OVERVIEW

Together with his older brother Tamerlan, Dzhokhar Tsarnaev detonated two homemade bombs at the 2013 Boston Marathon, thus committing one of the worst domestic terrorist attacks since the 9/11 atrocities. Radical jihadists bent on killing Americans, the duo caused battlefield-like carnage. Three people died. And hundreds more suffered horrific, life-altering injuries. Desperately trying to flee the state, the brothers also gunned down a local campus police officer in cold blood. Reports and images of their brutality flashed across the TV, computer, and smartphone screens of a terrified public — around the clock, often in real time. One could not turn on the radio either without hearing something about these stunningly sad events.

Dzhokhar eventually got caught, though Tamerlan died after a violent confrontation with the police.

Indicted on various charges arising from these ghastly events, Dzhokhar stood trial about two years later in a courthouse just miles from where the bombs went off. Through his lawyers, he conceded that he did everything the government alleged. But he insisted that Tamerlan was the radicalizing catalyst, essentially intimidating him into acting as he had. See 18 U.S.C. § 3592(a)(4) (providing that relative culpability is a mitigating factor relevant to the imposition of a death penalty). Apparently unconvinced, a jury convicted him of all charges and recommended a death sentence on several of the death-eligible counts — a sentence that the district judge imposed (among other sentences).

A core promise of our criminal-justice system is that even the very worst among us deserves to be fairly tried and lawfully punished — a point forcefully made by the then-U.S. Attorney for Massachusetts during a presser at the trial's end. To help make that promise a reality, decisions long on our books say that a judge handling a case involving prejudicial pretrial publicity must elicit "the kind and degree" of each prospective juror's "exposure to the case or the parties," if asked by counsel, see Patriarca v. United States, 402 F.2d 314, 318 (1st Cir. 1968) — only then can the judge reliably assess whether a potential juror can ignore that publicity, as the law requires, see United States v. Vest, 842 F.2d 1319, 1332 (1st Cir. 1988). But despite a diligent effort, the judge here did not meet the standard set by Patriarca and its successors.

Another error forces us to act as well, this one involving the judge's denial of Dzhokhar's post-trial motion for judgments of acquittal. Navigating a complex and changing area of the law, the judge let stand three of Dzhokhar's convictions for carrying a firearm during crimes of violence, in violation of 18 U.S.C. § 924(c). The judge thought that each of the underlying offenses constituted a crime of violence. But with respect (and with the luxury of time that district judges rarely have), we believe the current state of the law propels us toward the opposite conclusion.

The first error requires us to vacate Dzhokhar's death sentences and the second compels us to reverse the three § 924(c) convictions. On remand, then, the district court must enter judgments of acquittal on the relevant § 924(c) charges, empanel a new jury, and preside over a new trial strictly limited to what penalty Dzhokhar should get on the death-eligible counts. And just to be crystal clear: Because we are affirming the convictions (excluding the three § 924(c) convictions) and the many life sentences imposed on those remaining counts (which Dzhokhar has not challenged), Dzhokhar will remain confined to prison for the rest of his life, with the only question remaining being whether the government will end his life by executing him.

What follows is an explanation of our reasoning, as well as our take on certain issues that may recur on remand.

HOW THE CASE CAME TO US

The facts of today's appeal are painful to discuss. We apologize for their graphic detail. We also do not attempt to cover all of the case's complicated events in this section — instead we offer only a summary, adding more information during our discussion of particular issues.

Bombings

On Patriots' Day 2013 — April 15, to be exact, a local holiday celebrating the first battles of the American Revolution — the Tsarnaev brothers set off two shrapnel bombs near the finish line of the world-famous Boston Marathon, leaving the area with a ravaged, combat-zone look. BBs, nails, metal scraps, and glass fragments littered the streets and sidewalks. Blood and body parts were everywhere — so much so that it seemed as if "people had just been dropped like puzzle pieces onto the sidewalk" (a description taken from a witness's trial testimony). The smell of smoke and burnt flesh filled the air. And screams of panic and pain echoed throughout the site. "Mommy, mommy, mommy," a five-year-old boy cried out over and over again, his leg cut to the bone. Others yelled "help us," "we're going to die," or "stay with me."

Now brace yourself for how Marathon-goers Krystle Campbell, Lingzi Lu, and Martin Richard spent the last few minutes of their lives.

Krystle Campbell was watching the race with her friend Karen Rand (now known as Karen McWatters) when the first bomb went off. Rand got knocked to the ground. But she dragged herself over to Campbell, burning herself on hot metal pieces as she did so. She put her face next to Campbell's and held her hand. Campbell was "complete[ly] mutilat[ed]" from the waist down. Speaking very slowly, Campbell said her "legs hurt" — even though they had been blown off. Moments later, her hand went limp in Rand's. And she never spoke again, bleeding to death right there.

Lingzi Lu was with her friend Danling Zhou as the second bomb exploded. Zhou got a gash across her stomach, requiring her to hold her insides in. While Lu had her arms and legs, she did have a significant thigh injury. "[B]asically her leg had been filleted open down to the bone," a doctor on the scene later said. He tried to tourniquet the wound. But she had lost a lot of blood and didn't have much of a pulse. Despite knowing that she was going to die, the doctor asked a nearby person to start CPR. A firefighter later moved in and pumped air into Lu's mouth with his mask. And a police officer did chest compressions, telling her, "[S]tay with us. You can do this. You're going to be okay. Stay strong." He and others put Lu on a backboard and placed her in an ambulance. But a paramedic told them to get her off because she "was gone" and he needed to keep the ambulance free for those who could be "save[d]."

The second bomb also sent BBs and nails tearing through eight-year-old Martin Richard's body, cutting his spinal cord, pancreas, liver, kidney, spleen, large intestine, and abdominal aorta, and nearly severing his left arm. He bled to death on the sidewalk — with his mother leaning over him, trying to will him to live. Searching for his two other children, Martin's father, Bill, first found son Henry (age twelve) and then daughter Jane (age six). Jane tried to stand up but fell — because her left leg was gone. Bill carried her for a bit. And then an off-duty firefighter tourniqueted the leg, saving her life.

Not only did the Tsarnaev brothers kill Krystle Campbell, Lingzi Lu, and Martin Richard, they also consigned hundreds of others to a lifetime of unimaginable suffering. Some lost one or more limbs, blown off as they stood near the finish line or amputated later because they were so badly mangled. Others lost sight, still others hearing. And years after the bombings, many still had debris in their bodies. One survivor had shrapnel in her that occasionally worked its way to the surface and had to be removed; another had a ball bearing stuck in his brain — to give just a few examples.

Manhunt and Capture

Leaving the scene, Dzhokhar and Tamerlan drove to Cambridge and stopped at a Whole Foods. Dzhokhar went in, grabbed a bottle of milk, paid for it, and left. About a minute later, he returned to the store and exchanged the bottle for a different one.

Back at his school the next day (UMass Dartmouth, as it is known locally), Dzhokhar resumed his normal routine. He worked out with a friend at the campus gym, for example. "I'm a stress free kind of guy," he tweeted.

Aided by a description given by a man from his hospital bed, as well as by videos from security cameras and bystanders' cell phones, law enforcement released images of the bombers two days later on April 18, and asked the public to help identify them and provide information about their whereabouts. The FBI (short for the Federal Bureau of Investigation) produced a "wanted poster" on its website and asked the local community to give any details that could lead to their arrests.

That night, still April 18, Tamerlan and Dzhokhar put pipe bombs, a handgun, and a shrapnel bomb (similar to the ones they exploded at the Marathon) into Tamerlan's Honda Civic and drove off from his Cambridge home. Passing by the Massachusetts Institute of Technology ("MIT," from now on), they spotted the squad car of campus police officer Sean Collier. Approaching the squad car from behind, the brothers shot Collier dead at close range — twice in the side of the head, once between the eyes, and three times in the hand. They tried and failed to take his gun (the holster's retention system tripped them up, apparently). Startled by an MIT student riding by on a bike, they got back in the Honda and sped away.

The brothers then drove to Brighton where they crossed paths with Dun Meng, who was sitting in his parked Mercedes SUV. After pulling up behind him, Tamerlan got out of the Honda and knocked on Meng's passenger window. Meng rolled the window down. Tamerlan leaned in, opened the door, jumped in, aimed a gun at him, and demanded cash. A frightened Meng handed over his wallet. Explaining that he had exploded the bombs at the Marathon and had just killed Collier, Tamerlan ordered Meng to drive. So drive Meng did. Dzhokhar followed behind in the Honda.

Tamerlan eventually had Meng pull over on a street in Watertown. Tamerlan got into the Mercedes's driver's seat. Meng got into the Mercedes's front passenger's seat. And after parking the Honda, Dzhokhar got into the Mercedes's back passenger's seat. Tamerlan drove to an ATM. Dzhokhar withdrew $800 from Meng's account using Meng's bank card and PIN ($800 was the card's withdrawal limit).

Tamerlan drove back to his Honda to get a music CD with nasheeds on them (nasheeds are Islamic chants). He played the CD in the Mercedes — music that sounded a "bit weird" and "religious" to Meng. Tamerlan then stopped for gas in Cambridge. Fearing this might be his last chance to escape, Meng made a break for it, sprinting across the street to a different gas station where he begged the attendant to call the police. Tamerlan and Dzhokhar took off in the Mercedes. Meng told the arriving officers that the carjackers were the Boston Marathon bombers. He also told them that his Mercedes had a built-in tracking system.

Good police work then led authorities to the Tsarnaev brothers in Watertown, where the two had returned to get the Honda. Tamerlan got out of the Mercedes. Dzhokhar got out of the Honda. Tamerlan started shooting at the officers. And he and Dzhokhar threw some pipe bombs and a larger bomb that looked "like a big cooking pot." A couple of the bombs exploded.

After getting shot, and having possibly run out of bullets, Tamerlan tossed his gun at one of the officers and ran toward them. They wrestled him to the ground, however. Dzhokhar got back in the Mercedes. And while trying to make his getaway, he ran over his brother, who died hours later.

Also hurt in the shootout was MBTA police officer Richard Donohue. He nearly bled to death after a bullet hit his groin area. It took months in a hospital to recover from his injury.

Back to Dzhokhar. He could only drive about two blocks, because the police had damaged the Mercedes's tires. So he exited the car and fled on foot. In a nearby backyard, he found a boat shrink-wrapped in plastic and climbed inside. And he stayed there overnight, bleeding from his wounds.

Dzhokhar did, however, have enough strength to write a manifesto justifying his actions. On two wooden slats attached to the boat he carved the words, "Stop killing our innocent people and we will stop." "God has a plan for each person," he wrote on the fiberglass hull (with a pencil he found on the boat). "Mine was to hide in this boat and shed light on our actions." "[J]ealous" of Tamerlan's martyrdom, he accused "[t]he U.S. Government [of] killing our innocent civilians." Stressing that he could not "stand to see such evil go unpunished," he warned that "we Muslims are one body, you hurt one, you hurt us all." And finishing up, he wrote, "Now I don't like killing innocent people it is forbidden in Islam but due to said [ ] it is allowed."

With Dzhokhar still at large, then-Massachusetts Governor Deval Patrick asked nearly a million citizens of Boston and the five neighboring locales (Belmont, Cambridge, Newton, Waltham, and Watertown) to "shelter in place" — that is, he told them to remain behind closed doors and "not to open the door for anyone other than a properly identified law enforcement officer." He also asked schools and businesses to close — only hospitals and law enforcement would stay open.

Later that day, on April 19, Watertown resident David Henneberry noticed that his boat had some loose shrink wrap and went out to fix it — by this time Governor Patrick had lifted the shelter-in-place order, even though Dzhokhar was still a fugitive. As Henneberry climbed up a ladder, he saw blood in the boat and a person lying there with a hooded sweatshirt pulled over his head. So he raced back inside his house and called 911.

Officers responded rapidly. And after Dzhokhar ignored repeated requests to surrender, they threw flash-bang grenades into the boat and fired a barrage of bullets at it. Officers finally arrested him about 90 minutes after Henneberry's call.

An ambulance took Dzhokhar to a hospital where he underwent hours of emergency surgery to treat his injuries — injuries that included a gunshot wound to the left side of his face and multiple gunshot wounds to his extremities. Doctors sutured his left eye shut and wired his jaw closed. He could not hear out of his left ear. He had to be intubated and took narcotic pain meds intravenously. FBI agents questioned him off and on for more than 13 hours over the next 36 hours without Miranda-izing him — and without letting lawyers from the federal public defender's office see him either. Most questions he answered by nodding yes or no, or by writing his responses down in a notebook. Repeatedly, he asked for a lawyer. But the agents told him that he first needed to answer their questions "to ensure that the public ... was no longer in danger." Exhausted and in pain, he asked the agents several times to stop questioning him. But the record does not indicate whether the agents honored these requests. At some point he told them that after the bombings, he and Tamerlan fled Boston by car, and "[o]n the way back to Cambridge, they stopped at a Whole Foods ... to buy some milk" — the two "were observing the Muslim tradition of fasting on Mondays and Thursdays and needed milk to break the fast."

Unsurprisingly, the bombings and their aftermath dominated Boston-area TV, radio, newspapers, and magazines — not to mention web and social-media sites. Locals quickly adopted the "Boston Strong" slogan to convey a message of courage and resilience. And the Boston Strong movement remains vibrant to this very day.

Legal Proceedings

Eventually, a Boston-based federal grand jury indicted Dzhokhar for crimes arising from his unspeakably brutal acts. The indictment also included a number of specific allegations necessary for seeking capital punishment under the Federal Death Penalty Act ("FDPA"), 18 U.S.C. §§ 3591 - 99, which governs aspects of this case. And the government later notified him that it would seek the death penalty on all 17 death-eligible counts (Counts 1-10 and 12-18). See 18 U.S.C. § 3593(a). Because of the extensive pretrial publicity in the Boston area, Dzhokhar filed motions to change venue before the guilt phase started (a capital trial has two phases, a guilt phase and a penalty phase) — motions that the judge denied, though he did promise to conduct a thorough and searching voir dire. A French phrase that (roughly translated) means "to speak the truth," voir dire (as relevant here) "is a process through which a judge or lawyer examines a prospective juror to see if the prospect is qualified and suitable to serve on a jury." See United States v. Parker, 872 F.3d 1, 3 n.1 (1st Cir. 2017) (quotation marks omitted). But the judge stopped Dzhokhar's counsel from asking prospective jurors questions like "[w]hat did you know about the facts of this case before you came to court today (if anything)?" and "[w]hat stands out in your mind from everything you have heard, read[,] or seen about the Boston Marathon bombing and the events that followed it?"

During the guilt phase of his trial, Dzhokhar's lawyers did not dispute that he committed the charged acts. Rather, their guilt-phase defense rested on the idea that he participated in these horrible crimes only under Tamerlan's influence. In her opening statement, for instance, one of his attorneys said that "[i]t was him" and that the defense would not "attempt to sidestep" his "responsibility for his actions." But she said that his terrorist path was "created by his brother." And in her closing argument, she said that he "stands ready ... to be held responsible for his actions." Ultimately the jury convicted him on all counts after hearing testimony from nearly 100 witnesses and after receiving over 1,000 exhibits.

The jury later reconvened for the penalty phase. The parties combined to present some 60 witnesses and introduce over 180 exhibits. And after following the process just outlined, the jury recommended the death penalty on six of the death-eligible counts (Counts 4, 5, 9, 10, 14, and 15). The judge, for his part, sentenced Dzhokhar to die, while also giving him a number of concurrent and consecutive prison terms on the remaining counts — including 20 life terms.

And this timely appeal ensued.

DISCUSSION

Dzhokhar's briefs raise 16 issues for review, many with sub-issues and even sub-sub-issues. As we have previewed already, the judge's Patriarca-based error compels us to vacate the death sentences and his crime-of-violence errors require us to reverse three § 924(c) convictions. Not only do we explain those errors below. We also address other issues (even if just briefly) because we know they are likely to resurface on remand. See generally Swajian v. Gen. Motors Corp., 916 F.2d 31, 35 (1st Cir. 1990) (taking a similar approach in a similar situation). So our opinion proceeds as follows. Essentially taking the issues in the order presented to us, we start with venue but pivot to the jury-selection process — because the judge's promise to hold a searching voir dire helped drive his decision to deny a venue change, but his handling of voir dire did not measure up to the standards set by Patriarca and other cases. We then touch on some matters (in varying levels of detail) that could affect how the penalty retrial on the death-eligible counts proceeds — matters like the retrial's location, the government's failure to disclose evidence material to punishment, the judge's admission of evidence, the prosecution's behavior during opening statements and closing arguments, the judge's jury instructions, the prosecution's private communications with the judge, etc . And we last highlight the errors in the judge's crime-of-violence analysis (this is one of the most complex areas of American law, we must say — which is why even well-meaning judges and lawyers sometimes make mistakes).

Trial Venue and Jury Selection

We start with Dzhokhar's claims that the judge erred in how he handled the venue-change motions and the jury-selection process (we have a lot to go over, so please bear with us).

Background

It is no exaggeration to say that the reporting of the events here — in the traditional press and on different social-media platforms — stands unrivaled in American legal history (at least as of today). The highlights (or — as Dzhokhar sees some of it — lowlights) of the coverage include:

• Starting with the bombings themselves, the reporting covered the carnage-filled

terror scene — with the sights and sounds of the wounded and the dying in full display.

• The reporting then covered the ensuing search for the bombers — with images of Dzhokhar leaving a backpack behind Martin Richard and walking away before it exploded, with Governor Patrick's press-conference statements about sheltering-in-place, and with at-the-scene videos showing agents removing a bloodied Dzhokhar from the dry-docked boat.

• The reporting did not get every detail right, however — for example, some falsely claimed that Dzhokhar scrawled "Fuck America" in the boat.

• Other reporting mentioned his non- Miranda-ized statements to agents at the hospital — statements not introduced at trial.

• The reporting also explored the lives and deaths of Krystle Campbell, Lingzi Lu, Martin Richard, and Sean Collier — touchingly describing the pain borne by their families and foreshadowing much of the decedent-victim-impact evidence that the jury would hear. And the reporting anticipated much of the testimony from badly-injured survivors — though it sometimes spotlighted accounts from survivors who would never testify.

• The reporting captured the views of prominent community members about the penalty Dzhokhar deserved. For instance, the Boston Globe reported that despite his past opposition to capital punishment, the then-Boston mayor thought Dzhokhar should "serve[ ] his time and [get] the death penalty." And the Globe reported as well that a former Boston police commissioner believed the government did the right thing in seeking Dzhokhar's execution, given the evidence's strength.

• More still, the reporting generated lots of stories where everyday people in the area called Dzhokhar a "monster," a "terrorist," or a "scumbag[ ]." One article even asked if a particular photo of Dzhokhar was "what evil looks like."

First Venue Motion

In June 2014, Dzhokhar moved (before jury selection) for a change of venue, relying on this avalanche of pretrial publicity. He essentially argued that polling data collected by his expert showed potential jurors in the court's Eastern Division were more likely to consider him guilty than those in the district's Western Division, the Southern District of New York, and the District of Columbia. Convinced that the circumstances triggered a presumption of prejudice in the District of Massachusetts, he "recommend[ed] the District of Columbia as the venue with the least prejudicial attitudes."

Opposing this motion, the government argued that Dzhokhar failed to show that "12 fair and impartial jurors cannot be found" among the Eastern Division's "large, widespread, and diverse" populace. The government also claimed that his expert's analysis had a slew of problems, including the fact that courts in other "highly-publicized trials have found" his expert's "opinions unhelpful, misleading[,] or wrong." And the government suggested that "[f]ar from ‘demonizing’ " Dzhokhar, "the local press has largely humanized him ..., portraying him as a popular and successful student and the beloved captain of his high school wrestling team."

Applying the factors outlined in Skilling v. United States, 561 U.S. 358, 130 S.Ct. 2896, 177 L.Ed.2d 619 (2010), the judge denied Dzhokhar's motion in September 2014. Among other points, the judge noted that the district's Eastern Division has about "five million people," with many of them living outside of Boston — so, he emphasized, "it stretches the imagination to suggest that an impartial jury cannot be successfully selected from this large pool of potential jurors." And, the judge wrote, neither the defense expert's polling nor his newspaper analysis "persuasively show[ed] that the media coverage has contained blatantly prejudicial information that prospective jurors could not reasonably be expected to cabin or ignore." Moreover, some of the expert's results, the judge stressed, clashed with Dzhokhar's "position" because they showed that respondents in other jurisdictions were almost as likely to believe him guilty as respondents in Massachusetts's Eastern Division. Also, while "media coverage ha[d] continued" in the 18 months since the bombings, "the ‘decibel level of media attention,’ " the judge said (quoting Skilling ), had "diminished somewhat." For the judge, Dzhokhar had "not proven that this [was] one of the rare and extreme cases for which a presumption of prejudice is warranted." "[A] thorough evaluation of potential jurors in the pool," the judge continued, "will be made through questionnaires and voir dire sufficient to identify prejudice during jury selection."

Second Venue Motion, First Mandamus Petition, And Joint Proposed Jury Questionnaire

A few months later, in December 2014, Dzhokhar filed a second venue-change motion, protesting that a huge portion of the Eastern Division (again, the pool from which his jury would be drawn) "has been victimized by the attack on the Marathon and the related events" — which, combined with the continuing press coverage, made it impossible to seat an impartial jury. Responding to this motion, the government argued that "most of the articles" Dzhokhar mentioned in this memo "have little or nothing to do with this case, and the ones that do are largely factual and objective in nature." Without waiting for the judge's ruling on the second motion, Dzhokhar — also in December 2014 — petitioned this court for mandamus relief. See Tsarnaev II, 780 F.3d at 17 (discussing timeline).

With Dzhokhar's petition pending, however, the judge — now in early January 2015, and after construing the motion as one for reconsideration of the original venue-change denial — rejected his second venue-change bid (just days after he filed his mandamus petition). See id. Of note, the judge again expressed his confidence that the voir-dire process would ensure jury impartiality. "Should [that] process ... prove otherwise," wrote the judge, "the question of transfer can obviously be revisited." Still in early January 2015, a divided panel of this court then denied Dzhokhar's petition, concluding that he had "not made the extraordinary showing required to justify mandamus relief." See Tsarnaev I, 775 F.3d at 457.

While all this was going on, the parties — in December 2014 — submitted a joint proposed questionnaire for use in voir dire. Some of their suggested questions touched on the potential jurors' general thoughts about capital punishment. Others touched on their exposure to pretrial publicity — questions like: "What did you know about the facts of this case before coming to court today (if anything)?"

In a separate legal memo, the defense — citing Morgan v. Illinois, 504 U.S. 719, 112 S.Ct. 2222, 119 L.Ed.2d 492 (1992) — moved to add more specific questions to identify those prospective jurors who could consider imposing a life sentence not just abstractly, but in the particular circumstances of the case before them. One proposed question, for example, asked potential jurors to

[s]tate whether you agree or disagree with the following statements:

The death penalty is the ONLY appropriate punishment for ANYONE who:

A. murders a child. ? Agree ? Disagree

B. deliberately murders a police officer. ? Agree ? Disagree

C. deliberately commits murder as an act of terrorism. ? Agree ? Disagree

Dzhokhar's team wanted these questions to "probe for a common form of bias — the belief that the death penalty should always or automatically be imposed for certain types of murder ." The government opposed. Relying on and quoting Morgan, the government argued that the defense could ask "whether ‘they will always vote to impose death for conviction of a capital offense’ " generally — not "whether they will consider a sentence less than death" in response to "a laundry list of potential crime elements and aggravating factors." And according to the government, these questions were nothing but impermissible "stakeout questions" — i.e. , questions aimed at getting potential jurors to "stake out a position on the death penalty" before receiving instructions on the law. But according to the defense, "these are the opposite" of stakeout questions since "they seek only to probe whether jurors' minds are open to considering all of the evidence relevant to sentenc[ing]." The judge, however, decided not to include these questions on the questionnaire, saying he would cover those topics in voir dire.

