Boscia, 573 F.2d at 829 (mail fraud conspiracy). And, unsurprisingly, those cases did not turn out in their respective defendants' favor. Our sister circuits also emphasize the role of conspiracy charges in our default joint-trial framework. See United States v. Joiner, 418 F.3d 863, 868 (8th Cir. 2005) (“Generally, co-defendants charged with a conspiracy should be tried together because a joint trial gives ‘the jury the best perspective on all of the evidence and therefore increases the likelihood of a correct outcome.’ ” (citation omitted)); United States v. Knowles, 66 F.3d 1146, 1158 (11th Cir. 1995) (“This court is reluctant to reverse a district court's denial of severance, particularly in conspiracy cases.”); United States v. Dempsey, 733 F.2d 392, 398 (6th Cir. 1984) (“As a general rule, especially in conspiracy cases, parties who are jointly indicted should be tried together.”); United States v. Kelly, 569 F.2d 928, 938 (5th Cir. 1978) (“The general rule, especially in conspiracy cases, is that persons jointly indicted should be tried together and that severance should not be granted absent a showing of the most compelling prejudice.”). That same reasoning is persuasive here.

4.   Our caselaw addressing mutually antagonistic defenses has, at times, appeared to conflate irreconcilable defenses with the notion of prejudicial joinder generally. See United States v. Blunt, 930 F.3d 119, 128 (3d Cir. 2019) (discussing “mutually antagonistic defenses” of co-defendant spouses, but holding that severance was warranted because of evidence that could have helped the defendant-wife but was excluded to avoid prejudicing the defendant-husband). Mutually antagonistic defenses are only one means of demonstrating prejudicial joinder, however; other circumstances—such as evidence admitted through a co-defendant that would not otherwise come into a trial—provide independent reasons that joinder may be prejudicial. Zafiro, 506 U.S. at 539, 113 S.Ct. 933.

5.   Another piece of Murphy's testimony arguably implicated Green: namely, that the Infamous Ryders had ordered the murder of a police officer. Green raised—and the District Court sustained—an objection to that testimony under Federal Rule of Evidence 404(b), and issued a curative instruction to the jury.

6.   He did not repeat that statement at Wong's subsequent trial.

7.   Green did not object to this testimony on severance grounds, and thus arguably forfeited any claim stemming from that portion of Murphy's testimony. But we have explained that our forfeiture doctrine permits us “to excuse waiver or forfeiture concerns ․ especially when the government or appellee overlooks or disregards waiver or forfeiture.” United States v. Washington, 869 F.3d 193, 208 n.53 (3d Cir. 2017). Here, the Government makes no mention of waiver or forfeiture in its briefing. We thus consider in our prejudicial joinder analysis Murphy's mention of the ongoing federal investigation into Green and Wong.

8.   Green points to several other pieces of evidence that he claims were unavailable to him, but all of that information was put before the jury at his joint trial.

9.   Notably, this case provides a clear counterfactual to Green's conviction at his joint trial: Steven Wong's subsequent acquittal. While each jury is unique, Wong and Green occupied similar roles—that is, as the two “real” kidnappers—in Snowell and Murphy's recounting of events. Green's conviction is thus all the more conspicuous, given that both of his co-defendants were acquitted despite seemingly similar (or even lesser) degrees of culpability. However, the District Court was obviously unaware of these acquittals—they had not yet taken place at the time of Green's severance motions—so they cannot play a role in our broader severance analysis. McGlory, 968 F.2d at 340.

FISHER, Circuit Judge.