The judge also focused on the jointly-proposed question asking what potential jurors knew "about the facts of this case before coming to court today (if anything)." Conceding that this question "might get very interesting answers," the judge worried that it could "cause trouble because it will be so unfocused." "But if you want to live with it," the judge said to defense counsel, "this is a question that we'll probably be asking every voir dire person."

Despite having had a hand in submitting the questionnaire, the government now switched gears and argued that the question could cause the parties to have to "follow[ ] up on every fact asserted" — something that "would take forever." Apparently persuaded by the government's argument, the judge — after noting that the query could generate "unmanageable data" — ultimately struck the question, explaining that prospective jurors' "preconceptions" could instead be gauged by asking whether, "[a]s a result of what you have seen or read in the news media, ... you [have] formed an opinion" about Dzhokhar's guilt or the proper penalty, and if so, whether "you [can] set aside your opinion and base your decision ... solely on the evidence that will be presented to you in court." The defense objected, saying that "in a case like this[,] where ... you really have no idea what the juror may have swirling around in [his or her] head, it makes the juror the judge of [his or her] own impartiality." "To a large extent that's true," the judge countered, but "the other questions will help us" see if the potential jurors can set aside any preconceived notions about the case — which is "the biggest issue in voir dire, obviously."

Start of Jury Selection, Third Venue Motion, And Second Mandamus Petition

Around the beginning of January 2015, 1,373 potential jurors showed up at the John Joseph Moakley U.S. Courthouse for the start of jury selection. The judge divided them into six panels. As a preliminary matter, the judge twice told them that Dzhokhar was "charged in connection with events that occurred near the finish line of the Boston Marathon ... that resulted in the deaths of three people." And the judge had them fill out a 100-question questionnaire covering their backgrounds, social-media habits, exposure to pretrial publicity (the amount they had seen, whether they had "formed an opinion" about guilt or punishment, etc. ), and thoughts on the death penalty. The questionnaire also gave a "summary of the facts of this case," including that "two bombs exploded ... near the Boston Marathon finish line" and that "[t]he explosions killed Krystle Marie Campbell (29), Lingzi Lu (23), and Martin Richard (8), and injured hundreds of others." "MIT Police Officer Sean Collier (26) was shot to death in his police car," the questionnaire's summary added, and Dzhokhar "has been charged with various crimes arising out of these events." The questionnaire then asked prospective jurors their views on the death penalty for someone convicted of intentional murder and whether they could "conscientiously vote for life imprisonment without the possibility of release."

The judge and the parties identified the potential jurors by numbers. On appeal the parties focus on #138, #286, and #355. So we do too.

Before having the prospective jurors fill out the questionnaires, the judge gave some instructions. For instance, he told them "not to discuss this case with your family, friends[,] or any other person." They could "tell others that you may be a juror in the case," he also said, and could "discuss the schedule with your family and employer." But he warned them "not to discuss anything else, or allow anyone else to discuss with you anything else until you have been excused, or if you're a juror, until the case concludes." And he told them not to "communicate about this case or allow anyone to communicate about it with you by phone, text, message, Skype, email, social media, such as Twitter or Facebook."

As for the answers potential jurors gave on the questionnaires, Dzhokhar — for reasons that will shortly become clear — directs us to these replies by #138, #286, and #355.

Responding to a questionnaire inquiry about social-media use, #138 wrote that he used Facebook "[e]very other day" at "most." And he wrote "N/A" for a question asking whether he had "commented on this case ... in an online comment or post."

For her answer to the same question about social-media use, #286 wrote that she looked at Facebook, Instagram, and Twitter "daily" but did not "post daily." For her response to the question about whether she had "commented on this case ... in an online comment or post," she wrote "don't believe I have." And she wrote "N/A" to the question asking her to explain if she or a family member had been "personally affected by the Boston Marathon bombings or any of the crimes charged in this case (including being asked to 'shelter in place' on April 19, 2013)."

On his jury questionnaire, #355 disclosed that he worked as an "attorney." Answering a question about his death-penalty views, he wrote, "Since it is legal, it should be the rarest of punishments. It is much too prevalent in the country." On a scale of 1 to 10, with 1 meaning "that the death penalty should never be imposed" and 10 meaning that it should be imposed for all cases of "intentional murder," he circled 2. He circled another answer option that read, "I am opposed to the death penalty but I could vote to impose it if I believed that the facts and the law in a particular case called for it." He also wrote that "[k]illing people, especially gov't sponsored killing, is generally wrong" — "[w]hile I can imagine a scenario where facts and law call for it, it is an exceedingly rare case." And responding to the question "[i]f you found [Dzhokhar] guilty and you decided that the death penalty was the appropriate punishment ..., could you conscientiously vote for the death penalty," he checked the box for "I am not sure" — and then explained, "I cannot possibly prejudge his guilt or potential punishment at this stage."

After both sides agreed to excuse many of the 1,373 potential jurors, the judge called back 256 for individual voir dire — which lasted 21 days. And much happened during that time. Dzhokhar continued to ask the judge to ask prospective jurors case-specific questions (like the ones mentioned above) — but to no avail. For instance, Dzhokhar proposed that the judge ask them if they would automatically impose the death penalty if the defendant "killed a child by deliberately using a weapon of mass destruction"; "us[ed] a weapon of mass destruction to carry out an intentional killing"; "deliberately committed an act of terrorism that killed multiple victims"; or "intentionally murder[ed] a police officer in the line of duty." They "know that this is about a bombing," the judge ruled, "and they know that there are three people who were killed in the bombing." Plus they have "my preliminary instructions, ... telling them what the offenses were in general," the judge said. And, the judge stressed, "they have those specifics already in their minds as they ... answer the question about the ability to meaningfully consider life imprisonment." The judge thought, too, that "detailed questioning about what the juror thinks he or she knows about the events" could create the "wrong emphasis" and might inadvertently create bias where none existed before.

Dzhokhar also asked the judge to ask content-specific questions about pretrial publicity (like the ones mentioned earlier) — for example, "What stands out in your mind from everything you have heard, read[,] or seen about the Boston Marathon bombing and the events that followed it?" But the judge rejected that request, saying that "[w]e have detailed answers in the questionnaire concerning ... exposure to the media"; that he saw no need to "repeat" questions "covered in the questionnaire"; and that he thought "digging for details ... will not likely yield reliable answers."

Near the end of January 2015, #138 underwent individual voir dire. The judge reminded him that he had told "everyone to avoid any discussion of the subject matter of the case with anybody," though they "could talk about coming here, obviously, but ... also [had] to avoid any exposure to media articles about the case." And the judge asked #138 if he had "been able to do that." "Yeah," #138 replied, "I haven't looked at anything" or "talked to anybody about it." The judge then turned to the subject of #138's Facebook use (presumably as a follow-up to #138's questionnaire answers). "What's the nature of your use of it," the judge asked, "[i]s it essentially personal, social-type things?" And #138 said, "Yeah." Asked by the judge if he "comment[ed] on public affairs or anything like that," #138 answered, "Yeah, I see what my friends are doing and comment on that." Which prompted the judge to ask if anyone was "commenting about this trial" — to which #138 replied, "No."

The same day as #138's individual voir dire, Dzhokhar filed a third venue-change motion. Highlighting statistics based on questionnaire answers, he noted (among other arguments) that out of a pool of 1,373 prospective jurors, 68% thought he was "guilty, before hearing a single witness or examining a shred of evidence at trial," and 69% "have a self-identified connection or expressed allegiance to the people, places, and/or events at issue in the case." "Stronger" evidence "of presumed prejudice in Boston is difficult to imagine," he wrote. And as far as he was concerned, "[g]iven the extent of prejudice and personal connections," the judge could not count on voir dire to get "a jury that is both actually impartial and preserves the appearance of impartiality."

Disagreeing with Dzhokhar, the government argued (among other assertions) that Dzhokhar's own survey showed "that nearly 100% of respondents in Springfield, New York City, and Washington, D.C. said they had been exposed to publicity about this case." That makes sense, the government added, because "[t]he bombings and their bloody aftermath" made national and international news. And, the government emphasized, the percentage of people who believe Dzhokhar is guilty is greater in those locales than in the Eastern Division (according to the jury questionnaire responses): "84% in Springfield, 92% in New York City, and 86% in Washington, D.C." Also, according to the government, "of the 68% of potential jurors in this case who have formed an opinion that [Dzhokhar] is guilty, fully 60% said they could set aside that opinion and decide the case solely on the [trial] evidence." So as the government saw it, the questionnaires and voir dire could protect Dzhokhar's right to an impartial jury.

Before the judge ruled on the motion, Dzhokhar filed a second mandamus petition with us in early February 2015. But individual voir dire still continued. During her turn — and responding to questions from the judge (who was following up on her questionnaire answers) — #286 disclosed that she used Facebook, Instagram, and Twitter "just [for] social" purposes. "I watch TV," she explained, "and kind of tweet while I'm watching TV with other people that are watching the same programs that I'm watching." And she implied that she had not been "locked down" with her family — saying that while at work on April 19, she "jok[ed]" with her boss that she had to go home. "I live in Boston," she said, "and Boston was on lockdown. I'm, like, I have to go home. We're on lockdown." Defense counsel then asked her if any family or friends had talked with her "about the Marathon bombing[s] ... [o]r any of the events of that week." "[M]aybe in general or something but not really," she said. Dzhokhar's lawyer then asked, "Can you tell us what stands out in your mind that you read" about the case? But the government objected. And the judge sustained the objection.

A day later, with Dzhokhar's mandamus petition still pending, the judge denied Dzhokhar's third venue-change motion — "for reasons both old and new." We focus here on the judge's new reason. Conceding both that "[c]hecking a box" on a questionnaire "may result in answers that appear more clear and unambiguous than the juror may have intended or than is actually true," and that handwritten "answers" frequently "can ... be unclear, inapposite, or incomplete," the judge concluded that the voir dire underway was "successfully identifying potential jurors who are capable of serving" fairly and impartially.

A week after the judge's ruling, #355 was individually voir dired. #355 said that he did not think that his work as a criminal lawyer would affect his impartiality. "[I]f asked to vote on" "the death penalty," he explained, "I would probably vote against it because of my belief that it is overused." But he later added that "[i]f, after hearing the [judge's] instructions, and if I believed it ... fit into one of those rare cases where I believed the death penalty should be imposed, having understood the law as given to me, then ... I could vote to impose the death penalty." Asked by the government whether he could "imagine any case that [he] would think is appropriate for the death penalty," #355 said, "I think Slobodan Milosevic was close, if not a prime example." Asked by the defense whether he could "actually vote to impose" the death penalty in an appropriate case, #355 stated, "I think I could." "Are you pretty confident of that answer?" the defense asked. "Yes," #355 replied.

The government moved to strike #355 "for his bias" as a criminal defense lawyer and "for his death penalty answers." To the government's way of thinking, #355 was "substantially impaired" because "the only time ... he could think that he could impose the death penalty would be in a case of genocide." The defense opposed the motion, pointing out that #355 said he could "make a decision" to impose the death penalty "in a given set of facts." The judge granted the motion, however. "I would not exclude [#355] because of his ... criminal defense work," the judge noted. But relying on his "sense of him," the judge concluded that #355 was not adequately "open to the possibility of the death penalty" — especially given "the genocide issue," which made #355's "zone of possibility ... so narrow" that he was "substantially impaired." All in all, the judge was not convinced that #355 "was going to be truly open in the way that would be necessary."

In the last week of February 2015, the judge provisionally qualified 75 prospective jurors — a group that included #138 and #286. The judge ended up excusing 5 of the 75 for hardship. And it was from this group of 70 that the parties would choose a jury.

That same week, a divided panel of this court denied Dzhokhar's second mandamus petition because he had not shown a clear and indisputable right to a venue change (which is what he had to show to get mandamus relief). See Tsarnaev II, 780 F.3d at 15, 19-20. To give one of the panel majority's reasons: Although Dzhokhar argued that we had to "presume prejudice for any jury drawn from the Eastern Division of Massachusetts," the panel majority found that "his own statistics reveal that hundreds of members of the [jury pool] have not formed an opinion that he is guilty" — and "[t]he voir dire responses have confirmed this." Id. at 21. The panel majority also believed that a rigorous and thorough voir dire would secure an impartial jury. See id. at 21-24.

Motions to Excuse #138 and #286, Fourth Venue Motion, and Peremptory Strikes

On the same day Tsarnaev II came down, Dzhokhar filed motions to strike #138 and #286 for cause from the provisionally qualified jury pool — motions premised on alleged newly discovered information.

In his motion against #138, Dzhokhar claimed that he had just learned that #138 "was dishonest ... about comments on Facebook" and had defied the judge's "instructions" within mere "hours of receiving them." For support, he pointed to the following:

• On the day #138 went to court to complete his juror questionnaire, he posted on Facebook, "Jury duty....this should be interesting ... couple thousand people already here."

• Two of his Facebook "friends" responded. One said, "How'd you get stuck going to Boston?" The other said, "Did you get picked for the marathon bomber trial!!!??? That's awesome!" #138 replied, "Ya awesome alright haha there's like 1000s of people."

• Over the next few hours, people left more comments, saying things like, "If you're really on jury duty, this guys got no shot in hell" and "They're gonna take one look at you and tell you to beat it."

• Despite hearing the judge's preliminary warnings — "not to discuss anything else, or allow anyone else to discuss with you anything else until you have been excused," not to "communicate about this case or allow anyone to communicate about it with you by phone, text message, Skype, email,

social media, such as Twitter or Facebook," etc. — #138 returned to the Facebook thread later that day and posted, "Shud be crazy [Dzhokhar] was legit like ten feet infront of me today with his 5 or 6 team of lawyers ... can't say much else about it tho ... that's against the rules." His Facebook friends responded, "Whoa!!"; "Since when does [#138] care about rules?"; and "Play the part so u get on the jury then send him to jail where he will be taken care of." #138 replied, "When the Feds are involved id rather not take my chances ... them locals tho ... pishhh ain't no thaang." "Yea super careful," a Facebook friend wrote back, "bc should you get picked any mention of anything can get you booted or call for mistrial."

Dzhokhar argued that these actions by #138 showed "a willingness to flout the rules, a lack of maturity, and a failure to appreciate the seriousness of these procedures." So he asked the judge to "excuse[ ]" #138 "for cause."

Dzhokhar's motion against #286 claimed that "[a]fter her voir dire questioning," he discovered information about #286's social media "postings at the time of the Boston Marathon Bombings and their aftermath" that undermines "her juror questionnaire" answers. He emphasized the following:

• On the day of the bombings, #286 tweeted, "Need something to make you smile and warm your heart after today's tragedy at #BostonMarathon, take a look at #BostonHelp." About Martin Richard, she tweeted, "Little 8yr old boy that was killed at marathon, was a Savin Hill little leaguer :-( RIP little man #Dorchester #bostonmarathon." #286 was from Dorchester too (Dorchester is a neighborhood in Boston).

• During the shelter-in-place situation, #286 tweeted that she was "locked down" with her family, adding, "it's worse having to work knowing ur family is locked down at home!! Finally home locked down w/them #boston."

• After Dzhokhar's capture, #286 retweeted expressions of celebration — including a tweet that said, "Told y'all. Welcome To Boston The City Of CHAMPS! We get our shit DONE! #BostonStrong." Another of her retweets said, "Congratulations to all of the law enforcement professionals who worked so hard and went through hell to bring in that piece of garbage." And another of her retweets read, "Monday started in celebration and ended in tragedy. Today began in tragedy and ended in celebration. You can't keep us down. #BostonStrong."

• Over the following year, #286 retweeted additional posts about the victims — including retweets of a photo of Martin Richard's sister singing the national anthem at Fenway Park (home to the Boston Red Sox), a photo of Martin's older brother running the Boston Athletic Association Youth Relay Race in 2014, and photos of officers Sean Collier and Richard Donohue at their police academy graduation (the caption over the photos read, "Please keep both in your prayers").

Dzhokhar contended that #286's tweets and retweets showed "a community allegiance that is certain to color her view of the case," making her claim that she could fairly and impartially "consider life versus the death penalty at trial exceedingly suspect." So he asked the judge to "excuse[ ]" #286 "for cause" or "recall[ ]" her "for follow-up questioning." In a memo opposing Dzhokhar's motions, the government called his challenges to #138 and #286 untimely because he did not object when the judge provisionally qualified them. And that untimeliness, said the government, could not be excused based on newly discovered evidence because the motions relied on "social media postings ... that predated voir dire, often by years." On the merits, the government asserted that #138 did not disobey the judge's instructions by "simply reporting that [Dzhokhar] was ten feet in front of him at one point and had a team of five or six lawyers" (though the government did not address how #138 told the judge during individual voir dire that his Facebook friends were not "commenting" on the "trial"). And as for #286, labeling her tweets and retweets "innocuous," the government argued that she "may well not have considered ‘tweeting’ (or especially ‘retweeting’) a photograph to be the same as ‘comment[ing] on this case in a letter to the editor, in an online comment or post, or on a radio talk show’ " (the government made no mention of her "piece of garbage" retweet or her being "locked down" with her family, however).

The judge orally denied Dzhokhar's motions at a conference the first week of March 2015. "I reviewed the jury questionnaires," the judge said, and the voir dire

transcripts. First of all, I agree with the government that the objections are late and it is — we have a procedure. We have done it with some care and taken the time to do it. And I think the time to raise the issues was in the course of that process and not thereafter. So I am not inclined — and will not — reopen the voir dire for late discovery matters that could have been discovered earlier.

Continuing on, the judge added that he found Dzhokhar's objections "largely speculative." "There are various possible explanations," he said,

and none of them is ... serious enough to warrant changing our provisional qualification, and in particular, none of the issues that were raised seem ... to suggest the presence of a bias that would be harmful to jury impartiality in this case. They're collateral matters about things, they are — people close to them may have done, but none of them speak to actual bias in the case. So we leave the roster as it is.

Around this time — early March 2015 — the defense filed a fourth venue-change motion — essentially arguing that of the 75 provisionally qualified jurors, 42 "self-identified ... some connection to the events, people, and/or places at issue in the case"; 23 "stated ... that they had formed the opinion that [Dzhokhar] 'is' guilty, with ... 1 ... of those ... 23 stating ... that he would be unable to set aside that belief"; and that 48 "either believe that [Dzhokhar] is guilty, or have a self-identified connection, or both." The government opposed, contesting (among other things) the defense's statistical methodology.

While that motion was pending, the defense used all 20 of its peremptory strikes, see Fed. R. Crim. P. 24(b)(1), but did not strike #138 or #286 (the judge denied the defense's request for 10 more peremptories). The government used all of its peremptory challenges too. Both #138 and #286 got on the jury (#286 ultimately served as the jury foreperson). Of the 12 jurors seated by the judge, 9 got there without disclosing the specific content of the media coverage they had seen — recall how the judge rejected the defense's efforts to learn not just whether prospective jurors had seen media coverage of this case but what specifically they had seen. And of those 9, 4 believed based on pretrial publicity that Dzhokhar had participated in the bombings. But all 12 did say that they could adjudicate on the evidence as opposed to personal biases or preconceived notions.

On the first day of trial — also in early March 2015 — the judge orally denied the defense's pending venue-change request, without an on-the-spot explanation.

Basic Appellate Arguments

Dzhokhar presents a raft of venue- and juror-selection claims on appeal.

Starting with venue, Dzhokhar contends that his trial in the district's Eastern Division violated his constitutional right to an impartial jury for either of two reasons: "The community's exposure to the bombings and ensuing pre-trial publicity ... warranted a presumption of prejudice," or "the jurors' questionnaire and voir dire responses establish[ed] actual prejudice." Hoping to counter that claim, the government argues that "[p]rejudice should not be presumed in a venue with a population of almost five million and where more than half of the prospective jurors had either not prejudged guilt or had stated under oath that they could set aside their view that [Dzhokhar] was guilty." And the government also insists that the record nowhere shows that the pretrial publicity "actually biased" the potential or seated jurors.

From there, Dzhokhar argues that both #138 and #286 lied under oath during voir dire about their social-media postings. And, he says, by not striking them for cause, the judge robbed him of his constitutional rights to due process, an impartial jury, and a reliable sentencing decision. The government responds that even if #138 and #286 "had fully disclosed everything" that Dzhokhar says "they should have, they would not have been stricken for cause." Which is why the government thinks that Dzhokhar "cannot show entitlement to a new trial or an evidentiary hearing."

Dzhokhar then argues that the judge stacked the deck against him by excusing #355 based on the mistaken belief that #355's death-penalty views would have "substantially impaired" him in his ability to perform as a juror. Calling him "[a]n educated professional who had devoted" considerable thought to the question, Dzhokhar notes that #355 said that he could vote for the death penalty in the right case — despite his personal views on capital punishment. Conversely, the government contends that #355 "gave hesitant and carefully hedged answers about the death penalty," plus "was unable to think of any category of crimes beyond genocide where he believed the death penalty would be appropriate."

Penultimately (at least for this part of this opinion), Dzhokhar faults the judge for "taking a crabbed view of Morgan." The nub of his complaint is, to quote his brief, that a faithful application of Morgan required the judge to ask prospective jurors "whether they could take into account mitigating evidence and consider a sentence of life imprisonment not just in the abstract, but in light of specific allegations in his case." Unpersuaded, the government sees no legal error under Morgan, because the suggested questions "were impermissible 'stakeout' questions" that basically asked potential jurors "to prejudge the appropriateness of the death penalty in this case without consideration of the [judge's] instructions or mitigating factors." And, the government adds, even if Morgan required the judge to tell prospective jurors about "certain case-specific facts," Dzhokhar's suggested questions were unnecessary because they already knew about the case's key facts from the judge's preliminary instructions and the juror questionnaire — "and they could have considered those facts when answering questions about their views on the death penalty."

Relying on Patriarca and its offspring, Dzhokhar lastly argues here that the judge erred in denying his request to ask potential jurors content-specific questions about "what they had seen, read, or heard about his case." The pretrial publicity, he writes (quoting Patriarca ), created a " ‘significant possibility that jurors [had] been exposed to potentially prejudicial material’ " and so "trigger[ed]" a "duty to inquire." Which, according to him, means that the judge had to ask "not just whether prospective jurors had seen media coverage of this case, but what , specifically, they had seen." And by not doing so, the judge (in Dzhokhar's words) produced "a jury biased by prejudicial publicity." Trying to meet this argument, the government — citing Mu'Min v. Virginia, 500 U.S. 415, 111 S.Ct. 1899, 114 L.Ed.2d 493 (1991) — principally contends that Supreme Court precedent "reject[s] the claim that such an inquiry is required."

Analysis

We start, as the parties do, with the judge's decision not to change venue — a decision that receives abuse-of-discretion review. See United States v. Casellas-Toro, 807 F.3d 380, 385 (1st Cir. 2015). Anyone alleging an abuse of discretion faces an uphill climb. See generally United States v. Rivera-Carrasquillo, 933 F.3d 33, 44 (1st Cir. 2019) (explaining that a judge abuses his discretion "if no reasonable person could agree with the ruling"), cert. denied, ––– U.S. ––––, 140 S.Ct. 2691, 206 L.Ed.2d 835 (2020). And that is certainly true when a party asks us to critique a denial of a motion grounded on alleged jury partiality, because — as no less an authority than the Supreme Court has said — "[i]n reviewing claims of this type, the deference due to district courts is at its pinnacle ." Skilling, 561 U.S. at 396, 130 S.Ct. 2896 (emphasis added).

Two of us find serious points against Dzhokhar's venue-change arguments.

First, the polling data shows that many in Boston were undecided about whether Dzhokhar should receive death — even after all the publicity. The defense expert's own survey data revealed that only 36.7% of people in Boston favored the death penalty for Dzhokhar before the trial, leaving 63.3% undecided or leaning against death. Since the Eastern Division has a population of about five million, that leaves several million people (even minus children, etc. ) open to a life sentence despite all the publicity. The data from the voir-dire questionnaire is even more telling, showing that only 23% of the prospective-juror panel had formed an opinion that Dzhokhar should die — and just 16% held that view and said they could not be convinced otherwise. This means that of the 1,373 potential jurors, over 1,000 of them had not predetermined that death was the right sentence.

Second, the same polling data shows public awareness and attitudes were not materially different in, for example, Springfield or New York City. In Springfield, 51.7% said Dzhokhar was "definitely guilty" and 32.2% said he was "probably guilty" based on the pretrial publicity, compared with 47.6% and 44% in Manhattan, and 57.8% and 34.5% in Boston. The numbers regarding penalty are also similar — in fact fewer respondents preferred life without the possibility of parole in Springfield (45.4%) than in Boston (51.2%).

Also and importantly, the polling inquiry does not ask respondents to judge for themselves whether they are biased. Nor does it fail to overlook subliminal biases. Instead, the pollster asked whether Dzhokhar should get the death penalty. The answer to that question likely reflected whatever actual bias might have been operating, knowingly or otherwise. So (extrapolating from the relevant figures) the fact that so many hundreds of thousands in the pool of potential jurors were undecided, and that the percentage of those persons was not materially less than the percentage for New York City, does support the judge's venue decision in a powerful way.

We note too that this is not a case where almost everybody locally knows something and very few elsewhere know of it. Plus the data seemingly contradicts any claim that the Boston Strong movement and the sheltering in place account for undue prejudice — were it otherwise, the difference in attitudes would probably be much greater in, for example, New York City.

Third, most of the publicity was true — something we now know from Dzhokhar's guilt admission in his lawyer's guilt-phase opening and closing statements (notably, Dzhokhar does not say he would have raised an innocence defense in another venue). See Murphy v. Florida, 421 U.S. 794, 802, 95 S.Ct. 2031, 44 L.Ed.2d 589 (1975) (noting the truth of the pretrial publicity); see also Skilling, 561 U.S. at 382-83, 130 S.Ct. 2896 ; Casellas-Toro, 807 F.3d at 387. So after the first day of trial, a juror from Boston and one from California would know essentially the same things about the case — even though the California juror would have seen less of the publicity. Contrast that with Rideau v. Louisiana, 373 U.S. 723, 83 S.Ct. 1417, 10 L.Ed.2d 663 (1963), where an inadmissible taped confession by a guilt-contesting defendant was televised only on the local news. As for the untrue pieces of publicity, they seem trivial given the true and relevant information — for example, the report that Dzhokhar's boat message said "fuck America" got quickly disproved once jurors saw the actual message; and at any rate, the words "fuck America" added very little if anything to what he actually wrote.

Fourth and finally, comparing Dzhokhar's case to other venue-change cases — specifically Skilling and Casellas-Toro — makes it hard to say the judge abused his discretion. Starting with the size and diversity of the metropolitan area (the first Skilling factor), Boston is very large and diverse — much closer to Houston in Skilling, 561 U.S. at 382, 130 S.Ct. 2896 (a "large, diverse pool" of 4.5 million eligible jurors in the area made the "suggestion that 12 impartial individuals could not be empaneled ... hard to sustain"), than to the Indiana community in Irvin v. Dowd, 366 U.S. 717, 719, 81 S.Ct. 1639, 6 L.Ed.2d 751 (1961) (about 30,000 people), or to Puerto Rico in Casellas-Toro, 807 F.3d at 386 ("a compact, insular community" of 3 million people that "is highly susceptible to the impact of local media" (quotation marks omitted)). The nature of the publicity (the second Skilling factor) was, as discussed, largely factual and the untrue stuff was no more inflammatory than the evidence presented at trial. As to passage of time (the third Skilling factor), two years had elapsed between the crime and the trial — which is closer in magnitude to the four years in Skilling (a point cutting against a venue change) than the two months in Casellas-Toro (a point favoring a venue change). And the verdict shows no affirmative signs of bias (the fourth Skilling factor), since the jury recommended Dzhokhar die for six of seventeen death-eligible counts, similar to Skilling's acquittal on certain counts and contrasted with Casellas-Toro's guilty verdict on counts lacking sufficient evidence. Dzhokhar asks us to give no weight to the jury's decision not to recommend death on most of the death-eligible counts. He reasons that the jury's decision-making can be explained by the fact that the government can only kill him once. But the jurors had seventeen death counts to consider. And they decided on death for six of those. That selectivity cannot be explained (as he presumes) by his the-government-can-only-execute-me-once theory. For if he were right, the jury would have returned a death verdict on just one count. To pick six and not one or seventeen, the jurors must have had some other rationale in mind that required them to draw distinctions between facts that they thought warranted a death verdict and those that they thought did not. And that sort of nuance favors a view of the jury as intent on following the law and the facts.

Dzhokhar's chief argument — that the nature of the crime (terrorism) might be viewed as an attack on the Boston community specifically — does not appear in the framework of these cases. But it seems just as likely that a juror in, say, New York City would view the crime as an attack on all of America (as he himself did) — thus negating any advantage in changing venues.

So if pressed to decide the venue question now, two of us would likely find the judge abused no discretion in finding venue proper in Boston in 2015. But we need not make such a decision now. That is because, as explained next, we must remand the penalty-phase portion of this trial for a retrial — regardless of how we rule on venue. And it also is because, given the sizable passage of time, the venue issue should look quite different the second time around, likely in 2021. See Skilling, 561 U.S. at 383, 130 S.Ct. 2896.

Now for our remand-for-a-penalty-phase-retrial explanation. Even assuming (favorably to the government) that the judge did not reversibly err on the venue question, he still had to oversee a voir-dire process capable of winnowing out partial jurors through careful questioning — indeed, in denying Dzhokhar a venue change, the judge premised his analysis in part on a pledge to run a "voir dire sufficient to identify prejudice." But performance fell short of promise, providing (as Dzhokhar's counsel said at oral argument) a sufficient ground to vacate his death sentences — even on abuse-of-discretion review. See United States v. Casanova, 886 F.3d 55, 60 (1st Cir. 2018). See generally United States v. Connolly, 504 F.3d 206, 211-12 (1st Cir. 2007) (noting that "an erroneous view of the law" is always an abuse of discretion). With the venue assumption in place, we lay out our reasoning.

Patriarca is the key. A pretrial-publicity case, Patriarca involved an organized-crime prosecution where the press called one of the defendants " ‘Boss’ of the New England ‘Cosa Nostra’ " and reported how a lawyer for a government witness nearly died in a car-bomb incident. See 402 F.2d at 315-16. Convinced that the news accounts might make prospective jurors think (wrongly, apparently) that the defendants had something to do with the bombing, the defense teams moved to change the trial's venue — and lost. Id. at 316-17.

The defendants appealed, relevantly arguing that the judge "erred in denying" the change-of-venue motion "because of prejudicial publicity." Id. at 315. We noted "that the amount of coverage diminished sharply after the week following the bombing." Id. at 317. We also noted that the defense had the chance "to mitigate any possible effect of pretrial publicity — [namely,] on the voir dire." Id. Counsel for one of the defendants had asked the judge to "ask a question of the jury in connection with this case, in the light of all the publicity." Id. at 317-18. And the judge said that he would ask the jurors "if there is any member ... who feels that he would not be able to give the defendants a fair and impartial jury." Id. at 318. Counsel said "thank you." Id. The judge put the question to the jury, got "[n]o response" from the members, and so saw no reason not to proceed to trial. See id. Given this set of circumstances, we found no sign of abused discretion in the judge's venue decision. Id.

But crucially, we felt "bound" to address "sua sponte" — i.e. , without prompting from either side — the scope of voir dire judges should conduct "[i]n cases where there is, in the opinion of the [judge], a significant possibility that jurors have been exposed to potentially prejudicial material." Id. Specifically, we directed that

on request of counsel, ... the [judge] should proceed to examine each prospective juror apart from other jurors and prospective jurors, with a view to eliciting the kind and degree of his exposure to the case or the parties , the effect of such exposure on his present state of mind, and the extent to which such state of mind is immutable or subject to change from evidence.

Id. (double emphasis added).

And in driving this directive home, we explicitly endorsed section 3.4 of the American Bar Association's then-recent Standards Relating to Fair Trial and Free Press. See id. (emphasizing that "we are in accord with the suggestions of section 3.4"). Section 3.4, in turn, said that in cases involving prejudicial pretrial publicity, voir-dire "questioning shall be conducted for the purpose of determining what the prospective juror has read and heard about the case ." See Am. Bar Ass'n, Standards Relating to Fair Trial and Free Press § 3.4(a), at 130 (Tentative Draft Dec. 1966) (emphasis added).

The rationale for the Patriarca standard is obvious. Decisions about prospective jurors' impartiality are for the judge, not for the potential jurors themselves. See, e.g., United States v. Rhodes, 556 F.2d 599, 601 (1st Cir. 1977). And that is because prospective jurors "may have an interest in concealing [their] own bias" or "may be unaware of it." Smith v. Phillips, 455 U.S. 209, 221-22, 102 S.Ct. 940, 71 L.Ed.2d 78 (1982) (O'Connor, J., concurring); see also Sampson v. United States, 724 F.3d 150, 164 (1st Cir. 2013) (" Sampson II") (emphasizing that "a person who harbors a bias may not appreciate it and, in any event, may be reluctant to admit her lack of objectivity"). So asking them only "whether they had read anything that might influence their opinion" does not suffice, for that question "in no way elicit[s] what, if anything," they have "learned, but let[s] [them] decide for themselves the ultimate question whether what they [have] learned had prejudiced them." Rhodes, 556 F.2d at 601.

With these principles in mind, we have held that a judge in a high-profile case "fully complied with" Patriarca by asking potential jurors if they "had read or heard anything about the case in the newspapers, on television[,] or radio" — and if so, by "prob[ing] further as to the extent of such knowledge." See United States v. Medina, 761 F.2d 12, 20 (1st Cir. 1985) (emphasis added). We have also found "no inconsistency" with Patriarca when a judge in another high-profile case "asked the prospective jurors, collectively," if they "had heard 'anything at all' about the case" — and then asked those who had "to recount" at side bar "all that [they] knew about the case." See Vest, 842 F.2d at 1332 (emphasis added). And we have held that a judge in yet another high-profile case satisfied Patriarca when he asked potential jurors if they "had seen or read anything about the case" — and then asked those who had about "the circumstances under which [they] had been exposed to publicity." See United States v. Orlando-Figueroa, 229 F.3d 33, 43 (1st Cir. 2000) (emphasis added).

Despite his best intentions, Dzhokhar's judge did not meet the Patriarca standard, however — even though the case met Patriarca's conditions for requiring extensive inquiry. Dzhokhar, do not forget, "request[ed]" voir dire on the contents of the material that the potential jurors had seen. See Patriarca, 402 F.2d at 318. And there was "a significant possibility" that the prospective jurors had been "exposed to potentially prejudicial material." See id. Again, the pervasive coverage of the bombings and the aftermath featured bone-chilling still shots and videos of the Tsarnaev brothers carrying backpacks at the Marathon, of the maimed and the dead near the Marathon's finish line, and of a bloodied Dzhokhar arrested in Watertown (to name just a few). Also, while the media (social, cable, internet, etc. ) gave largely factual accounts, see Tsarnaev II, 780 F.3d at 21-22, some of the coverage included inaccurate or inadmissible information — like the details of his un- Miranda-ized hospital interview and the opinions of public officials that he should die.

With Patriarca's prerequisites satisfied, the judge had to ascertain not just the "degree" but the "kind" of "exposure to the case or the parties" that the prospective jurors had experienced, see 402 F.2d at 318 — that is, "what [they] ha[d] read and heard about the case," see Am. Bar Ass'n, Standards Relating to Fair Trial and Free Press § 3.4(a), at 130 (cited in Patriarca, 402 F.2d at 318 ). But as to 9 of the 12 seated jurors, the judge fell short on this front. To repeat what we wrote earlier, the judge qualified jurors who had already formed an opinion that Dzhokhar was guilty — and he did so in large part because they answered "yes" to the question whether they could decide this high-profile case based on the evidence. The defense warned the judge that asking only general questions like that would wrongly "make[ ]" the potential jurors "judge[s] of their own impartiality" — the exact error that the Patriarca line of cases seeks to prevent. But the judge dismissed the defense's objection, saying that "[t]o a large extent" jurors must perform that function. Yet by not having the jurors identify what it was they already thought they knew about the case, the judge made it too difficult for himself and the parties to determine both the nature of any taint (e.g. , whether the juror knew something prejudicial not to be conceded at trial) and the possible remedies for the taint. This was an error of law and so an abuse of discretion. See Connolly, 504 F.3d at 211-12 ; see also Mangual v. Rotger-Sabat, 317 F.3d 45, 61 (1st Cir. 2003) (echoing the truism that "[i]t is an abuse of discretion for the district court to apply an erroneous standard of law").

The government offers a number of arguments to the contrary. But none of them changes the result.

The government first argues that the Patriarca language we bank on is "dicta." True, the pertinent appellate claim there concerned a venue-change denial. See 402 F.2d at 315. But after rejecting that claim, we "fe[lt] bound " to address the sufficiency of the voir dire — which we did by stating that in high-profile cases, district judges "should proceed to examine each prospective juror ... with a view to eliciting the kind and degree of his exposure to the case." Id. at 318 (emphasis added). And as a later case confirms, Patriarca intended to and does state "the standards of this circuit." See Medina, 761 F.2d at 20 (emphasis added). So the government's dicta argument does not work.

Nor does the government's suggestion that the voir dire here actually "elicit[ed] the kind and degree" of the potential jurors' exposure to the case. In making this claim, the government (paraphrasing the questionnaire) notes that prospective jurors had to disclose "what newspapers, radio programs, and television programs [they] viewed and with what frequency, as well as how much media coverage [they] had seen about the case." And that suffices, the government says, because we have not read Patriarca to require content-specific questioning. But this is wrong for several reasons. For one thing, learning that prospective jurors read, say, the Boston Globe daily and have seen a lot of coverage about the case is not the same as learning that they read Globe articles quoting civic leaders saying Dzhokhar should die —statements that could not constitutionally be admitted into evidence. See Bosse v. Oklahoma, ––– U.S. ––––, 137 S. Ct. 1, 2, 196 L.Ed.2d 1 (2016) (per curiam). For another thing, the government's rejoinder rests on a misreading of Patriarca — an opinion that does require inquiry into what information potential jurors have been exposed to. Again, Patriarca endorsed the ABA's standards calling for content-specific questioning "for the purpose of determining what the prospective juror has read and heard about the case." Also and critically, post- Patriarca caselaw clarified that the defect Patriarca aimed to cure was delegating to prospective jurors the job of evaluating their impartiality — a defect that content-specific questioning can fix. See Vest, 842 F.2d at 1332. Consider Vest. Following the correct approach, the district judge there did not ask potential jurors "to decide for themselves the 'ultimate question' of impartiality" — instead, "once a juror admitted to any knowledge of the case," the judge "individually questioned" him or her "as to the facts and extent of such knowledge." Id. And contrary to the government's characterization, Vest concerned not just individual versus group voir dire, but also content-specific versus non-content-specific questioning.

Quoting Mu'Min, the government then makes its biggest argument — namely, that this post- Patriarca opinion by the Supreme Court (emphasis ours) "rejected the argument that the Constitution requires [judges] to question prospective jurors ‘about the specific contents of the news reports to which they had been exposed.’ " But there is a major flaw in the government's theory. Mu'Min arose on direct review of a state-court criminal conviction — which meant the Supreme Court's "authority" was "limited to enforcing the commands of the [federal] Constitution ." 500 U.S. at 422, 111 S.Ct. 1899 (emphasis added). Dzhokhar, contrastingly, was "tried in federal court[ ]" — and thus was "subject to" the "supervisory power " of the federal appellate courts. See id. (emphasis added). And this distinction makes all the difference, because "[w]e enjoy more latitude in setting standards for voir dire in federal courts under our supervisory power than we have in interpreting" the federal Constitution "with respect to voir dire in state courts." See id. at 424, 111 S.Ct. 1899 (italics omitted); see also Kater v. Maloney, 459 F.3d 56, 66 n.9 (1st Cir. 2006) (noting that Mu'Min "carefully distinguished between constitutional requirements which states must meet and the exercise of its broader supervisory authority over cases tried in federal courts").

Patriarca did not say that the endorsed standard sprang from the Constitution. And neither did Patriarca explicitly say that it emanated from our supervisory powers — yet we see plenty of signs that it did indeed emanate from that source. For starters, neither side in Patriarca made voir dire an issue. And it is highly unlikely that we would have engaged in a constitutional excursion without prompting by the parties. Also and relatedly, given the well-entrenched doctrine of constitutional avoidance, it is equally unlikely that we would have gone out of our way to issue a constitutional decision. See Am. Foreign Serv. Ass'n v. Garfinkel, 490 U.S. 153, 161, 109 S.Ct. 1693, 104 L.Ed.2d 139 (1989) (per curiam) (explaining that the doctrine counsels against issuing "unnecessary constitutional rulings"). Plus as we have noted, Patriarca relied on the ABA standards. And those standards are meant as templates for courts, not as constitutional pronouncements. Cf. Br. of Am. Bar Ass'n as Amicus Curiae at 2, Martinez v. Ryan, 566 U.S. 1, 132 S.Ct. 1309, 182 L.Ed.2d 272 (2011) (No. 10-1001), 2011 WL 3584754, at *2 (explaining that the ABA Standards for Criminal Justice "represent a collection of 'best practices' based on the consensus views of a broad array of professionals involved in the criminal justice system"). Mu'Min itself recognized that other federal appellate courts have mandated content-specific questioning in the exercise of their discretionary supervisory powers, not as a matter of constitutional law. See 500 U.S. at 427, 111 S.Ct. 1899. One of those courts — the Fifth Circuit — later specifically said that " Mu'Min does not abrogate" that court's earlier holding "that, where pretrial publicity creates a significant possibility of prejudice, the district court must make an independent determination of the impartiality of jurors." United States v. Beckner, 69 F.3d 1290, 1292 n.1 (5th Cir. 1995) (discussing Davis ).

The government also defends some (not all) of the judge's reasons for declining to ask content-specific questions. But concerns about "unmanageable data" from content-specific questions — in a case where 1,373 prospective jurors each completed a 100-question questionnaire and the judge designated 21 days for voir dire — seem misplaced. So too does any fear that content-specific questioning could accidentally create bias where none existed. If potential jurors recall a particular piece of reporting well enough to bring it up at voir dire, and the reporting is prejudicial, then potential bias was already present. Far from "reinforc[ing] potentially prejudicial information," content-specific questioning would have brought such material front and center. The parties and the judge could then assess the publicity's effect on the prospective jurors' ability to reach a fair verdict, thus putting the judge in a position to take any necessary measures to protect Dzhokhar's fair-trial rights.

Patriarca was a noncapital case, unlike Dzhokhar's. And the pretrial publicity in Patriarca pales in comparison to the pretrial publicity surrounding Dzhokhar's case. Surely then, with his life at stake, Dzhokhar deserved the type of voir dire that Patriarca calls for. See generally Sampson II, 724 F.3d at 159-60 (suggesting that protections are generally heightened in capital cases, because death is different from other kinds of penalties).

In denying Dzhokhar mandamus relief on the venue-change issue, it is reasonable to infer that the mandamus panels reasonably expected that the judge would conduct the kind of searching voir-dire inquiry required by our caselaw. But regrettably, we conclude that his efforts fell short for the reasons just stated. Dzhokhar's appellate counsel admitted at oral argument that this error was harmless at the guilt stage, given his trial concession (through his trial lawyer) that he had done what the government accused him of doing. The government suggests that "any error was harmless" at the penalty stage too, because prospective jurors said that they could serve impartially. Not only does the government push a theory that clashes with our caselaw — caselaw that again says that in a situation like Dzhokhar's, a judge cannot delegate to potential jurors the work of judging their own impartiality. But the government never explains how its flawed argument proves the error's harmlessness beyond a reasonable doubt. See 18 U.S.C. § 3595(c)(2)(C) (providing that "[t]he court of appeals shall not reverse or vacate a sentence of death on account of any error which can be harmless" if "the [g]overnment establishes beyond a reasonable doubt that the error was harmless"). So the government's harmless-error theory is a nonstarter.

Given our ruling that the judge's Patriarca-based error demands vacatur of the death sentences, the remaining issues listed above (in the "Basic Appellate Arguments" section) require only the briefest discussion — but discussion nevertheless, at least on matters that may arise again on remand.

Regarding Dzhokhar's claim that #138 and #286 lied during voir dire, we repeat a point made in our caselaw again and again (and again) because it is so very important to our system of justice: If a defendant "com[es] forward" at any point in the litigation process "with a ‘colorable or plausible’ " juror-misconduct claim, "an ‘unflagging duty’ falls to the district court to investigate the claim." United States v. French, 904 F.3d 111, 117 (1st Cir. 2018) (quoting United States v. Zimny, 846 F.3d 458, 464 (1st Cir. 2017) ), cert. denied sub nom. Russell v. United States, ––– U.S. ––––, 139 S. Ct. 949, 203 L.Ed.2d 137 (2019). See generally Sampson II, 724 F.3d at 169 (stressing that "[j]urors who do not take their oaths seriously threaten the very integrity of the judicial process"). But our decision on the content-specific-questioning issue makes it unnecessary to address the misconduct charge.

And that same ruling also makes it unnecessary to touch on Dzhokhar's argument that the judge wrongly excused #355 for his views on the death penalty.

That leaves us with Dzhokhar's claim that the judge kept him from asking case-specific death-penalty questions critical to seating an impartial jury — questions like whether potential jurors could consider mitigating circumstances on the ultimate life-or-death issue, given "the specific allegations in his case: the killing of multiple victims, one of them a child, in a premediated act of terrorism." On this issue — which also gets abuse-of-discretion review, see Casanova, 886 F.3d at 60 — it is enough for us to say that even assuming without deciding that the judge had to inform prospective jurors of certain case-specific facts, he did do so here. Recall the judge's preliminary instructions to prospective jurors before they filled out the questionnaires — that Dzhokhar was "charged in connection with events that occurred near the finish line of the Boston Marathon ... that resulted in the deaths of three people." Recall too the questionnaires' "summary of the facts of this case" — that "two bombs exploded ... near the Boston Marathon finish line[,] ... kill[ing] Krystle Marie Campbell (29), Lingzi Lu (23), and Martin Richard (8), and injured hundreds of others"; that "MIT Police Officer Sean Collier (26) was shot to death in his police car"; and that Dzhokhar "has been charged with various crimes arising out of these events." And like the judge, we think that because potential jurors knew these details, the voir dire adequately covered Dzhokhar's case-specific questions.

Having said all that needs saying on these subjects, we press on — for the reader's information, everything from here on out until we reach the crime-of-violence issue also falls within the category of issues that could easily reappear on remand.

Mitigation Evidence About Tamerlan's Possible Homicidal Past

We shift our focus to Dzhokhar's claim that the judge damaged the defense's mitigation case by barring evidence tying Tamerlan to a triple murder in 2011, and by keeping the defense from seeing a confession Tamerlan's friend made to the FBI about how he and Tamerlan had committed those crimes.

Background

On September 11, 2011 — the tenth anniversary of the 9/11 terrorist attacks — someone (or a number of someones) robbed and killed three drug dealers in an apartment in Waltham, Massachusetts. All three were found bound, with their throats slit. These crimes remain unsolved to this day.

Fast forward to 2013. Soon after the Marathon bombings, federal and state law-enforcement officers interviewed Tamerlan's friend, Ibragim Todashev — a mixed-martial-arts fighter who had come to the United States from Chechnya in 2008 and met Tamerlan shortly afterwards. Officers interviewed Todashev, then living in Florida, four separate times in April and May. The first two interviews focused on his relationship with Tamerlan and his possible knowledge of the bombings. At some point, agents began suspecting that Todashev had a hand in the 2011 murders. During the final interview on May 21, Todashev said he knew something about the murders and asked if he could get a deal for cooperating.

After waiving his Miranda rights, Todashev gave the following account. Tamerlan recruited Todashev to rob the men. They drove to a Waltham apartment, held the men at gunpoint (with a gun Tamerlan had brought), beat them, and bound them with duct tape. Not wanting to leave any witnesses, Tamerlan cut each man's throat while Todashev waited outside (Todashev did not want any part of the throat cutting, apparently). Tamerlan then waved Todashev back in to help remove all traces of evidence.

Todashev agreed to write out a confession. But as he was doing so, he attacked the agents — one of whom shot and killed him. The FBI documented Todashev's statements in memos known as 302 reports. And a state trooper recorded most of his statements at his final interview. The Florida attorney general's office investigated the circumstances of Todashev's death and found the agent had acted reasonably in using deadly force.

Dzhokhar's lawyers repeatedly asked the judge pretrial to make the government produce all reports and recordings of Todashev's statements about the Waltham crimes, either directly to them or to the judge for an in-camera inspection. The government opposed each of the defense's motions, arguing that the sought-after materials were not discoverable under the federal or local criminal rules or under Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963), and were protected by the law enforcement investigatory privilege. In the government's telling, because prosecutors had informed the defense that Tamerlan had "participated in the Waltham triple homicide," it did not have to disclose the actual reports and recordings. After inspecting some of the items in camera, the judge refused to disclose any of the materials documenting Todashev's statements. Agreeing with the government that prosecutors had "conveyed the fact and general substance of Todashev's statements," the judge said that the FBI's 302 report of Todashev's final interview did "not materially advance [the mitigation] theory beyond what is already available to the defense." The judge also said that disclosing Todashev's statements "risk[ed] revealing facts seemingly innocuous on their face, such as times of day or sequences of events," that "would have a real potential to interfere with the ongoing state investigation."

While all this was going on, a lawyer representing Dzhokhar's college friend Dias Kadyrbayev — who faced prosecution for hiding Dzhokhar's backpack and computer — told the government that his client "may be able to provide" some information, including that "Kadyrbayev learned in the fall of 2012 from Dzhokhar ... that Tamerlan ... was involved in the Waltham murders" and that "Dzhokhar ... told Kadyrbayev that [Tamerlan] 'had committed jihad' in Waltham." The government disclosed Kadyrbayev's lawyer's proffer to Dzhokhar's counsel.

Anyway, because of the judge's rulings, the defense never learned key details about the murders (as disclosed by Todashev) — including:

• Tamerlan brought the "tools" he and Todashev used to commit the crimes (a gun, knives, duct tape, cleaning supplies).

• Tamerlan and Todashev got into the apartment because Tamerlan knew one of the victims, Brendan Mess — Tamerlan and Mess were close childhood friends.

• Tamerlan had Todashev duct tape one of the victim's hands and feet. And Tamerlan duct taped the others.

• Tamerlan beat Mess to try to get him to say where more money was in the apartment.

• Todashev had agreed with Tamerlan to rob the men. But after they had bound and robbed them, Tamerlan decided to kill the men — a decision that made Todashev shake with nerves, because while he did not want to participate in the murders, he felt he "had to" since he "did not have a way out."

• Tamerlan slashed each man's throat.

• Tamerlan gave Todashev $20,000 from the money they had stolen.

The government later moved in limine to bar Dzhokhar from introducing any evidence about the Waltham murders at the guilt or penalty phases. Among other theories, the government called Todashev's statements about Tamerlan's role "unreliable" since he had an obvious motive to pin the murders on someone else (what the government did not tell the judge, however, was that agents had previously relied on Todashev's statements in applying for a search warrant to look for evidence from the Waltham homicides in Tamerlan's car). The government also argued that, apart from Todashev's statements, it had no "evidence that Todashev and/or Tamerlan ... actually participated in the Waltham triple homicides." The government further claimed that Todashev's statements should not come in because he "cannot be cross examined," because he "obviously was not of sound mind" since he attacked armed agents, and because admitting this evidence would confuse the jurors by opening the door to "a great deal of information having nothing to do with" Dzhokhar's crimes. And the government claimed that "[t]here's no evidence that the defense can point to anywhere, including ... Todashev's own statement, that Tamerlan ... controlled him in any way."

Dzhokhar's lawyers argued in opposition that evidence showing Tamerlan's having committed the crimes was highly probative of the brothers' respective roles in the bombings and was sufficiently reliable to be admitted under the evidentiary standards applicable at the penalty phase. They also stressed that whether to credit Todashev's statements was for the jury.

The judge orally granted the government's in-limine motion, finding that "there simply is insufficient evidence to describe what participation Tamerlan may have had" in the Waltham murders. From his check of the evidence — which "include[d] an in-camera review of some Todashev 302s," but not the recordings of the confession — the judge thought that "it [was] as plausible ... that Todashev was the bad guy and Tamerlan was the minor actor." So the judge concluded that the murder evidence "would be confusing to the jury and a waste of time, ... without any probative value."

Dzhokhar's mitigation theory portrayed him as influenced by Tamerlan to take part in the Marathon bombings. "[I]f not for Tamerlan," said his lawyer to the penalty-phase jury, "this wouldn't have happened." And the defense sought to prove several mitigating factors about their relationship and their relative culpability — including:

• Dzhokhar "acted under the influence of his older brother";

• "because of Tamerlan's age, size, aggressiveness, domineering personality, [and] privileged status in the family," Dzhokhar "was particularly susceptible to his ... influence";

• "Dzhokhar['s] ... brother Tamerlan planned, led, and directed the Marathon bombing[s]";

• "Dzhokhar ... would not have committed the crimes but for his older brother Tamerlan"; and

• "Tamerlan ... became radicalized first, and then encouraged his younger brother to follow him."

Without the Waltham evidence, the defense supported its mitigation theory with testimony like:

• Tamerlan became radicalized first, began proselytizing his views, and sent jihadi materials to Dzhokhar;

• the oldest brother in a Chechen family like the Tsarnaevs usually receives deference (an associate professor from Princeton University testified that "it's expected that the younger brothers will listen to the older brother");

• Tamerlan occasionally broke the rules of the gym he belonged to (he used

other members' equipment without asking, for instance);

• Tamerlan sometimes got argumentative at a mosque (for example, he twice called the Imam a "hypocrite");

• Tamerlan yelled at a store owner for selling halal turkey for Thanksgiving (halal is a term associated with Islamic dietary laws); and

• Tamerlan once might have physically abused his then girlfriend (he later married her).

Conversely, the government tried to convince the jury that Dzhokhar should die because he and Tamerlan were equally culpable in the bombings and that Tamerlan had played no role in Dzhokhar's decision to participate. During the penalty phase, the government argued that the defense's mitigation evidence consisted of little more than "testimony that Tamerlan was bossy." The government also described Tamerlan as a "handsome," "charming," "loud" guy who "sometimes lost his temper." And the government implored the jurors to "ask [themselves] if there's anything about Tamerlan ... that will explain ... how Dzhokhar ... could take a bomb, leave it behind a row of children, walk ... down the street, and detonate it." Insisting that no evidence supported the notion that Tamerlan had "coerced or controlled" Dzhokhar, the government labeled the brothers "a partnership of equals" and so "bear the same moral culpability for what they did."

Basic Appellate Arguments

Dzhokhar presents essentially two arguments about the judge's handling of the Waltham evidence. The first claim is that the judge violated his right to present a complete mitigation defense by keeping from the jury major proof of Tamerlan's brutal past, his ability to enlist others in acts of extreme cruelty, and thus his relative culpability — an error the government exploited by distorting Tamerlan's character and suggesting no evidence showed his influence over Dzhokhar. The second claim is that the judge violated his Brady rights by refusing to give the defense a 302 report and recordings of Todashev's confession — evidence that, "if presented," would have shown "why Tamerlan was to be feared, and his ability to influence others to commit horrific crimes."

The government takes a different view of the matter. According to the government, the Waltham evidence was not relevant mitigation evidence because nothing suggests Tamerlan's alleged commission of the Waltham crimes had any link to Dzhokhar's commission of the crimes here. And, says the government, even if the Waltham evidence had some slight relevance, the judge rightly excluded it because the risks of confusing the issues and misleading the jury outweighed any probative worth. The government also thinks that any error by the judge was harmless beyond a reasonable doubt because "overwhelming[ ]" evidence (the government's word) showed Dzhokhar willingly engaged in the crimes charged here. Wrapping up, the government says that the undisclosed "information was not discoverable under Brady" and "was subject to the law enforcement privilege."

Analysis

We give abuse-of-discretion review to preserved disputes over whether the judge wrongly excluded mitigating evidence at the penalty phase, showing "great deference" to his balancing of the evidence's probative worth against its possible prejudice. See United States v. Sampson, 486 F.3d 13, 42 (1st Cir. 2007) (" Sampson I"). We also give abuse-of-discretion review to preserved disputes over whether the judge wrongly kept Brady material from the defense. See United States v. Bulger, 816 F.3d 137, 153 (1st Cir. 2016).

With these preliminaries out of the way, we turn to Dzhokhar's first claim: that the judge committed prejudicial error by keeping the Waltham evidence from the jury at the penalty phase.

Because it is "desirable for the jury to have as much information before it as possible when it makes the sentencing decision," the Supreme Court has for years said that if "the evidence introduced and the arguments made ... do not prejudice a defendant, it is preferable not to impose restrictions." Gregg v. Georgia, 428 U.S. 153, 203-04, 96 S.Ct. 2909, 49 L.Ed.2d 859 (1976). So a defendant convicted of capital crime has a constitutional right to put before the jury, "as a mitigating factor , any aspect of [his] character or record and any of the circumstances of the offense that [he] proffers as a basis for a sentence less than death." Lockett v. Ohio, 438 U.S. 586, 604, 98 S.Ct. 2954, 57 L.Ed.2d 973 (1978) (plurality opinion); see also Eddings v. Oklahoma, 455 U.S. 104, 110, 102 S.Ct. 869, 71 L.Ed.2d 1 (1982) (adopting the rule announced by the Lockett plurality). Mitigating factors include aspects of "the defendant's background, record, or character or any other circumstance of the offense that mitigate against imposition of the death sentence," see 18 U.S.C. § 3592(a)(8) — like, for instance, information bearing on the extent and nature of each defendant's role in the charged crime, see Green v. Georgia, 442 U.S. 95, 97, 99 S.Ct. 2150, 60 L.Ed.2d 738 (1979) (finding a constitutional violation where the judge excluded penalty-phase evidence showing a codefendant's primary role).

This standard is broad, reflecting the idea "that punishment should be directly related to the personal culpability of the criminal defendant." See Penry v. Lynaugh, 492 U.S. 302, 319, 109 S.Ct. 2934, 106 L.Ed.2d 256 (1989), abrogated on other grounds by Atkins v. Virginia, 536 U.S. 304, 122 S.Ct. 2242, 153 L.Ed.2d 335 (2002) ; see also Enmund v. Florida, 458 U.S. 782, 801, 102 S.Ct. 3368, 73 L.Ed.2d 1140 (1982) (emphasizing that "punishment must be tailored to [a defendant's] personal responsibility and moral guilt"). And consistent with this lenient approach, mitigating information need not be admissible under the rules of evidence to get in. See 18 U.S.C. § 3593(c). All of which is why the normally low relevance threshold in noncapital cases is lower still when it comes to mitigation evidence in capital cases: Relevant "mitigating evidence" encompasses any "evidence which tends logically to prove or disprove some fact or circumstance which a fact-finder could reasonably deem to have mitigating value." See Smith v. Texas, 543 U.S. 37, 44, 125 S.Ct. 400, 160 L.Ed.2d 303 (2004). Once this modest "threshold ... is met," the Constitution " ‘requires that the jury be able to consider and give effect to’ a capital defendant's mitigating evidence." See Tennard v. Dretke, 542 U.S. 274, 285, 124 S.Ct. 2562, 159 L.Ed.2d 384 (2004) (quoting Boyde v. California, 494 U.S. 370, 377-78, 110 S.Ct. 1190, 108 L.Ed.2d 316 (1990) ); see also Green, 442 U.S. at 97, 99 S.Ct. 2150 (holding that a "mechanistic[ ]" use of the hearsay rule to keep a capital defendant from introducing mitigating evidence at sentencing in a capital case offends due process (quotation marks omitted)).

None of this is code for anything goes, however. For a judge can exclude "information" if "its probative value is outweighed by the danger of creating unfair prejudice, confusing the issues, or misleading the jury." See 18 U.S.C. § 3593(c) ; see also Sampson I, 486 F.3d at 45 (stating that the "low barriers to admission of evidence in a capital sentencing hearing ‘do[ ] not mean that the defense has carte blanche to introduce any and all evidence that it wishes’ " (quoting United States v. Purkey, 428 F.3d 738, 756 (8th Cir. 2005) )).

The government in our case recognized that Dzhokhar's penalty-phase defense turned on what Tamerlan's role was. Which probably explains why in its own penalty-phase arguments, the government continually called the brothers equally culpable and stressed Tamerlan's lack of influence over Dzhokhar. The jurors cared about the brothers' relative culpability as well, a point made quite clear by their not recommending death for Dzhokhar on the capital counts involving Tamerlan's conduct in setting off the first bomb. And given how the proceedings played out, the probative value of showing that the bombings were not the first time Tamerlan committed acts of brutality and persuaded others to help him is obvious. So we cannot agree with the judge that the Waltham evidence lacks (emphasis ours) "any " probative force.

Inspired by his jihadi beliefs, Tamerlan's lead role in the Waltham killings — felonies (according to the kept-out evidence) that he committed without Dzhokhar — makes it reasonably more likely that he played a greater role in the crimes charged here than Dzhokhar. And as we said a moment ago, evidence showing a defendant's minor role in the offense is relevant mitigating evidence under the rule broadcast in Lockett. See Enmund, 458 U.S. at 797-98, 102 S.Ct. 3368.

But there is more to be said in Dzhokhar's favor than that.

The Waltham evidence was also highly probative of Tamerlan's ability to influence Dzhokhar. Because of the judge's decisions, Dzhokhar did not present proof showing how he learned (months after the fact, per college-acquaintance Kadyrbayev) that Tamerlan had butchered the men, one of whom was a close friend — actions motivated by Tamerlan's vision of jihad. But evidence of this sort could reasonably have persuaded at least one juror that Dzhokhar did what he did because he feared what his brother might do to him if he refused (and remember, a jury may consider any mitigating factor at least one juror found proved by a preponderance of the information). Or put slightly differently, at least one juror could reasonably have found that because of what had happened in Waltham, Tamerlan was not just "bossy" (to use the prosecutor's word) but a stone-cold killer who got a friend to support his fiendish work. And if Tamerlan could influence Todashev (a mixed-martial-arts bruiser who followed Tamerlan because he "did not have a way out"), Tamerlan's influence over Dzhokhar (his younger brother with no prior history of violence) could be even stronger. All of which strengthens two of Dzhokhar's mitigating factors — his susceptibility to Tamerlan's influence, and his having acted under Tamerlan's influence.

And despite its hard work, the government's responses do not persuade us otherwise.

The government's lead argument is that the Waltham evidence cannot clear the low relevancy hurdle because that evidence (at least in its mind) would have told the jurors nothing about the brothers' relative culpability here. Not so. Again, Dzhokhar premised his mitigation theory on his being less culpable than Tamerlan because he would not have committed the charged crimes but for Tamerlan's influence. And Tamerlan's earlier domineering and deadly acts had relevance to this theory. The judge admitted other, lesser evidence of Tamerlan's belligerence — like his screaming at others for not conforming to his view of how a good Muslim should act. And the judge did so because he deemed that evidence relevant to Tamerlan's "domination." Even this limited evidence convinced some jurors to find the existence of mitigating factors touching on Tamerlan's prior radicalization, leadership role in the bombings, and influence over Dzhokhar. And if Tamerlan's yelling at someone for selling halal turkeys had the effect of showing his dominance and radicalization, then evidence of his having conscripted a friend into a jihad-inspired robbery and killing scheme would have increased that effect exponentially.

The government is wrong to imply that the jury had to make leaps of imagination to connect what Tamerlan did in Waltham to his influence over Dzhokhar. If the judge had admitted this evidence, the jurors would have learned that Dzhokhar knew by the fall of 2012 that Tamerlan had killed the drug dealers in the name of jihad. They also would have known that it was only after these killings that Dzhokhar became radicalized as well: Evidence actually admitted showed that Dzhokhar first flashed signs of radicalization — as is obvious from his texts on jihad — after spending a holiday break with Tamerlan several weeks or so after learning about the Waltham murders. So if the jurors had heard Todashev's description of how he felt powerless to withdraw from the Waltham crimes once Tamerlan chose to turn an armed robbery into a triple murder, at least one juror might have found that Dzhokhar felt the same way when it came to the bombings in early 2013.

And if the judge had admitted the Waltham evidence — evidence that shows (like no other) that Tamerlan was predisposed to religiously-inspired brutality before the bombings and before Dzhokhar's radicalization — the defense could have more forcefully rebutted the government's claim that the brothers had a "partnership of equals." The Waltham evidence would have helped the defense show that Tamerlan inspired his younger brother not only to believe in jihad but also to act on those beliefs — just as he had in Waltham (again, the government does not suggest that Tamerlan did not commit the murders). Similarly, the evidence could have helped the defense counter the government's argument that Tamerlan and Dzhokhar "bear the same moral culpability" and that Dzhokhar acted "independently" in placing the bomb at the finish line — for the evidence showed that Tamerlan, unlike Dzhokhar, had a history of horrific violence, which he justified as jihad; that Tamerlan, unlike Dzhokhar, had previously instigated, planned, and led brutal attacks; and that Tamerlan, unlike Dzhokhar, had influenced a less culpable person (Todashev) to participate in murder.

The government still could have argued to the jurors — as it does to us — that Dzhokhar was nevertheless a willing criminal. The government also could have challenged the evidence's reliability, arguing that other than his self-serving statement about thinking he "had to" help clean up the scene, nothing proves Tamerlan bullied Todashev into doing anything. See 18 U.S.C. § 3593(c) (providing that either party can rebut any information received at the hearing). And maybe the government could have argued that the evidence undercuts Dzhokhar's mitigation theory — saying something like, Tamerlan had to pay Todashev money to get him to go along, while Dzhokhar joined on for free; and Todashev opted not to kill, while Dzhokhar killed with no reluctance or regret. But all of this goes to weight and credibility and not to admissibility — i.e. , the effect of Tamerlan's prior violence on Dzhokhar's radicalization, on his willingness to go from texting to bombing, was something the jurors should have gotten to decide for themselves. See, e.g., United States v. Guzmán-Montañez, 756 F.3d 1, 9 (1st Cir. 2014) (explaining that "[w]hen the issue lies on credibility of the evidence, it is up to the jury to decide" and adding that "[t]he factfinder is free to conduct its own interpretation of the evidence"); Nelson v. Quarterman, 472 F.3d 287, 313 (5th Cir. 2006) (en banc) (noting that the "strength" or "sufficiency" of mitigating evidence is for the jury to decide).

The government insists that because the circumstances of the Waltham killings are too dissimilar to the bombings, the Waltham evidence has no relevance here. But in both situations, Tamerlan committed murder with help from someone who gave no prior sign of a willingness to commit such acts. And in both situations, Tamerlan used his interpretations of Islam to justify his actions. So the government's too-dissimilar argument has no merit either.

Shifting from the relevancy question, the government defends the judge's actions by insisting that the Waltham evidence's admission would have led to mini-trials over whether Todashev's version of the killings "was believable" or just a pack of lies told to minimize his responsibility for those crimes. But the concern is overblown. As Dzhokhar notes, the defense could have relied, for instance, on the government's sworn search-warrant materials (to search Tamerlan's car after the bombings) that credited Todashev's statements to the FBI. The government now tries to soft-pedal its crediting of Todashev's account in the search-warrant affidavit as "a far cry from embracing those claims" at trial. But when the agent swore out the affidavit and the prosecutor submitted the materials to the magistrate judge, the government confirmed its belief in Todashev's veracity. See generally Franks v. Delaware, 438 U.S. 154, 164-65, 98 S.Ct. 2674, 57 L.Ed.2d 667 (1978) (explaining that "[w]hen the Fourth Amendment demands a factual showing sufficient to comprise 'probable cause,' the obvious assumption is that there will be a truthful showing," and adding that "it is to be 'truthful' in the sense that the information put forth is believed or appropriately accepted by the affiant as true"). We know of no reason why the sworn affidavit — which the government asked the magistrate judge to credit — should now be disbelieved. To this we add that the judge retained control over how much of this evidence could have come in. He also could have limited the evidence as appropriate or cut off the presentation if the evidence became too extensive — a more suitable remedy than barring all evidence of Tamerlan's murderous past. So in the end we think the Waltham evidence was sufficiently reliable to go to the jurors, who could then decide whether to believe it and how much weight (if any) to assign it in mitigation. See Guzmán-Montañez, 756 F.3d at 9.

The government is also off-base in saying that "[t]he Waltham evidence would have confusingly focused the jury's attention on Tamerlan 's character and the circumstances of an unrelated offense." But the parties and the judge put the mitigating factors before the jury, front and center — factors that made clear that Tamerlan's character and prior conduct were relevant because they bore on the broader circumstances of Dzhokhar's commission of the charged crimes. Arguing to the jury, the government called Dzhokhar's mitigation theory (centered on Tamerlan's influence over him) baseless because no evidence supported it. But the Waltham evidence could have been that evidence. And it would not have confused the jurors to have learned about it. Caselaw tells us to presume that juries follow instructions. See, e.g., Richardson v. Marsh, 481 U.S. 200, 211, 107 S.Ct. 1702, 95 L.Ed.2d 176 (1987). And the jurors' penalty-phase verdicts — not recommending death on 11 of the 17 death-eligible counts — show they fully understood that Tamerlan's relative culpability was mitigating only to the extent it bore on the brothers' respective roles in committing the charged crimes. Which compels us to reject the government's claim that the jurors would have lost sight of this distinction.

So we find the judge abused his discretion in banning the Waltham evidence. Compare McKinney v. Arizona, ––– U.S. ––––, 140 S. Ct. 702, 706, 206 L.Ed.2d 69 (2020) (stressing "that a capital sentencer may not refuse as a matter of law to consider relevant mitigating evidence"), with United States v. Rodriguez, 919 F.3d 629, 634 (1st Cir. 2019) (explaining "that a material error of law always amounts to an abuse of discretion"). The government thinks that any error was harmless beyond a reasonable doubt. But the government's harmlessness claim is essentially a reprise of its argument in support of exclusion: In its view, just as the Waltham evidence is irrelevant because it does not show that Dzhokhar participated in the bombings under Tamerlan's influence, for the same reasons, its exclusion could not have affected the jurors' decision. Again, though, the exclusion of the Waltham evidence undermined Dzhokhar's mitigation case. Sure, as the government argues, a jury armed with the omitted evidence still might have recommended death. But the omitted evidence might have tipped at least one juror's decisional scale away from death. In other words, the government cannot show to a "near certitude," see Victor, 511 U.S. at 15, 114 S.Ct. 1239, that the excluded evidence — that Tamerlan cold-bloodedly killed the drug dealers, all in the name of jihad — would not have convinced even one juror that (contrary to the government's jury argument) Dzhokhar did not "bear the same moral culpability" as Tamerlan, see Skipper v. South Carolina, 476 U.S. 1, 8, 106 S.Ct. 1669, 90 L.Ed.2d 1 (1986) (holding that because the judge's ruling excluding mitigating evidence "may have affected the jury's decision to impose the death sentence," the error was "sufficiently prejudicial" to require vacatur of the defendant's death sentence).

This leaves us then with Dzhokhar's Brady-based challenge: that the judge also erred by denying the defense access to additional evidence both favorable and material to him — specifically, the report and recordings of Todashev's FBI confession.

Prosecutors have an "inescapable" duty "to disclose known, favorable evidence rising to a material level of importance." Kyles v. Whitley, 514 U.S. 419, 438, 115 S.Ct. 1555, 131 L.Ed.2d 490 (1995) (discussing Brady ). Favorable evidence includes both exculpatory and impeachment evidence that is relevant either to guilt or punishment. See, e.g., United States v. Bagley, 473 U.S. 667, 674-76, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985) ; Giglio v. United States, 405 U.S. 150, 154, 92 S.Ct. 763, 31 L.Ed.2d 104 (1972). Material evidence includes information that creates a "reasonable probability" of a different outcome, see Kyles, 514 U.S. at 433, 115 S.Ct. 1555 — and in a capital case that encompasses data that "play[s] a mitigating, though not exculpating, role," see Cone v. Bell, 556 U.S. 449, 475, 129 S.Ct. 1769, 173 L.Ed.2d 701 (2009). But make no mistake: "A reasonable probability does not mean that the defendant ‘would more likely than not have [gotten] a different [result] with the evidence,’ only that the likelihood of a different result is great enough to ‘undermine[ ] confidence’ " in the proceeding's outcome. See Smith v. Cain, 565 U.S. 73, 75, 132 S.Ct. 627, 181 L.Ed.2d 571 (2012) (last alteration in original) (quoting Kyles, 514 U.S. at 434, 115 S.Ct. 1555 ). To find the withheld evidence not material, the judge must conclude that the other evidence is so overwhelming that, even if the undisclosed evidence had gotten in, there would be no "reasonable probability" of a different result. And this standard is not met just because the government "offers a reason that the jury could have disbelieved [the withheld evidence], but gives us no confidence that it would have done so." Id. at 76, 132 S.Ct. 627.

Several pages earlier we noted how the judge ruled that the government had already given the defense the gist of Todashev's statements and so the sought-after material did "not materially advance [the mitigation] theory beyond what is already available to the defense." But as we also explained, that material had information that the defense never saw below, including: that Tamerlan planned the Waltham crime, got Todashev to join in, and brought the key materials (gun, knives, duct tape, and cleaning supplies) to the apartment; that Tamerlan thought up the idea of killing the three men to cover up the robbery; and that Todashev felt "he did not have a way out" from doing what Tamerlan wanted. Todashev's confession showed — probably more than any other evidence — how and why Tamerlan inspired fear and influenced another to commit unspeakable crimes and thus strongly supported the defense's arguments about relative culpability. And armed with these withheld details, the defense could have investigated further and developed additional mitigating evidence. To us, this means there is a reasonable probability that the material's disclosure would have produced a different penalty-phase result. So the confession constituted Brady material, making it reversible error for the judge to rule the evidence off-limits from discovery.

And we find that the judge also erred by relying on the qualified law enforcement investigatory privilege. As the party asserting the privilege, the government had the burden of showing that withholding the FBI report and recordings would achieve the privilege's "underlying purpose" of not jeopardizing an ongoing investigation. See Puerto Rico, 490 F.3d at 62-64 (recognizing "a qualified privilege for law enforcement techniques and procedures"). The showing must be specific not speculative, concrete not conclusory. See id. at 62. But the government offered no specific ways in which disclosure would have endangered the ongoing Waltham-murders investigation — particularly if disclosure occurred under a protective order. See Ass'n for Reduction of Violence v. Hall, 734 F.2d 63, 66 (1st Cir. 1984) (emphasizing that where possible, a court should accommodate a moving party's interest in disclosure through excising privileged sections, editing or summarizing documents, or okaying discovery subject to a protective order). The Waltham murders occurred in 2011. By 2015, when Dzhokhar's penalty phase began, the only identified suspects — Tamerlan and Todashev — were both dead. And the government did not ask the district attorney's office to explain whether and why the privilege was still viable. Ultimately, the judge's speculation about how disclosing Todashev's statements might compromise the investigation was just that: speculation . Which as we just observed is not sufficient.

The long and the short of it is that the judge's handling of the Waltham evidence provides an additional basis for vacating Dzhokhar's death sentences.

Mitigation Evidence About Dzhokhar's Mental Condition

We now consider Dzhokhar's claim that the judge infracted his constitutional right against compelled self-incrimination and the criminal rules of procedure by conditioning the admission of his "non-testimonial neuropsychological evidence" on his "be[ing] interrogated, without limits, by government experts."

Background

A death-penalty defendant wishing to make an issue of his mental health and present expert evidence on that subject must notify the government within specified time limits. See Fed. R. Crim. P. 12.2(b). If he does that, the judge may order a rebuttal exam by the government's expert "under procedures ordered by the [judge]." See Fed. R. Crim. P. 12.2(c)(1)(B). Judges often appoint assistant U.S. attorneys from another district as "fire-walled" attorneys to handle this process. See, e.g., United States v. Sampson, 335 F. Supp. 2d 166, 244-45 (D. Mass. 2004). The results and reports from a rebuttal exam must be sealed and not given to the prosecution or the defense unless the defendant is found guilty and confirms his intent to rely on mental-condition evidence during the penalty phase. See Fed. R. Crim. P. 12.2(c)(2). If this happens, the defendant must then give the government any results and reports of his mental condition "about which [he] intends to introduce expert evidence." See Fed. R. Crim. P. 12.2(c)(3). But prosecutors cannot use any statement he made during an exam conducted under this regime unless he first introduces evidence of his mental condition, see Fed. R. Crim. P. 12.2(c)(4) — a rule designed to protect a defendant's Fifth Amendment privilege against compulsory self-incrimination, see Fed. R. Crim. P. 12.2 advisory committee notes. See also generally Estelle v. Smith, 451 U.S. 454, 466-69, 101 S.Ct. 1866, 68 L.Ed.2d 359 (1981) (recognizing that a psychiatrist's court-ordered competency exam in a capital case raised the same Fifth (and Sixth) Amendment concerns as an incustody interrogation by law enforcement); Buchanan v. Kentucky, 483 U.S. 402, 424, 107 S.Ct. 2906, 97 L.Ed.2d 336 (1987) (explaining that if a defendant tries to establish a mental-status defense, the government may constitutionally force him to submit to an interview with a mental-health expert for "limited rebuttal purpose[s]").

Citing Criminal Rule 12.2, Dzhokhar's lawyers filed a pretrial notice of intent to introduce expert evidence about "his mental condition as it bears on the issue of punishment." Simultaneously, they told the prosecution that they planned on presenting "neuropsychological testimony" that would rely on the results of various tests done on Dzhokhar (an intelligence test, for example). The judge created a separate sealed docket and appointed two fire-walled assistant U.S. attorneys from Connecticut and New York to manage the government's rebuttal exams and to represent the government in any related litigation. In an ex-parte proffer to the judge, Dzhokhar's attorneys claimed that the neuropsychological and neuroimaging exams revealed information that would support his mitigation theory.

Through the fire-walled attorneys, the government told the defense that it intended to have two experts examine Dzhokhar. The first would do a clinical interview of him and administer some tests (the just-mentioned intelligence test being one of them). The second would do a psychiatric exam focusing on his "life choices," especially "those decisions and actions underlying the charged criminal conduct," while also "explor[ing] the effects, if any of [his] social history, personality, mental state, social environment, family influences, peer pressure, and any duress to which [he] may have been subjected."

The defense objected to the government's notice. Insisting that prosecutors have "only a limited rebuttal right," the defense asked the judge to limit the exams of the government experts to "the same type of testing conducted by the defense experts" — i.e. , "objective" tests, like the "computer based tests," "pen and paper tests," "physical tests," and "neuroimaging test[s]" that the defense experts used. None of these tests, the defense argued, would elicit or rely on statements by Dzhokhar expressing his views about his own symptoms or history. So according to the defense, the judge had to bar the government experts from "asking questions beyond those specified in the test instruments themselves, or otherwise engaging [Dzhokhar] in any communication intended to elicit testimonial evidence, including[ ] opinions, view[s], beliefs, historical information or anything else." That is because asking such questions would "compel him to testify against himself," or so the defense contended.

Responding, the fire-walled attorneys argued that Dzhokhar could not "dictate and limit" their experts' "testing by selecting certain tests and then objecting to different tests that inform the subject matter under inquiry." Dzhokhar, they noted, had not revealed exactly "what type of mental disease or defect defense he is intending to assert" — though they suspected he would "claim that he was particularly susceptible to his brother's persuasion" based on "a dependent personality disorder." And they insisted that the government's proposed exams would rebut that claim by showing he could think independently. Admittedly, some of the information the experts got might not be admissible, they said — but they insisted that "curtailing the government's right to prepare at this juncture d[id] nothing more than allow the defendant to present completely un-contradicted testimony."

After a hearing, the judge overruled Dzhokhar's objection. Rule 12.2 did not "limit[ ] the rebuttal to simply the same ... tests or investigations" that the defense performed, the judge said. "[A]n appropriate rebuttal," the judge pointed out, might be "to say that the wrong tests were done or that insufficient tests were done." The judge did not want to prevent the government experts from using tests that they in their "professional judgment" deemed "appropriate." But the judge made clear that his ruling did not "mean necessarily" that the exams' results would be "admissible" or "usable" at trial — it all depended on what the "defense present[ed]." Attempting one last stand, defense counsel argued that "we are offering nothing testimonial from [Dzhokhar]," just "the results of pen-and-paper tests." So the defense asked the judge to "make the rebuttal call now as opposed to exposing [him] to an interrogation by a government agent, essentially." But the judge remained unmoved. "I don't think I can make it until I know what the exam[s] might reveal and what [any government expert] might be offered to say." Consequently, the exams could go forward, even though they "may not produce admissible evidence" — an approach that he said posed no constitutional problems.

After the ruling (but before the government experts could examine him), Dzhokhar withdrew his Rule 12.2 notice. "The broad scope" of the planned exams, his lawyers wrote, "without the presence of counsel," plus "the use the [judge] indicated can be made" of the exams' results, violate Dzhokhar's constitutional rights and clash with Rule 12.2. As the defense saw things, "[t]hese conditions separately and cumulatively impose too great a cost on the introduction of [his] proposed [mental-health] evidence."

Basic Appellate Arguments

Broadly speaking, Dzhokhar claims that the judge created a "constitutional mismatch." In his telling, the results of the testing he planned to introduce were not testimonial in any constitutional sense. During his test-taking — which he calls a "pen and paper" exercise, like having his "reflexes" evaluated — he neither offered his beliefs or thoughts on "historical or life events," nor talked about "the crimes charged against him[ ] or his family background." Contrastingly — at least according to Dzhokhar — and "as the price of" admitting his "non-speech" mental-health evidence, the judge essentially ruled that the government experts could interrogate him on a wide range of topics, including "the charged criminal conduct." And, the argument goes, by setting this price, the judge forced him to withdraw his notice, damaging his mitigation case — which violated his rights both under the Fifth Amendment and Rule 12.2.

The government counters with several arguments. Among other things, the government contends — citing Luce v. United States, 469 U.S. 38, 105 S.Ct. 460, 83 L.Ed.2d 443 (1984) — that by withdrawing his Rule 12.2 notice and not presenting his mental-health evidence, Dzhokhar failed to preserve this issue for appeal. Preservation aside, the government claims that he cannot show that the Fifth Amendment privilege — which protects against real rather than speculative dangers — actually applied here. The judge did not order him to submit to any mental-health exams, the government notes, and prosecutors did not comment on his failure to introduce mental-condition evidence — which makes this situation the direct opposite of compulsory. Also, the government reminds us, the judge repeatedly said that he would not necessarily admit the results of the government experts' exams. And, the argument proceeds, even if the judge did admit the exam results, prosecutors could only use them for rebuttal purposes — which the government says is perfectly permissible under Kansas v. Cheever, 571 U.S. 87, 94, 98, 134 S.Ct. 596, 187 L.Ed.2d 519 (2013). Additionally, to quote again from the government's brief, Dzhokhar's mental-health evidence "would have had little or no mitigating effect," because other evidence rebutted what the government surmises was his mitigation defense (the government highlights testimony about his academic achievements, like his getting mostly A's in middle school).

Pertinently for our purposes, Dzhokhar's reply contests the government's preservation point. As he sees things, a pair of Supreme Court cases — Brooks v. Tennessee, 406 U.S. 605, 92 S.Ct. 1891, 32 L.Ed.2d 358 (1972), and New Jersey v. Portash, 440 U.S. 450, 99 S.Ct. 1292, 59 L.Ed.2d 501 (1979) — allows a defendant to challenge on appeal obstacles that " ‘reach[ ] constitutional dimensions’ without first taking the stand," which he argues is the situation here.

Analysis

The parties note that different standards of review may come into play here: for instance, fresh-eyed review ("de novo," in legal parlance) for claims involving the judge's interpretation of the protections provided by privilege against forced self-incrimination, see Amato v. United States, 450 F.3d 46, 49 (1st Cir. 2006), but abuse-of-discretion review for claims about the exclusion of testimony under Rule 12.2, see United States v. Cartagena-Carrasquillo, 70 F.3d 706, 710 (1st Cir. 1995), and for claims concerning the scope of rebuttal testimony, see United States v. Sebaggala, 256 F.3d 59, 66 (1st Cir. 2001).

Often "[t]he simplest way" to decide an issue is "the best." See Stor/Gard, Inc. v. Strathmore Ins. Co., 717 F.3d 242, 248 (1st Cir. 2013) (quoting Chambers v. Bowersox, 157 F.3d 560, 564 n.4 (8th Cir. 1998) (R. Arnold, J.)). And that is so here.

Skipping over the parties' preservation points (which lets us avoid having to work through a complex series of arguments and cases), we conclude that Dzhokhar's reliance on the Fifth Amendment privilege fails. To get anywhere, he must show that he had "reasonable cause to apprehend danger" from submitting to interviews with the government experts, see Hoffman v. United States, 341 U.S. 479, 486, 71 S.Ct. 814, 95 L.Ed. 1118 (1951) — for as the Supreme Court has long emphasized, "the privilege protects against real dangers, not remote and speculative possibilities," see Zicarelli v. N.J. State Comm'n of Investigation, 406 U.S. 472, 478, 92 S.Ct. 1670, 32 L.Ed.2d 234 (1972). But the judge here said over and over again that he would not automatically admit the results of the government experts' exams, and that even if he did admit them, prosecutors could use them only for "rebuttal" — which is copacetic under the Fifth Amendment. See Cheever, 571 U.S. at 94, 98, 134 S.Ct. 596. So no appreciable danger of a Fifth Amendment violation would have arisen unless (1) Dzhokhar incriminated himself during the government experts' exams, (2) he still chose to present mental-health evidence, (3) the judge let a government expert testify based on Dzhokhar's self-incriminating comments, and (4) the expert's testimony was not proper rebuttal. To ask us to find this rank conjecture sufficient (as Dzhokhar does) is asking too much. See Minor v. United States, 396 U.S. 87, 98, 90 S.Ct. 284, 24 L.Ed.2d 283 (1969) (explaining that one must show " ‘real and appreciable’ risks to support a Fifth Amendment claim").

On then to another issue.

Surviving Victims' Testimony

In this section we tackle Dzhokhar's claim that the judge erred by admitting "victim impact" testimony by survivors at the penalty phase. As briefed here, his challenge is factually and legally intricate. But because there is a straightforward route to resolving it, we can streamline and simplify our discussion.

The FDPA says that when the government seeks a death sentence, it must "serve on the defendant[ ] a notice ... setting forth the aggravating factor or factors" it believes justify the death penalty. 18 U.S.C. § 3593(a). In its notice here, prosecutors specified several statutory aggravators they envisioned proving in pursuing the death penalty against Dzhokhar — including:

• his "knowingly creat[ing] a grave risk of death to 1 or more persons in addition to" the victims who died, see 18 U.S.C. § 3592(c)(5) ;

• his "committ[ing] the offense in an especially heinous, cruel, and depraved manner in that it involved serious physical abuse to the victim," see id. § 3592(c)(6) ; and

• his "committ[ing] the offense after substantial planning and premeditation to cause the death of a person and commit an act of terrorism," see id. § 3592(c)(9).

The notice also listed several nonstatutory aggravators, see id. § 3593(a)(2) — including:

• his "target[ing] the Boston Marathon, an iconic event that draws large crowds of men, women[,] and children to its final stretch, making it especially susceptible to the act and effects of terrorism"; and

• his "participat[ing] in additional uncharged crimes of violence," like "assault with intent to maim, mayhem[,] and attempted murder."

Dzhokhar's appellate argument only focuses on the victim-impact aggravator. And it proceeds in four steps: (1) He notes (emphasis ours) that § 3593(a)(2) provides that nonstatutory aggravators

may include factors concerning the effect of the offense on the victim and the victim's family, and may include oral testimony, a victim impact statement that identifies the victim of the offense and the extent and scope of the injury and loss suffered by the victim and the victim's family, and any other relevant information.

(2) He then says that the reference to the "victim and the victim's family" precludes penalty-phase testimony on "the impact of survivors' injuries on those survivors (or their families) themselves." (3) That is so, he contends, because even though this FDPA subsection does not define "victim," Congress used "victim" in three other subsections and in the Act's legislative history to refer to a "victim" who died. And (4) asking us to apply the usual rule of statutory interpretation that identical words bear identical meaning throughout the same act, he believes that such an analysis should lead us to conclude that the judge misinterpreted the statute to allow "victim impact evidence from surviving victims at the penalty phase."

If preserved, we review challenges to the judge's interpretation of the FDPA afresh (i.e. , de novo), see United States v. Troy, 618 F.3d 27, 35 (1st Cir. 2010), and challenges to his rulings admitting or excluding evidence for abuse of discretion, see Sampson I, 486 F.3d at 42. Dzhokhar says he preserved his challenges at trial; the government says he did not and so must now prove plain error on the judge's part. A famously demanding standard, plain error requires the proponent to show not just error, but error that is plain, that affects his substantial rights, and that seriously impaired the fairness, integrity, or public perception of the trial. See, e.g., United States v. Takesian, 945 F.3d 553, 563 (1st Cir. 2019). Because we see no error in any event, we need not resolve their dispute about the standard of review.

We can leave the resolution of the interpretive question about the FDPA for another day, because (as the government notes) even assuming without granting that Dzhokhar is correct here, the surviving spectators' testimony had relevance to the jury's weighing of aggravating factors other than victim impact.

For organizational convenience, the survivors' testimony Dzhokhar complains about can be grouped into these categories:

1. "[R]eactions to facing death": Jeffrey Bauman, for example, described making "peace" with death because he "had a great life." Roseann Sdoia said she knew she "was bleeding out" but resolved to "stay calm and stay conscious" because if she panicked she "would die." And Celeste Corcoran described how she at first "wanted to die" because she was in so much pain but realized she needed to "be there" for her family.

2. "[U]ncertainty about what had happened to other family members": Eric Whalley, for instance, said he and his wife each thought the other had died. And Stephen Woolfenden said he was "terrified" when first responders whisked his son away because he "didn't know if [he] was ever going to see [him] again."

3. "[F]eelings of helplessness watching their injured child or partner suffer": Rebekah Gregory, for example, said she could hear her son calling "mommy" after the blast and felt "helpless as a mother" because she could not go to him. Lying there on the pavement, she said a prayer, "God, if this is it, take me but let me know that Noah is okay." And Jessica Kensky discussed her frustration at not being able to care for her husband (one of the bombs left his "foot and part of his leg ... completely detached, hanging on kind of by a thin thread").

4. "[T]he long-term implications of becoming an amputee": Roseann Sdoia, for instance, said her amputation made it "extremely difficult" to learn to walk and run again, and to deal with the snow. Jessica Kensky said becoming a "bilateral amputee" was "terrifying." She wanted to keep "some memory" of her legs, to "paint [her] toenails," and to "put [her] feet in the sand." And the pain from surgeries and treatments was "[a]bsolutely horrendous," putting her "in a very dark place" of "really not wanting to live" anymore. And Adrianne Haslet-Davis said her husband could not attend the trial because he had checked himself into a mental-health facility.

Given how low the relevance threshold is, we cannot say that the judge slipped in finding this evidence at least minimally relevant to an aggravating factor other than victim impact.

The survivors' reactions to facing death (category 1) helped show that Dzhokhar "knowingly created a grave risk of death to 1 or more persons in addition to" the persons who died — a statutory-aggravating factor. See 18 U.S.C. § 3592(c)(5). That they felt they might die helps show they actually faced a grave risk of death. And their specific descriptions of what that felt like made it more likely that the jury would credit their statements about being at death's door. Plus their testimony could help the jury in weighing the grave-risk-of-death factor as part of its death-penalty decision. See Sampson I, 486 F.3d at 44 (upholding admission of graphic evidence about a murder because it "would help the jury to determine how much weight it should give" the aggravating factors). Above and beyond all that, their testimony could help prove other aggravators — that Dzhokhar substantially planned "an act of terrorism," a statutory-aggravator, see 18 U.S.C. § 3592(c)(9) ; and that he "participated in additional uncharged crimes of violence" like "assault with intent to maim, mayhem[,] and attempted murder," a nonstatutory aggravator.

The survivors' uncertainty regarding what happened to their family members (category 2) was relevant to the grave-risk of-death aggravator. See id. § 3592(c)(5). In most cases, family members got separated because so many victims were on the verge of dying that rescuers had to evacuate them as soon as possible. And the multiple family separations highlighted how "grave" the "risk" was. The evidence was also relevant to the statutory terrorism aggravator. It is hard to think of anything more terrifying than to lose track of one's child or parent in a life-or-death situation. The evidence was additionally relevant to a nonstatutory aggravator — that he "targeted the Boston Marathon," an event "especially susceptible to the ... effects of terrorism" because of its "large crowds of men, women[,] and children."

The survivors' feelings of helplessness as loved ones suffered (category 3) was relevant for the same reasons. Their inability to aid their family members after the bombings magnified the terror that Dzhokhar sought to create.

Finally, the testimony about the long-term implications of becoming an amputee (category 4) helped establish the statutory grave-risk-of-death aggravator. The survivors' comments about surgeries and suicidal thoughts showed that the risk of death continued past the immediate aftermath of the bombings. And the long-term effects of their injuries were directly relevant to the existence and weight of these factors.

Enough said about the surviving victims' testimony. Whole Foods Video

We now take up Dzhokhar's challenge to the admission of a video showing him buying milk at a Whole Foods soon after the bombings. What he essentially wants us to do is to remand for a hearing on his claim that agents derived the video from involuntary statements he made at the hospital after his arrest. Each side spends much energy debating two principal points. The first is whether Dzhokhar waived this claim — the government says he did just that by not moving to suppress the video before trial; while Dzhokhar says prosecutors made a pretrial promise not to use his confession, which excuses any untimeliness in his raising the claim. The second is whether agents discovered the video through an independent source untainted by any constitutional violations — the government says a tip from Tamerlan's wife led agents to the video; while Dzhokhar (noting the government identified her as the tipster after he had filed his opening brief) says his involuntary statements (not his sister-in-law's tip) steered the agents to the video. But because Dzhokhar is getting another penalty-phase trial, the parties and the judge should address these matters — e.g. , was the source for the video genuinely independent of his hospital confession, and if not, was the confession voluntary? — if the government again opts to offer the video into evidence and Dzhokhar objects.

An Expert's Testimony About ISIS and the Prosecution's Use of an Islamic Song and a Photo of Dzhokhar Raising His Middle Finger

In the next section of his opening brief, Dzhokhar raises three claims. He first knocks the judge for wrongly admitting (in the guilt phase) expert testimony on ISIS (the popular acronym for the Islamic State of Iraq and Syria). He then accuses the prosecutor of committing misconduct by juxtaposing (in the guilt-phase closing) a photo slideshow of the post-bombing carnage with a recording of an Islamic song found on a computer in Tamerlan's home that Dzhokhar used to surf the internet. And he lastly accuses another prosecutor of committing misconduct by displaying (in the penalty-phase opening) posters of the four murder victims beside a photo of him raising his middle finger at a cell-block security camera.

Background

Before trial, the defense had moved to exclude any expert testimony "about terrorist leaders and attacks in which [Dzhokhar] was not involved" unless the government could show that he knew of the materials under discussion and "endorsed" or "absorbed" them. The judge did not rule on the motion until just before the prosecution called Dr. Matthew Levitt as an expert witness on international terrorism during the guilt phase. Recognizing that Federal Rule of Evidence "403 is an important consideration," and cautioning the government not to "step too far on this," the judge denied the motion, saying that Levitt could "testify about the history of recent terrorist activity, particularly the encouragement of jihadi actions by particular prominent figures." The government, for its part, promised to be "very sensitive" to the defense's concerns.

Levitt talked to the jury about the "global jihad movement." What drove this movement, he said, was not a formal organizational structure but a decades-old "idea" that "there is a need for a global effort on behalf of Muslims to unite as a nation" and to "defend itself" through "acts of violence." The movement's ideology, he added, permitted the killing of innocents and focused its wrath on the United States. And, he further explained, calling on followers to conduct independent terrorist attacks "at home" had "become a major theme of radical propaganda."

Levitt noted that this was true not only of "al-Qaeda" but "now [also] the so-called Islamic state or ISIS." The defense objected to "bringing in" ISIS. But the judge ruled the testimony admissible "[a]s ... general background." Levitt then said that "ISIS" — which both fought and cooperated with al-Qaeda — "is the latest incarnation of this global jihad movement." And he explained that "ISIS, like al-Qaeda, has glossy magazines" and "very impressive online radical and radicalization literature" that tells supporters they "don't have to come" to a foreign battlefield — "just do something back home."

Later in his testimony, Levitt described how the conflict in the Russian republic of Chechnya had become a "rallying cry" that jihadists used to "radicalize people." He then said that the "Syrian conflict" — which started "four years" before in 2011 — had also "become a rallying cry around the world." The defense objected "to the whole discussion of Syria that goes beyond the date of any of the events alleged in the indictment" — to which the judge said, "Overruled." Levitt then explained that "[s]ticking even to the first two years of the Syrian conflict two years ago," there were "different things that drew jihadis to this conflict," including "jihadi ideology and want[ing] to go fight with the next incarnation of al-Qaeda."

We fast-forward to the prosecution's guilt-phase closing arguments. There, the prosecutor argued that the Tsarnaev brothers had been "radicalized to believe that jihad was the solution to their problems." He reviewed the evidence of Dzhokhar's radical beliefs — which included Dzhokhar's boat manifesto, plus his "library" of jihadist videos, writings, and "nasheeds" (Islamic chants) that Dzhokhar watched, read, and listened to on his computer and other devices. He noted that after the Tsarnaev brothers carjacked Dun Meng, they "went back to Watertown" to get "a CD containing those jihad nasheeds on it" for some "portable inspiration" as they continued their escape. He also noted that Dzhokhar had created a twitter account with the display name "Ghuraba" — an Islamic word that means "stranger."

Later in his guilt-phase closing, the prosecutor said that Dzhokhar had "murdered four people" and "wounded hundreds" to "make a statement" and to "be a terrorist hero." "This is how [Dzhokhar] saw his crimes," the prosecutor stated while displaying a PowerPoint presentation. The presentation combined photos with the audio of a nasheed. Involving a singer chanting "Ghuraba" repeatedly, the nasheed played over a slide of bombmaking instructions (from Inspire , al-Qaeda's English-language magazine), a photo of Dzhokhar seated in front of a black flag with Arabic script, and three images of severely wounded victims in the aftermath of the bombings. The nasheed played for about 19 seconds. After the chanting stopped, the prosecutor said that "this is the cold reality of what his crimes left behind." And then he showed additional slides of the bombings' aftermath in silence.

Calling the prosecutor's playing "this haunting music over the [photos]" a bid to "inflame religious or ethnic prejudice," the defense moved for a mistrial after the guilt-phase summations and before the jury began deliberating (the defense did not object during the closing, probably to not draw undue attention to the presentation). The government responded that both the audio file and the photos were in evidence and that the slideshow provided "perspective" on Dzhokhar's "state of mind, his radicalization." The judge denied the defense's motion, adopting "the government's radicalization position."

We now skip ahead to the prosecution's penalty-phase opening statement. There, the prosecutor displayed on easels 3-foot by 4-foot photos of Lingzi Lu, Krystle Campbell, Sean Collier, and Martin Richard. A fifth easel in the middle had a black cloth covering it. Near the end of her statement, the prosecutor said:

On July 10th, 2013, almost three months after Dzhokhar Tsarnaev had murdered Krystle Marie Campbell, Lingzi Lu, Martin Richard, and Officer Sean Collier, he was here in this courthouse. He knew the United States had charged him for his crimes. In the room that he was in, there was a video camera. [He] was alone. There was no brother with him. And once more, just as he had done with the boat [in Watertown], he had one more message to send.

The prosecutor then pulled the black cloth off the middle easel, revealing a 3-foot by 4-foot photo of Dzhokhar in his cell thrusting his middle finger at a surveillance camera. Concluding, the prosecutor remarked:

This is Dzhokhar Tsarnaev, unconcerned, unrepentant, and unchanged. Without remorse, he remains untouched by the grief and the loss that he caused. And without assistance, he remains the unrepentant killer that he is. It is because of who [he] is that the United States will return and ask you to find that the just and appropriate sentence for [him] is death.

After the opening statement, a lawyer for Dzhokhar noted as a "point of record-keeping" that the prosecution had "displayed the cell block photograph" during its opening. Counsel claimed "that the prejudicial" and "inflammatory" effect "of what we think was an out of context and ... quite distorted still [shot] from the cell block was greatly enhanced ... by its juxtaposition between these very attractive and touching photographs of the victims in life." The judge did not comment on the issue.

Basic Appellate Arguments

Dzhokhar calls Levitt's ISIS testimony both "irrelevant and prejudicial," noting that the group (which he had no ties to) "was well known for its barbarism at the time of his trial, but unknown — indeed, hardly existent — at the time of his crimes." He labels the prosecution's audiovisual presentation misconduct. According to him, by "pairing religiously evocative images and gruesome photographs of the bombings, and overlaying both with an Arabic chant," the prosecution "played to commonly held biases against Muslims: that they are foreign, frightening, and violenceprone." And he alleges an instance of misconduct in the prosecution's extracting the image of him "raising his middle finger at a cellblock camera, juxtapos[ing] it with photographs of the four decedents in the case, and then [telling] the jurors" — "with no factual basis" — "that this obscene gesture was [his] 'message' to his victims." All of which, his theory runs, shows the jury voted for death under the influence of "[p]assion and [p]rejudice."

Taking the opposite view, the government argues that Levitt's testimony was more pertinent than prejudicial, because his comments helped the jurors see how the global jihad movement inspires home-grown militants to commit "independent terrorist attacks" — comments he delivered briefly and in an academic tone. The government also defends the propriety of the PowerPoint presentation, saying the playing of "a 19-second audio clip of a nasheed ... over photos of [Dzhokhar] and the bombing's aftermath" were "tied specifically to the trial evidence regarding [his] inspiration for the bombing." And the government sees no prosecutorial misconduct regarding the middle-finger poster, because — contrary to the defense's representation — the prosecutor did not "say [Dzhokhar's] middle finger was a message ‘to his victims’ " but instead only said that his "gesture was intended to send the same 'message' that he had written in a boat before his arrest, when he wrote that the bombings were a 'message' to the United States Government."

Analysis

Taking first things first, we consider Levitt's guilt-phase testimony. Preserved relevance and prejudice claims, like these ones, prompt "abuse-of-discretion review — a famously deferential standard that requires a challenger to show that no rational person could accept the judge's decision." See United States v. Rodríguez-Soler, 773 F.3d 289, 293 (1st Cir. 2014). But even deference has limits. See United States v. Ayala-García, 574 F.3d 5, 18 (1st Cir. 2009). And our deference reaches its limit here.

The relevance of Levitt's ISIS testimony is hard to see. For example, we do not understand how the actions of a group that did not meaningfully exist before Dzhokhar's crime could have made any fact of consequence more likely (the government admits that the threat posed by ISIS was generally "after the Boston Marathon attacks"). See United States v. Kilmartin, 944 F.3d 315, 335 (1st Cir. 2019) (emphasizing that "[e]vidence is relevant as long as it has some tendency to make a fact of consequence more or less probable"). And calling the evidence "background," as the judge did, does not move the needle. Again, because ISIS barely existed at the time of the bombings, Levitt's testimony could not have provided "background" or "stich[ed] together an appropriate context in which the jury could assess the evidence introduced during the trial." See United States v. McKeeve, 131 F.3d 1, 13 (1st Cir. 1997). And by falsely linking Dzhokhar to this infamously brutal group, the unfair prejudicial effect of Levitt's ISIS comments far outweighed any probative value that it had.

But what saves the government is that this error is harmless beyond a reasonable doubt. See 18 U.S.C. § 3595(c)(2) (explaining that a circuit court cannot reverse or vacate a federal death sentence if the error is harmless beyond a reasonable doubt). Running about two transcript pages, the contested testimony was briefly given and tonally academic. Plus an overwhelming amount of other evidence showed that Dzhokhar drew inspiration from radical Islamic propaganda, including from articles in a magazine published by al-Qaeda (Inspire ) and from lectures given by an Imam connected to al-Qaeda. So any suggestion during the guilt phase that he got inspiration from another radical Islamic group (ISIS) would not have affected the jury's sentencing verdict. Moreover, the jurors unanimously found the existence of several statutory intent factors, statutory aggravating factors, and nonstatutory aggravating factors supported the death penalty. And Dzhokhar does not challenge the evidentiary support for any of them. And given the overwhelming force of these factors — driving home the devastating effects of Dzhokhar's actions — we find beyond a reasonable doubt that the jury would have imposed death even if the judge had excluded the ISIS testimony. See generally Jones, 527 U.S. at 404-05, 119 S.Ct. 2090 (noting that a reviewing court doing a harmless-error check of a death sentence can consider whether "the jury would have reached the same conclusion" absent the error).

Turning to the prosecutorial-misconduct claims, we note that we review preserved claims de novo (that is, without giving the district judge's decision any weight), and unpreserved claims for plain error. See, e.g., United States v. Sepúlveda-Hernández, 752 F.3d 22, 31 (1st Cir. 2014). The parties dispute whether Dzhokhar preserved all of his claims. But the dispute is academic, because any error also fails on harmless-error review.

When faced with a prosecutorial-misconduct allegation, we first look to see if the prosecutors acted improperly. See, e.g., United States v. Veloz, 948 F.3d 418, 435 (1st Cir. 2020). If they did, we then see if their misconduct "so poisoned the well that the trial's outcome was likely affected," id. (quoting French, 904 F.3d at 124 ) — "weigh[ing] factors such as the severity of the misconduct, the context in which it occurred, the presence or absence of curative instructions, and the strength of the evidence," United States v. Walker-Couvertier, 860 F.3d 1, 10 (1st Cir. 2017). Ultimately, reversal is justified "only where there would be a miscarriage of justice or where the evidence preponderates heavily against the verdict." United States v. Rodríguez-De Jesús, 202 F.3d 482, 486 (1st Cir. 2000) (quoting United States v. Gonzalez-Gonzalez, 136 F.3d 6, 12 (1st Cir. 1998) ).

Both sides have strong arguments. As for the audiovisual presentation, Dzhokhar correctly says that prosecutors offered no proof that the nasheed used had any significance to him (let alone that he ever played it or that it had any connection to the crimes themselves), and they never played that nasheed during the trial. And the government correctly notes that the presentation consisted of photos and an accompanying audio file that the judge had admitted into evidence. Dzhokhar calls the presentation — scored with a "foreign-sounding soundtrack" — too emotional or frightening, intended not to inform but "to stoke religious bias." To this the government replies — with quotes from United States v. Mehanna, 735 F.3d 32, 64 (1st Cir. 2013) (quotation marks omitted) — that "terrorism-related evidence is often emotionally charged," even "alarming" and "blood curdling," yet "much of this emotional overlay is directly related to the nature of the [terrorist] crimes." And, the government suggests, if that is true of Mehanna, where the defendant committed no violent acts, it is all the more reasonable to expect emotionally charged evidence and argument here, where Dzhokhar partook in terrorist attacks that killed four and grievously injured hundreds of others.

Rather than resolving this close question, we assume error — though the error did not irreversibly poison the well. As per usual, the judge told the jurors that counsel's arguments are not evidence. And as required by the FDPA, see 18 U.S.C. § 3593(f), the judge told them they could not consider Dzhokhar's religious beliefs or national origin in deciding whether to recommend death. Also as required by the FDPA, they specifically certified in the penalty-verdict form that "consideration of the ... religious beliefs" or "national origin ... of Dzhokhar ... was not involved in reaching" their decision. Last but not least, in the context of this case — with the overwhelming evidence of Dzhokhar's devotion to radical jihadist ideology and with his guilt unquestioned — the jury's penalty-phase verdict was not likely affected by 19 seconds of music played weeks earlier during the guilt phase.

Sounding a familiar refrain, we note that each party makes good points on the next issue too. Dzhokhar insists that by displaying the image of his middle-finger salute alongside photos of those who had died, the prosecutor fueled the jurors' passions by saying — without any factual support — that this was his "message" to the murder victims. The government insists that the prosecutor never said that he was flipping off his victims — rather, she said that what he did was meant to convey the same "message" as his "boat" manifesto: that a religious duty to wage jihad against the United States justified his actions. And the government reminds us that courts do not lightly infer that a prosecutor intends an "ambiguous remark" to carry its most harmful meaning. Not leaving that claim unanswered, Dzhokhar says that given how the prosecutor displayed the photos, "[t]he inference that these images were related to one another is not only the most damaging meaning, but also the most obvious."

Yet even if the government is wrong and Dzhokhar is right, he cannot win. The evidence overwhelmingly showed his disdain for his victims. Far from simply gesturing at them, he set off a bomb designed to kill them by sending pieces of metal tearing through their bodies. And after doing this, he later tweeted that he was "a stress free kind of guy." So even if the jurors understood the prosecutor as saying that Dzhokhar directed his middle-finger salute at his victims, the prejudice from that inference would pale when compared with evidence of his violently and intentionally killing them — without showing any remorse.

Even though the government won the prosecutorial-misconduct challenges on harmless error, we suggest that prosecutors not repeat these tactics on remand. And by tactics we mean the prosecution's (a) showing the jury a photo slideshow of the post-bombing carnage scored with a nasheed that had zero connection to the crime itself; and (b) displaying a poster-sized photo of Dzhokhar sticking out his middle finger placed between the same-sized photos of the decedents, thereby implying his gesture constituted his "message" to the victims — even though no evidence showed he in fact directed his gesture toward the victims.

Penalty-Phase Jury Instructions About Weighing Aggravating and Mitigating Factors

We turn our attention to Dzhokhar's suggestion that the judge stumbled by not telling the jurors that to recommend death they had to find beyond a reasonable doubt that the aggravating factors outweighed any mitigating ones.

Background

To place the matter into proper perspective (and to save the reader from having to flip back to a footnote many pages ago), we highlight certain aspects of how capital sentencing works. For a defendant to get a death sentence under the FDPA, the jurors must make several penalty-phase determinations — including: they must find unanimously and beyond a reasonable doubt that he acted with the statutorily required intent, see 18 U.S.C. § 3591(a)(2) ; they must find unanimously and beyond a reasonable doubt that at least one statutory aggravator is present, see id. § 3593(e)(2); see also id. § 3592(c); and they must find unanimously that the aggravators (statutory and nonstatutory) sufficiently outweigh any mitigators, see id. § 3593(e).

Before the penalty-phase deliberations started, the defense asked the judge to tell the jurors that they could only call for Dzhokhar's death if they found the aggravators outweighed any mitigators "beyond a reasonable doubt." Without hearing any argument, the judge said that "Circuit law" precluded him from giving an instruction like that. The judge probably had in mind Sampson I — a case holding that because "the requisite weighing constitutes a process, not a fact to be found," a jury need not make the weighing determination beyond a reasonable doubt. See 486 F.3d at 32.

Basic Appellate Arguments

The Supreme Court tells us "that only a jury, and not a judge, may find [beyond a reasonable doubt] facts that increase a maximum penalty, except for the simple fact of a prior conviction." See Mathis v. United States, ––– U.S. ––––, 136 S. Ct. 2243, 2252, 195 L.Ed.2d 604 (2016) (citing Apprendi v. New Jersey, 530 U.S. 466, 490, 120 S.Ct. 2348, 147 L.Ed.2d 435 (2000) ). Dzhokhar says that the weighing determination is a "fact[ ]" that ups a defendant's "maximum possible punishment from life to death." So he argues that the judge erred by not telling the jurors that they had to find the aggravators outweighed the mitigators under the reasonable-doubt standard. He admits that Sampson I forecloses his claim. But he thinks that the Supreme Court "abrogated" Sampson I in Hurst v. Florida, 577 U.S. 92, 136 S. Ct. 616, 193 L.Ed.2d 504 (2016).

The government disagrees. Sampson I, the government says, is still good law because nothing in Hurst weakens Sampson I's holding that the "outweighs" decision — coming into play only after the jurors find the defendant death-eligible beyond a reasonable doubt — is not a fact determination, but a moral one about what is just.

Analysis

Dzhokhar's claim rises or falls on the notion that Hurst requires that jurors make the weighing determination beyond a reasonable doubt. Using our independent (or de novo) judgment, see Sampson I, 486 F.3d at 29, we think his argument must fall.

Hurst invalidated Florida's capital-sentencing scheme. See 136 S. Ct. at 619-20. Under that scheme, "the maximum sentence a capital felon" could get based on "the [jury] conviction alone [was] life imprisonment." Id. at 620. He could get a death sentence only if the judge later determined that (1) "sufficient aggravating circumstances exist" and that (2) "there are insufficient mitigating circumstances to outweigh the aggravating circumstances." Id. at 620, 622 (quotation marks omitted). Hurst said that determination (1) — that sufficient aggravators exist — violated the defendant's constitutional right to a jury trial, because it "impermissibly allowed ‘a sentencing judge to find an aggravating circumstance, independent of a jury's factfinding, that is necessary for imposition of the death penalty.’ " McKinney, 140 S. Ct. at 707 (quoting Hurst, 136 S. Ct. at 624 ). But importantly here, Hurst made no holding regarding determination (2) — that the mitigators do not outweigh the aggravators. See 136 S. Ct. at 624 (summarizing the case as holding that Florida's sentencing procedure, "which required the judge alone to find the existence of an aggravating circumstance , is ... unconstitutional" (emphasis added)).

About a week after Hurst came out, the Supreme Court issued Kansas v. Carr. Carr held that the Constitution does not "require[ ] capital-sentencing courts ... to affirmatively inform the jury that mitigating circumstances need not be proven beyond a reasonable doubt." See 577 U.S. 108, 136 S. Ct. 633, 642, 193 L.Ed.2d 535 (2016) (quotation marks omitted). In doing so, Carr

doubt[ed] whether it is even possible to apply a standard of proof to the mitigating-factor determination.... Whether mitigation exists ... is largely a judgment call (or perhaps a value call); what

one juror might consider mitigating another might not.

Id. And then Carr emphasized the discretionary nature of the weighing process, saying

the ultimate question whether mitigating circumstances outweigh aggravating circumstances is mostly a question of mercy — the quality of which, as we know, is not strained. It would mean nothing , we think, to tell the jury that the defendants must deserve mercy beyond a reasonable doubt; or must more-likely-than-not deserve it.

Id. (emphasis added).

Carr causes problems for Dzhokhar in two ways. One, if the Supreme Court in Hurst intended to impose the reasonable-doubt standard on the weighing process — as Dzhokhar argues — the Court in Carr would not have said days later that telling the jury to use that standard "would mean nothing." And two, Carr's "mercy" talk supports Sampson I's statement that "[t]he outcome of the weighing process is not an objective truth that is susceptible to (further) proof by either party." See 486 F.3d at 32.

Now consider McKinney v. Arizona, a Supreme Court opinion from this year. McKinney held that while cases like Hurst require a jury to "find the aggravating circumstance that makes the defendant death eligible," they "did not require jury weighing of aggravating and mitigating circumstances." See 140 S. Ct. at 707-08 (holding that an appellate court can reweigh aggravators and mitigators if the judge failed to properly consider a mitigator). So McKinney helps sink Dzhokhar's claim that Hurst requires the jury to make the weighing determination beyond a reasonable doubt — a view we hold because McKinney makes crystal clear Hurst addressed only the finding of aggravating facts and had nothing to do with the weighing process.

The bottom line of this discussion is that our Sampson I opinion — holding that the reasonable-doubt standard does not apply to the weighing process — remains good law.

Penalty-Phase Jury Instructions About Jury Deadlock

Dzhokhar makes a second claim of instructional error: that the judge botched the proceedings by not telling the jurors that failure to reach a unanimous recommendation on the death penalty would result in his imposing a life sentence without the possibility of release.

Background

As a matter of helpful repetition, we emphasize again that jurors in a federal capital case cannot recommend that a defendant die unless (as relevant here) they unanimously find beyond a reasonable doubt a requisite intent factor and an aggravating circumstance, and then unanimously find that all the aggravators outweigh any mitigators to justify his getting death. See Jones, 527 U.S. at 376-77, 119 S.Ct. 2090 (reviewing the FDPA). And if they cannot make a unanimous recommendation, the judge steps in and can impose either a life sentence without release or any lesser sentence allowed by law. See id. at 380-81, 119 S.Ct. 2090. The parties concur that jury deadlock on any of these prerequisites here would have resulted in a sentence of life without release.

With this understanding in place, we return to the particulars of Dzhokhar's case — right before the penalty-phase deliberations. The defense asked the judge to tell the jurors that he would impose a sentence of life without release if, after weighing the aggravating and mitigating circumstances, they could not unanimously agree on a sentencing recommendation. The proposed instruction read:

If the jury is unable to reach a unanimous decision in favor of either a death sentence or of a life sentence, I will impose a sentence of life imprisonment without possibility of release upon the defendant. That will conclude the case. At this sentencing stage of the case, the inability of the jury to agree on the sentence to be imposed does not require that any part of the case be retried. It also does not affect the guilty verdicts that you have previously rendered.

The defense conceded that the Supreme Court's Jones decision "authorize[d] district courts" to refuse to give such an instruction. See 527 U.S. at 381, 119 S.Ct. 2090 (holding that judges are not required to instruct sentencing juries on the consequences of a deadlock in the weighing process). But the defense claimed that without it, the jury might "wrongly assume that a failure to agree on sentence would require the case to be retried before a new jury." And this mistaken belief, the defense added, would "coerc[e]" some jurors into accepting a death verdict to avoid having to "put the victims and the survivors and the entire community through this entire case again."

The judge rejected the defense's request, saying that the suggested instruction could "undercut[ ]" the "process anticipated by" the FDPA by essentially empowering "one juror" to "simply decid[e] that the decision was his or hers" without sufficiently engaging in the deliberative process. But the judge explained that he would address the defense's coercion concerns by giving "a very strong instruction" that "each individual juror is to give his or her own [verdict] and not agree just to agree with others."

After the parties presented their penalty-phase evidence, the judge told the jurors that they had to decide — unanimously and beyond a reasonable doubt — whether the government established the existence of one of the "gateway" mental-intent factors (which we will later discuss) and one of the statutory aggravating factors. The judge noted that if they could not so agree, he then would sentence Dzhokhar to life imprisonment without release. But the judge said that if they could so agree, they then had to consider whether the aggravators outweighed any mitigators to justify a death sentence. And if, after the weighing process, they unanimously found that death or life without release was the proper sentence, the judge stated that they should mark the corresponding part of the verdict form.

The judge did not instruct the jurors about what would happen if they deadlocked in making a sentencing recommendation. But he did tell them that "[b]efore you reach any conclusion based on a lack of unanimity on any count, you should continue your discussions until you are fully satisfied that no further discussion will lead to a unanimous decision." And he emphasized that "[a]ny one of you is free to decide that a death sentence should not be imposed," that "[e]ach juror must individually decide" whether to recommend death, and that "no juror is ever required to impose a sentence of death."

The defense later objected to the judge's "refusal" to instruct the jurors about what would occur if they deadlocked on the penalty recommendation.

Basic Appellate Arguments

Dzhokhar criticizes the judge for not telling the jurors that if they could not unanimously agree on whether to recommend death, then he (Dzhokhar) would automatically get life without release. He thinks this because to him the jurors likely drew a "negative inference" from the instructions at earlier stages — i.e. , that because the judge said that deadlock at the intent and aggravator stages would result in an automatic life sentence without release, his not saying anything about deadlock at the weighing stage would cause them to infer that the "failure to reach unanimity" there "would yield a result other than a mandatory life sentence." And, still repeating arguments made below, this "omission[ ]" (to quote again from his brief) "created an intolerable risk of coercing holdout jurors for life to acquiesce in a death verdict in order to spare the victims' families, the survivors, and the Boston community the significant financial and emotional strain" of a second penalty-phase "trial."

The government takes a diametrically opposed position. It says that the judge's decision not to instruct about the effect of a deadlock on the sentence decision squares with the Supreme Court's Jones opinion. It also says that Dzhokhar's argument about the jurors drawing a negative inference "is speculative at best" and so cannot undermine the judge's ruling.

Analysis

The parties disagree about what standard of review applies to this claim. We review anew (de novo, as the cases say) preserved claims that the jury instructions mislead the jurors, "taking into account the charge as a whole and the body of evidence presented at trial." Sampson I, 486 F.3d at 29. Dzhokhar believes he said enough below to preserve his arguments. The government believes differently. But we assume without deciding that Dzhokhar is correct because he loses here under either standard.

Jones lights the path to decision. Construing § 3594 of the FDPA, Jones held that if the jury fails to reach a unanimous verdict on punishment for a capital crime, the judge must enter a sentence other than death — so there is no mistrial or second penalty-phase proceeding. See 527 U.S. at 380-81, 119 S.Ct. 2090. Jones also held that a judge need not tell the jurors about the consequences of deadlock at that stage. See 527 U.S. at 381, 384, 119 S.Ct. 2090. "[I]n a capital sentencing proceeding," Jones explained, "the Government has ‘a strong interest in having the jury express the conscience of the community on the ultimate question of life or death.’ " Id. at 382, 119 S.Ct. 2090 (quoting Lowenfield v. Phelps, 484 U.S. 231, 238, 108 S.Ct. 546, 98 L.Ed.2d 568 (1988) ). And telling the jurors about the consequences of nonunanimity, Jones said, could undermine those vital interests because it might amount to "an open invitation for the jury to avoid its responsibility and to disagree." Id. at 383-84, 119 S.Ct. 2090 (quoting Justus v. Virginia, 220 Va. 971, 266 S.E.2d 87, 92 (1980) ). Jones also stressed that if a defendant thinks the judge's charge "caused jury confusion," he must show "a reasonable likelihood that the jury has applied the challenged instruction[s]" in a legally flawed way. Id. at 390, 119 S.Ct. 2090 (quoting Estelle v. McGuire, 502 U.S. 62, 72, 112 S.Ct. 475, 116 L.Ed.2d 385 (1991) ).

Dzhokhar has not shown a reasonable likelihood that the jury applied the instructions incorrectly. Our reasons for so concluding are threefold.

In the first place, the judge's instructions that he would sentence Dzhokhar to life without release if the jury deadlocked at the intent-and-aggravator-factor stages passed legal muster. See Jones, 527 U.S. at 380-81, 119 S.Ct. 2090. So too did the judge's decision not to instruct on the effect of their deadlocking on the sentence decision. See id. at 381-84, 119 S.Ct. 2090. Devastating to his claim, Dzhokhar cites no precedent holding that if the judge instructs on the effect of an impasse at one stage, he must also do so at every other stage.

In the second place, Dzhokhar's negative-inference theory — the omission of a consequences-of-deadlock instruction at the weighing stage signaled to the jury that a deadlock there would lead to a mistrial and a new penalty phase — rests on nothing but speculation. See Carr, 136 S. Ct. at 643 (stressing that "[a] meager 'possibility' of confusion is not enough" (quoting Boyde v. California, 494 U.S. 370, 380, 110 S.Ct. 1190, 108 L.Ed.2d 316 (1990) )). We doubt that the jurors recognized the inconsistency that his lawyers see, particularly since "[j]urors do not sit in solitary isolation booths parsing instructions for subtle shades of meaning in the same way that lawyers might." See Brown v. Payton, 544 U.S. 133, 143, 125 S.Ct. 1432, 161 L.Ed.2d 334 (2005) (quoting Boyde, 494 U.S. at 380-81, 110 S.Ct. 1190 ). But even indulging his speculative inference drawing, we think that the jurors were at least as likely to conclude that the effect of a deadlock at the weighing stage would be the same as at the earlier stage — i.e. , that the judge would hand down a sentence of life without release. And even assuming the jurors wanted to avoid a new penalty phase, we believe the instructions could just as easily have caused them to compromise by choosing to recommend life without release. Anyhow, their verdict shows they did not feel compelled to return a death verdict, given how they recommended life on 11 of the 17 death-eligible counts.

Also hurting Dzhokhar is that Jones rejected a similar negative-inference claim. The Jones defendant claimed that an alleged ambiguity in the instructions might have caused the jury to think that if it failed to reach unanimity on the sentencing issue, the judge might give him a term less severe than life without release. 527 U.S. at 387, 119 S.Ct. 2090. But Jones rebuffed this negative-implication argument, holding that the defendant had "parse[d]" the instructions "too finely" and that — after considering the instructions as a whole — the inferences he relied on did not "create a reasonable likelihood" of confusion over the deadlock's effect. Id. at 391-92, 119 S.Ct. 2090.

In the third place and finally, Dzhokhar cites no authority holding that an instruction that is constitutionally permissible can become unconstitutionally coercive by ambiguous negative inferences drawn from other instructions. That is probably because the caselaw is against him. For Jones holds "that instructions that might be ambiguous in the abstract can be cured when read in conjunction with other instructions." Id. at 391, 119 S.Ct. 2090. And reviewing the instructions holistically — instructions that stressed that "[a]ny [juror] is free to decide that a death sentence should not be imposed," that "[e]ach juror must individually decide" whether to recommend death, and that "no juror is ever required to impose a sentence of death" (which we presume they obeyed, see Marsh, 481 U.S. at 206, 107 S.Ct. 1702 ) — we see no basis for Dzhokhar's conjecture that any juror was coerced into voting for a death sentence to avoid causing a mistrial.

Ex-Parte Communications

Up for review here is Dzhokhar's claim that a "secret channel of communication" existed between prosecutors and the judge — "repeated private access" that violated his constitutional rights to due process and effective assistance of counsel.

Background

During Dzhokhar's prosecution, the government filed a number of documents ex parte. And the judge held a number of ex-parte conferences with the government. All of this resulted in 26 ex-parte docket entries, involving 4 court orders, 16 government motions or notices, and 6 ex-parte conferences.

As Dzhokhar's appeal moved along, the government (with the judge's approval) voluntarily disclosed 13 of the ex-parte filings to the defense. After some motion practice, the government disclosed a lightly-redacted transcript of an ex-parte conference on the Waltham evidence. So 12 ex-parte items remain undisclosed.

Basic Appellate Arguments

The nub of Dzhokhar's argument is that the judge's "back-channel talks" with the government robbed him of his "Fifth Amendment right to due process and his Sixth Amendment right to the assistance of counsel." Quoting a decision from us, he points out that

not only is it a gross breach of the appearance of justice when the defendant's principal adversary is given private access to the ear of the court, it is a dangerous procedure [because it invites the question whether] "[t]he firmness of the court's belief [in the prosecutor's position] may well have been due not only to the fact that the prosecutor got in his pitch first, but, even more insidiously, to the very relationship ... that permitted such [ex parte] disclosures."

See Haller v. Robbins, 409 F.2d 857, 859-60 (1st Cir. 1969). And he notes that the constitutional right to counsel applies to all critical stages of the prosecution. See Lafler v. Cooper, 566 U.S. 156, 165, 132 S.Ct. 1376, 182 L.Ed.2d 398 (2012) (underscoring that this "constitutional guarantee applies to pretrial critical stages that are part of the whole course of a criminal proceeding, a proceeding in which defendants cannot be presumed to make critical decisions without counsel's advice").

Seeing no violations, the government insists that the in-camera procedures helped the judge "to independently assess whether the materials were discoverable" — and because they "were not" discoverable, Dzhokhar had "no right to obtain them." As support, the government stresses that a judge's ex-parte, in-camera review of documents may be authorized under the Classified Information Procedures Act and Criminal Rule 16(d)(1). See United States v. Pringle, 751 F.2d 419, 426-28 (1st Cir. 1984). And quoting one of our cases, the government adds that the "requirements of confidentiality [can] outweigh the interest in adversarial litigation and permit a court to rule on an issue in camera and without the participation of an interested party." See United States v. Innamorati, 996 F.2d 456, 487 (1st Cir. 1993).

Analysis

The interests on each side of this controversy are profound. And both parties have points in their favor. But reviewing Dzhokhar's legal challenge de novo, see United States v. Lustyik, 833 F.3d 1263, 1267 (10th Cir. 2016) — i.e. , without giving the judge's take any special weight — we side with the government.

A criminal defendant's right to an adversary proceeding is central to our system of justice. See, e.g., Innamorati, 996 F.2d at 487. That right includes the right to have counsel at all critical stages of the criminal process. See, e.g., United States v. Cronic, 466 U.S. 648, 654, 659, 104 S.Ct. 2039, 80 L.Ed.2d 657 (1984). But the law permits some exceptions to this norm. See, e.g., Innamorati, 996 F.2d at 487. In exceedingly "rare situations" a judge may act in camera and with the benefit of only the prosecution's views, like when there is a need to stop disclosure of sensitive information — for example (and without limitation), material that could damage national security, see id., compromise an in-progress criminal inquiry, see Puerto Rico, 490 F.3d at 64, or fall outside the rule of Brady v. Maryland, see United States v. Claudio, 44 F.3d 10, 14 (1st Cir. 1995). And because the point is so powerful and cannot be made enough, we repeat what we said in Innamorati:

Outside of emergencies, ... the ex parte submission of information from a party to the court and the court's ruling on that information without notice to or participation of the opposing party is fundamentally at odds with our traditions of jurisprudence ... and can be justified only in the most extraordinary circumstances.

996 F.2d at 487.

The "burden of justification" here is on the government. See Claudio, 44 F.3d at 14. And it is a burden the government has carried.

The government notes that aside from a few documents on a restitution issue (which the judge never ruled on), all of the remaining ex-parte items involve "either classified or otherwise sensitive material" that prosecutors gave the judge for an in-camera review to see if "the material should be protected from disclosure or should instead be produced to the defense." Asking us to take his side, Dzhokhar zeroes in on an ex-parte proceeding held after the defense filed a motion challenging the prosecution's proposed trial exhibits. With the benefit of only the government's presentation, the judge there offered some suggestions about how the evidence could best be shown at trial. But neither he nor the government touched on Dzhokhar's objections. We do not understand why the judge had to consider this presentation issue on an ex-parte basis. But given the "unimportance of the material" discussed at this brief hearing, any error "inflicted no prejudice" on Dzhokhar. See Innamorati, 996 F.2d at 488.

As for the other ex-parte communications, we think that the necessity to keep sensitive information from the defense sufficiently justified the procedures employed in this case. And not for nothing, but these ex-parte measures actually helped protect Dzhokhar's due-process rights, for they allowed the judge to review and rule on the materials' discoverability — rather than leaving the decision in the hands of prosecutors. See generally Pennsylvania v. Ritchie, 480 U.S. 39, 59, 107 S.Ct. 989, 94 L.Ed.2d 40 (1987) (explaining that "[i]n the typical case where a defendant makes only a general request for exculpatory material under Brady ..., it is the [government] that decides which information must be disclosed"). On this point Innamorati put it best: "[T]he interests of justice are better served by encouraging the government to let the district court resolve" concerns about sensitive information "in close cases"; and a "[d]efendant[ ] in general would not gain from a regime that encouraged the government to decide the matter itself." See 996 F.2d at 488. See also generally Ritchie, 480 U.S. at 59-60, 107 S.Ct. 989 (finding the defendant's interest in discovering exculpatory information adequately protected by trial court's in-camera review of sensitive materials).

Dzhokhar does not gain any more traction by turning to the judge's ex-parte, in-camera handling of the Todashev material — proceedings prompted by the government's pressing the qualified law enforcement investigatory privilege. True, we today hold that the judge erred by denying the defense access to these items. But the potentially sensitive nature of the information involved justified the judge's "initial" ex-parte examination. See Innamorati, 996 F.2d at 488. Which is why we applaud rather than criticize the judge's use of established protocols for assessing the merits of this privilege claim.

Not only does Dzhokhar's due-process argument collapse — his right-to-counsel argument does too.

Citing Cronic, Dzhokhar insists that "[t]he ex parte communications concerning contested discovery" violated his right to counsel at a "critical stage" of his prosecution. Cronic held that if a defendant was completely denied the right to counsel for a "critical stage" of the trial, we irrebuttably presume that it was harmful (we do not ask whether the error was harmless). See 466 U.S. at 659 & n.25, 104 S.Ct. 2039 ; see also Bell v. Cone, 535 U.S. 685, 695-96, 122 S.Ct. 1843, 152 L.Ed.2d 914 (2002). But Dzhokhar never explains how these ex-parte proceedings qualified as critical stages. Then there is the just-discussed caselaw saying that ex-parte review is appropriate in those "rare" instances where the need to keep sensitive information from the opposing party "outweigh[s] the interest" in inquisitorial proceedings. See Innamorati, 996 F.2d at 487. See also generally Ritchie, 480 U.S. at 60, 107 S.Ct. 989 (recognizing that ex-parte proceedings "den[y]" the defendant "the benefits of an ‘advocate's eye,’ " but finding no constitutional problem there because the trial judge was "obligated to release information material to the fairness of the trial"). And that takes care of his right-to-counsel theory.

We end here with a caveat. This is a jury trial, not a bench trial where the judge decides the facts. And our reasoning does not necessarily apply to the latter without further consideration.

Fair-Cross-Section Requirement

Dzhokhar contends that an underrepresentation of African Americans in the grand and petit jury wheels violated his right to an impartial jury selected from a fair cross-section of the community. He calls the statistical methodology that our circuit uses to determine underrepresentation — the absolute-disparity method — "legally and statistically unsound." Conceding that we as a three-judge panel are stuck with this circuit's approach, he says that he raises the issue simply to preserve it for possible "en banc or Supreme Court review." So "[f]or present purposes," he adds, "nothing else need be said" — a point with which we agree.

Death Penalty for Offenders Under Age 21

That takes us to Dzhokhar's constitutional claim that as a person accused of having committed death-eligible crimes when he was under 21 (he was 19 at the time of the bombings), he is "categorically exempt from the death penalty."

Citing Roper v. Simmons, 543 U.S. 551, 125 S.Ct. 1183, 161 L.Ed.2d 1 (2005), Dzhokhar concedes — as he must — that the Supreme Court has "dr[awn] a bright line" for death eligibility "at age 18." He just thinks that the factors Roper considered relevant in granting death-penalty immunity to persons under 18 — that they lack the maturity we attribute to adults; that they are more vulnerable to peer pressure than are adults; and that their personality traits are less fixed, suggesting a higher likelihood of rehabilitation of juveniles than of adults, see id. at 569-79, 125 S.Ct. 1183 — apply equally to persons under 21. Looking for support, he argues that "scientific research" since Roper "has explained the effects of brain maturation, or the lack thereof, on the behavioral and decision-making abilities of late adolescents in their late teens and early twenties." He also says that there is a "growing national consensus against the death penalty" for offenders between 18 and 20. See Am. Bar Ass'n, Resolution 111 (2018), https://www.americanbar.org/content/dam/aba/images/abanews/mym2018res/111.pdf.

Unimpressed, the government writes that Dzhokhar discusses no research about "brain maturation that is substantially different from the research available" at the time Roper came down. Citing one of his sources, the government also writes that "not a single state with an active death penalty scheme" bans the execution of 18-to-20-year-olds. And if the United States made that group death-penalty immune, the government adds, quoting another of his sources, it "would be taking an unusual legal stance with respect to prevailing international norms."

Because Dzhokhar did not raise this issue below, we review for plain error — reversing only if (among other requirements) he can show an "indisputable" error, "given controlling precedent." See United States v. Correa-Osorio, 784 F.3d 11, 22 (1st Cir. 2015). This he cannot do, however, given Roper's square holding that 18 is "the age at which the line for death eligibility ought to rest." See 543 U.S. at 574, 125 S.Ct. 1183. The change he proposes is certainly worthy of careful consideration. As members of what the Constitution calls an "inferior" court, see U.S. Const. art. III, § 1, we simply note that whether a change should occur is for the Supreme Court to say — not us, see Morey v. United States, 903 F.2d 880, 883 (1st Cir. 1990).

Crime of Violence

We end with Dzhokhar's challenge to five convictions for using a firearm during a "crime of violence."

Background

The jury convicted Dzhokhar of (among other crimes) multiple violations of 18 U.S.C. § 924(c). As relevant here, that section has two prongs: the "use or carry" prong and the "possession" prong. The first punishes anyone who "during and in relation to any crime of violence ... uses or carries a firearm." Id. § 924(c)(1)(A) (emphasis added). The second punishes anyone who "in furtherance of any such crime [ ] possesses a firearm." Id. (emphasis added). The statute carries hefty minimum prison terms, especially for recidivists (and these sentences are over and above the ones they get for the underlying crime). See id. § 924(c)(1)(A)-(C). Another provision increases the maximum penalty to death if the defendant, "in the course of a violation of subsection (c)," kills "a person through the use of a firearm" and the killing is a murder as defined in the federal murder statute. See id. § 924(j)(1).

Section 924(c) defines "crime of violence" (a phrase we italicized above) as "an offense that is a felony" and

(A) has as an element the use, attempted use, or threatened use of physical force against the person or property of another, or

(B) that by its nature, involves a substantial risk that physical force against the person or property of another may be used in the course of committing the offense.

Id. § 924(c)(3). Courts commonly call subsection (A) the "elements clause" (sometimes also referred to as the "force clause") and subsection (B) the "residual clause." See, e.g., United States v. Davis, ––– U.S. ––––, 139 S. Ct. 2319, 2324, 204 L.Ed.2d 757 (2019). Two methods exist for deciding if a prior crime is a crime of violence: the "categorical approach" and the "modified categorical approach." See, e.g., United States v. Taylor, 848 F.3d 476, 491-92 (1st Cir. 2017).

If the prior crime involved a violation of an "indivisible" statute — i.e. , one that "sets out a single ... set of elements to define a single crime" — we apply the categorical approach. See Mathis, 136 S. Ct. at 2248. That means we see whether the prosecution had to prove that the defendant used, attempted to use, or threatened to use physical force against the person or property of another — not whether he actually did. See id. And because we care only whether the prior crime requires physical force — not whether his criminal conduct involved physical force — we focus on the least forceful conduct generally criminalized under the statute, knowing that there must be a realistic probability the statute would be used to criminalize the conduct. See United States v. Báez-Martínez, 950 F.3d 119, 124 (1st Cir. 2020). And "physical force" here means "force capable of causing physical pain or injury" to a person or physical damage to property. See Johnson v. United States, 559 U.S. 133, 138-40, 130 S.Ct. 1265, 176 L.Ed.2d 1 (2010) (" Curtis Johnson"); see also Davis, 139 S. Ct. at 2325-26.

Alternatively, if the prior crime involved a violation of a "divisible" statute — i.e. , one that defines multiple crimes with distinct elements — we apply the modified categorical approach (if the statute simply lists different means of committing a single crime, then it is indivisible and we use the categorical approach). See Mathis, 136 S. Ct. at 2249. This approach allows us to look beyond the face of the statute to a limited set of documents — known as " Shepard documents," which include the indictment, jury instructions, and verdict forms — to see "what crime, with what elements," the defendant committed. See id. (discussing Shepard v. United States, 544 U.S. 13, 125 S.Ct. 1254, 161 L.Ed.2d 205 (2005) ); see also United States v. Delgado-Sánchez, 849 F.3d 1, 8 (1st Cir. 2017). But the approach "serves a limited function," namely, to "help[ ] effectuate the categorical analysis" when we are faced with a divisible statute — in other words, after reviewing the relevant documents and identifying the specific crime underlying the defendant's conviction, we must then apply the categorical approach to that crime to see if it is a crime of violence. See Descamps v. United States, 570 U.S. 254, 260, 133 S.Ct. 2276, 186 L.Ed.2d 438 (2013).

Days after the judge sentenced Dzhokhar — giving him death on some of the death-eligible counts and various concurrent and consecutive terms on the remaining counts (including 20 life terms) — the Supreme Court invalidated the Armed Career Criminal Act's similarly worded residual clause as unconstitutionally vague. See Johnson v. United States, 576 U.S. 591, 135 S. Ct. 2551, 2557, 192 L.Ed.2d 569 (2015) (" Samuel Johnson"). For easy reading, we shorten Armed Career Criminal Act to "ACCA." The ACCA's residual clause defined "violent felony" as any crime punishable by a term of imprisonment exceeding one year that "involves conduct that presents a serious potential risk of physical injury to another." See 18 U.S.C. § 924(e)(2)(B). In tossing out that residual clause, Samuel Johnson (in brief) found "[t]wo features of the [ACCA's] residual clause" troublesome: it "leaves grave uncertainty about how to estimate the risk posed by a crime" and "about how much risk it takes for a crime to qualify as a violent felony." See 135 S. Ct. at 2557-58 (emphasis added). Relying on Samuel Johnson, Dzhokhar moved for a judgment of acquittal on all of the § 924(c) counts. He also asked for a new penalty-phase trial as well. According to his motion, the judge had told the jury (without objection) that all of the "predicate" offenses — malicious destruction of property, for example, or conspiracies to use a weapon of mass destruction, to bomb a place of public use, and to maliciously destroy property — constituted crimes of violence as a matter of law. But, he noted, the judge did not say which of § 924(c)'s clauses applied to which predicate. Insisting that the government could no longer rely on the residual clause after Samuel Johnson, he also claimed that none of the predicates categorically qualified as a crime of violence under the elements clause.

Opposing the motion, the government argued first that Dzhokhar had waived his challenge to the § 924(c) counts by not raising it sooner. The government premised this argument on two theories: that Dzhokhar had to raise defects in the indictment before trial and that he had to object to the judge's crime-of-violence instructions either before or after the judge gave them. Waiver aside, the government also argued that the different wordings between § 924(c)'s residual clause and the ACCA's residual clause made Samuel Johnson's void-for-vagueness analysis inapplicable to Dzhokhar's case. And relying on Curtis Johnson, the government insisted that the predicates qualified as crimes of violence under the elements clause because they involved the use, attempted use, or threatened use of violent physical force against the person or property of another.

The judge denied Dzhokhar's motion, finding § 924(c)'s residual clause not impermissibly vague and each contested predicate a crime of violence under the elements clause. In a footnote, the judge theorized how Dzhokhar may have waived any argument that the predicates failed to satisfy the elements clause. But the judge did not resolve the waiver issue because he found no error.

During the briefing phase of this appeal, the Supreme Court issued an opinion declaring § 924(c)'s residual clause overly vague. See Davis, 139 S. Ct. at 2336. With Davis on the books, that leaves only one potential path for treating the predicates as crime-of-violence offenses: the elements clause, a provision (as we said) that sweeps in crimes having "as an element the use, attempted use, or threatened use of physical force," see 18 U.S.C. § 924(c)(3)(A) — i.e. , "force capable of causing physical pain or injury" to a person or physical damage to property, see Curtis Johnson, 559 U.S. at 140, 130 S.Ct. 1265.

Basic Appellate Arguments

On appeal Dzhokhar limits his challenge to five § 924(c) convictions involving Counts 13, 15, 16, 17, and 18. Counts 13 and 15 alleged as predicates the malicious destruction of property, colloquially known as arson, resulting in death (as charged in Counts 12 and 14, respectively). See 18 U.S.C. § 844(i). And Counts 16, 17, and 18 alleged as predicates conspiracies to use a weapon of mass destruction, to bomb a place of public use, and to maliciously destroy property, all resulting in death (as charged in Counts 1, 6, and 11, respectively). See 18 U.S.C. §§ 2332a(a)(2), 2332f(a)(1) and (2), 844(i) and (n).

In essence, Dzhokhar's position boils down to this. Arson — the predicate crime for Counts 13 and 15 — fails to satisfy the elements clause because, first, one can commit the offense by maliciously destroying "any" property, including one's own and so does not require as an element that force be used against the person or property of another, as the elements clause requires; and second, one can commit the crime with a reckless mental state but the elements clause demands intentional conduct. Arguing further, Dzhokhar contends that the challenged conspiracies — the predicate crimes for Counts 16, 17, and 18 — fail to satisfy the elements clause because conspiracies criminalize mere agreements to commit an act and thus do not necessarily have as an element the actual, attempted, or threatened use of physical force.

The government responds, essentially, this way. It agrees that malicious destruction of property "simpliciter ... is not categorically a crime of violence." It admits that under our current precedent "reckless conduct, as opposed to intentional conduct, cannot constitute the use of force against the person or property of another." It accepts that "conspiring to commit a violent act does not necessarily have as an element the use, attempted use, or threatened use of physical force." And it consents to our vacating of Count 18 — predicated on conspiracy to commit arson — albeit on grounds different from those offered by Dzhokhar (thus sparing us the need to discuss Count 18 further).

But the government insists that when the indictment charges arson as a capital crime, "the jury must find as an element" at least one of the FDPA's gateway-intent factors — each of which "requires proof that the defendant engaged in intentional conduct that directly resulted in a victim's death," meaning he used a level of force required under the elements clause. It also insists that the death-resulting allegations "independently require[ ] proof that the victim was subjected to ‘physical force’ " as used in the elements clause. And it takes a similar approach with the remaining conspiracy predicates, claiming that the death-resulting allegations establish the type of force needed to satisfy the elements clause.

Analysis

The parties spend some time addressing our standard of review. Dzhokhar argues for a de novo appraisal, noting that we typically evaluate judgment-of-acquittal and crime-of-violence assessments without giving any deference to the district judge's views. See United States v. Santos-Soto, 799 F.3d 49, 56 (1st Cir. 2015) (judgment of acquittal); United States v. Turner, 501 F.3d 59, 67 (1st Cir. 2007) (crime of violence). The government pushes for plain-error review, repeating the waiver arguments it made in the district court: that Dzhokhar had to — but did not — raise the crime-of-violence issue pretrial or object to the crime-of-violence instructions either before or after the judge gave them.

We think Dzhokhar has the better of this standard-of-review exchange. United States v. Cruz-Rivera concluded that a defendant's judgment-of-acquittal motions preserved his § 924(c) predicate-offense challenge. 904 F.3d 63, 65 (1st Cir. 2018). And Cruz-Rivera did so even though the defendant had not moved to dismiss the indictment or objected to the jury charge instructing that the at-issue predicate constituted a crime of violence as a matter of law. See Br. for Appellee at 10, Cruz-Rivera, 904 F.3d 63 (No. 16-1321), 2018 WL 3035960, at *9-10 ; Br. for Appellant at 20, Cruz-Rivera, 904 F.3d 63 (No. 16-1321), 2018 WL 3261713, at *20. The government tries to downplay the importance of this decision by saying " Cruz-Rivera ... did not definitively opine on" the waiver question because "the government never challenged the preservation of the claim." When we give de novo review to an unpreserved claim because the government failed to argue plain error to us, we say so. See, e.g., United States v. Blewitt, 920 F.3d 118, 122 n.2 (1st Cir. 2019) (quoting United States v. Encarnación-Ruiz, 787 F.3d 581, 586 (1st Cir. 2015) ). But Cruz-Rivera said nothing of the sort — it only said that the defendant had "preserved this issue below." 904 F.3d at 65. So de novo review is called for.

To the merits then.

First up is whether Dzhokhar's arson convictions (on Counts 12 and 14) satisfy the elements clause. The arson statute at issue punishes the use "of fire or an explosive" to "maliciously damage[ ] or destroy[ ] ... any ... property used in interstate or foreign commerce or in any activity affecting interstate or foreign commerce." 18 U.S.C. § 844(i) (emphasis added). And the parties agree (or at least do not dispute) that "maliciously" there includes both intentional and reckless acts. See generally United States v. Grady, 746 F.3d 846, 848-49 (7th Cir. 2014) (adopting this definition and collecting circuit cases doing the same).

Our caselaw says that recklessness does not suffice the ACCA's materially identical elements clause. See Báez-Martínez, 950 F.3d at 126 (discussing our bright-line rule that "reckless conduct bereft of an intent to employ force against another falls short of the mens rea required under" the ACCA (emphasis removed and citation omitted)). And our caselaw routinely uses decisions interpreting the ACCA's elements clause in construing § 924(c)'s, see Taylor, 848 F.3d at 491 — no surprise, since both clauses encompass "the use, attempted use, or threatened use of physical force against the person ... of another," compare 18 U.S.C. § 924(e)(2)(B)(i), with id. § 924(c)(3)(A). Which perhaps explains why the government concedes that crimes requiring recklessness, as opposed to intent, do not qualify as crimes of violence under § 924(c)'s elements clause — at least under existing circuit precedent.

Our caselaw is also clear about what happens next. Applying the minimum-conduct rule (as a reminder, the elements-based approach focuses on "the least culpable conduct" criminalized, Báez-Martínez, 950 F.3d at 124 ), we must presume that Dzhokhar acted with recklessness, see Taylor, 848 F.3d at 492. So — as counterintuitive as it might first seem — his arson convictions are not crimes of violence for purposes of § 924(c)'s elements clause.

And none of the government's responses alters this conclusion.

The government argues that "[w]here ... arson is charged as a capital offense, the jury must find as an element at least one of the four 'gateway' special intent factors" in the FDPA. These factors, says the government, require proof that the defendant intentionally engaged in conduct that resulted in a victim's death and thus proof that he "intentionally used force sufficient to kill the victim."

This aspect of the government's response overlooks that the gateway factors are drawn from the FDPA, not § 844(i) itself. Under either the categorical or modified categorical approach, we generally look to the statute of conviction to determine the elements of the crime. See, e.g., Mathis, 136 S. Ct. at 2248. And nowhere in § 844(i) does there appear an intent element. The government has pointed us to no authority suggesting that we can look beyond the statute of conviction to an unrelated statutory scheme — like the FDPA — to add elements to a crime for these purposes. See Taylor, 848 F.3d at 491 (explaining that " ‘[e]lements’ are the ‘constituent parts’ of a crime's legal definition" (alteration in original) (quoting Mathis, 136 S. Ct. at 2248 )). To convict Dzhokhar on the arson offenses (Counts 12 and 14) — the predicates for the contested § 924(c) counts (Counts 13 and 15) — the jurors did not have to find any of the gateway-intent factors. Instead, they could convict even if he acted recklessly rather than intentionally. Had the penalty-phase jurors not found the gateway factors proven beyond a reasonable doubt as to the arson charges, Dzhokhar could not have gotten a judgment of acquittal on those counts (Counts 12 and 14); indeed, the indictment on those two counts does not reveal on its face that the government had to prove intent. Surely then those factors cannot be elements of the arson predicates.

The government next contends that the death-resulting allegations in Counts 12 and 14 provide an independent basis for us to conclude that the arson predicates satisfy § 924(c)'s elements clause. But even assuming without deciding that the death-resulting allegations are elements (the parties fight over whether they are), we know the minimum conduct necessary to commit arson resulting in death is still recklessness. See generally United States v. Gullett, 75 F.3d 941, 947-48 (4th Cir. 1996) (deeming evidence of malice sufficient to convict the defendant of violating the arson statute, § 844(i), specifically rejecting his argument that the jury had to find that he intended to damage the property). As we will discuss shortly, the fact that death results (when included as an element of the statute of conviction) may indicate the application of violent force. But it does not necessarily involve the intentional application of physical force (i.e. , a "use" in the language of the elements clause) as our caselaw requires. See Bennett v. United States, 868 F.3d 1, 7–9 (1st Cir. 2017) (holding that recklessly causing bodily injury does not constitute the "use ... of physical force against the person of another"), opinion withdrawn as moot, 870 F.3d 34, 36 (1st Cir. 2017), reasoning adopted by United States v. Windley, 864 F.3d 36, 37 n.2 (1st Cir. 2017). So the government's second basis for affirming these contested § 924(c) counts (Counts 13 and 15) is not compelling either.

Next up is whether Dzhokhar's conspiracy convictions (on Counts 1 and 6) satisfy the elements clause. Recall that prosecutors predicated the relevant § 924(c) counts (Counts 16 and 17) on his allegedly conspiring to use a weapon of mass destruction (Count 1), see 18 U.S.C. § 2332a(a)(2), and to bomb a place of public use (Count 6), see id. § 2332f(a)(1) and (2), each resulting in death. Section 2332a(a)(2) criminalizes anyone "who, without lawful authority, uses, threatens, or attempts or conspires to use, a weapon of mass destruction ... against any person ... within the United States," provided the "threat, attempt, or conspiracy[ ] would have affected interstate or foreign commerce." Section 2332f(a)(1) applies to anyone who "unlawfully delivers, places, discharges, or detonates an explosive ... in, into, or against a place of public use ... with the intent to cause death or serious bodily injury, or ... with the intent to cause extensive destruction of such a place." Section 2332f(a)(2) prohibits "attempts or conspirac[ies] ... under [ § 2332f(a)(1) ]." And "if death results" from these crimes, the statutes provide for punishment "by death or imprison[ment] for any term of years or for life." See id. § 2332a(a) ; see also id. § 2332f(c).

Helpfully, the parties agree that the at-issue convictions concern conspiracies to use a weapon of mass destruction and to bomb a place of public use (not attempts to do either crime, for example), with death resulting. Our "task" then "is to compare" the elements of those conspiracies "to the definition of a 'crime of violence' in the force clause." See United States v. García-Ortiz, 904 F.3d 102, 106 (1st Cir. 2018) (using the categorical approach where the parties "agree[d]" that the defendant's "conviction concerned Hobbs Act robbery (not extortion)"). So the question is whether the at-issue conspiracy offenses have as an element the use, attempted use, or threatened use of "violent [physical] force — that is, force capable of causing physical pain or injury to another person." See Curtis Johnson, 559 U.S. at 140, 130 S.Ct. 1265 ; see also Moncrieffe v. Holder, 569 U.S. 184, 190-91, 133 S.Ct. 1678, 185 L.Ed.2d 727 (2013) (requiring courts to consider whether the least serious form of the relevant offense meets that standard). If it does, then the relevant conspiracies qualify categorically as crimes of violence — if not, then not.

A conspiracy is — as the parties concur — a knowing agreement between two or more people "to commit a crime, intending that the underlying offense be completed." See United States v. Ledée, 772 F.3d 21, 32 (1st Cir. 2014). The crime of conspiracy is the agreement rather than the completed offense. See Iannelli v. United States, 420 U.S. 770, 777, 95 S.Ct. 1284, 43 L.Ed.2d 616 (1975) (explaining that "[c]onspiracy is an inchoate offense, the essence of which is an agreement to commit an unlawful act"). So "conspiracy's elements are met as soon as the participants have made an agreement." Sessions v. Dimaya, ––– U.S. ––––, 138 S. Ct. 1204, 1219, 200 L.Ed.2d 549 (2018). Thus — to borrow a line from the government's brief (emphasis omitted) — "simply conspiring to commit a violent act does not necessarily have as an element the use, attempted use, or threatened use of physical force," meaning "most conspiracies to commit what would otherwise be crimes of violence are not categorically crimes of violence under" § 924(c)'s elements clause.

But the "death results" element changes things. Báez-Martínez says that any crime for which "death results" (or any serious bodily injury results) is an element automatically satisfies the ACCA's "violent force" requirement. 950 F.3d at 132. So while most conspiracies are not crimes of violence, conspiracies that are categorically defined to result in death are (assuming the other requirements like intent are satisfied). And here, the statute makes "death results" an element of the crime. Section 2332a says that "if death results" from a violation of the at-issue conspiracy statutes, the punishment may be up to life in prison or death. 18 U.S.C. § 2332a(a). Section 2332f incorporates this penalty scheme. Id. § 2332f(c).

Dzhokhar argues that "death results" is not an element of § 2332a or § 2332f under the crime-of-violence categorical approach because, like the FDPA gateway factors, that element need only be proven to the jury at the penalty phase. But unlike the FDPA gateway factors, the "death results" element appears in the statute of conviction itself (or is incorporated into that statute, for § 2332f ). True, as Dzhokhar suggests, a (guilt-phase) jury could have convicted him under § 2332a or § 2332f without deciding that anyone died, and those convictions would stand even if the (penalty-phase) jury found that no deaths resulted. But these statutes, it seems to us, are divisible into two branches: one in which there is no "death results" element (and the penalty is up to life in prison), and one in which "death results" is an element (and the penalty can be death). See Mathis, 136 S. Ct. at 2256 (noting that "[i]f statutory alternatives carry different punishments, then under Apprendi they must be elements"). Yet we know Dzhokhar's conduct falls into the latter branch. And this we know from the indictment — which for Counts 1 and 6 says that the conspiracy resulted in the death of at least one person; and from the jury's guilt-phase verdict — which found beyond a reasonable doubt that the conspiracies resulted in at least one death. So under the modified categorical approach, the predicate offenses (Counts 1 and 6) are crimes of violence. And thus his convictions on Counts 16 and 17 must stand.

Our use of the modified categorical approach here aligns with the purpose behind that doctrine. The Supreme Court designed the categorical and modified categorical approaches to simplify the types of evidence we can look to in making a crime-of-violence assessment. See Taylor v. United States, 495 U.S. 575, 601, 110 S.Ct. 2143, 109 L.Ed.2d 607 (1990) (recognizing the "practical difficulties and potential unfairness of a factual approach"); see also Shepard, 544 U.S. at 17, 125 S.Ct. 1254. Were we to go beyond these Shepard documents, we could find ourselves lost in a sea of evidence presented at trial to the jury. And we might never be able to tell whether certain facts were proven beyond a reasonable doubt. But our analysis requires no guesswork, for (again) the Shepard-approved documents and the text of § 2332a and § 2332f indicate that at least one person died as a result of Dzhokhar's involvement in the conspiracy.

In a different context, we have noted that an indictment's death-resulting references "invoked" a statute's "sentencing regime," increasing "the maximum sentence available," and so is "pertinent only to sentencing ." United States v. Hilario-Hilario, 529 F.3d 65, 69 (1st Cir. 2008) (emphasis added). And except for a fact of a prior conviction, any fact that boosts a crime's maximum sentence or minimum sentence must be established beyond a reasonable doubt to a jury's satisfaction (unless the defendant agrees to a bench trial or formally admits the facts). See Apprendi, 530 U.S. at 490, 120 S.Ct. 2348 (maximum); Alleyne v. United States, 570 U.S. 99, 103, 133 S.Ct. 2151, 186 L.Ed.2d 314 (2013) (minimum); see also United States v. Gonzalez, 949 F.3d 30, 41-42 (1st Cir. 2020). See generally Burrage v. United States, 571 U.S. 204, 210, 134 S.Ct. 881, 187 L.Ed.2d 715 (2014) (stating that "[b]ecause the 'death results' enhancement [in 21 U.S.C. § 841(b) ] increased the minimum and maximum sentences to which [the defendant] was exposed, it is an element that must be submitted to the jury and found beyond a reasonable doubt"). Dzhokhar's reply brief touches on these points, at least inferentially. But while an additional sentencing element — like § 2332a's "death results" — would be "pertinent only to sentencing" for most purposes, see Hilario-Hilario, 529 F.3d at 69, for purposes of the modified categorical approach, we think here that it is right to consider this as an element of the crimes of conviction, see Mathis, 136 S. Ct. at 2256. CONCLUSION

Having completed our review, the net result is this: We reverse Dzhokhar's convictions on Counts 13, 15, and 18, with directions to acquit. And we vacate his death sentences on Counts 4, 5, 9, 10, and 14, with directions to hold a new penalty-phase trial consistent with this opinion and with Local Rule 40.1(k)(1) of the District of Massachusetts. But make no mistake: Dzhokhar will spend his remaining days locked up in prison, with the only matter remaining being whether he will die by execution.

TORRUELLA, Circuit Judge (Concurring in part, Joining in part, Concurring in Judgment).

I agree with the lion's share of the majority's reasoning and join all its holdings. I regretfully must write separately, however, to express my disagreement with its handling and tentative conclusion of Tsarnaev's claim that he could not receive a fair trial by an impartial jury in this venue.

Tsarnaev properly raised this issue with our blessing, and it therefore requires -- and deserves -- a straight answer. In my view, the district court's rulings on Tsarnaev's motions for transfer of the trial venue, affirmed on intermediate appeal by this court and tentatively adopted by the majority, was patently incorrect. Furthermore, the issue of unduly prejudicial pretrial publicity is likely to recur with more frequency in this modern day of technology. If an accused's Fifth and Sixth Amendment rights are to be other than a hollow platitude, it is imperative that this court's jurisprudence establish a realistic standard for cases such as this one, in which a steady stream of information by way of myriad sources inundated an already deeply affected community. If this case did not present a sufficient basis for a change of venue, there are no set of circumstances that will meet this standard, at least not in the First Circuit.

Let me be clear that at the sentencing retrial of this case, if the issue of venue is again raised, Tsarnaev will have to allege and prove prejudicial circumstances at the time of his motion, likely nearly a decade after the crime was committed. But that is a horse of another color. The question before us that must be decided is whether Boston was the appropriate venue for Tsarnaev's trial in 2015.

I. Discussion

In denying Tsarnaev's second mandamus petition, the mandamus majority assured that it "reviewed the entire voir dire conducted to this point by the [district] court and the parties," and that "the process ha[d] been thorough and appropriately calibrated to expose bias, ignorance, and prevarication." In re Tsarnaev ("Tsarnaev II"), 780 F.3d 14, 24-25 (1st Cir. 2015) ; see id. at 24 (noting that "[t]he careful selection process and the trial judge's expressed confidence in finding sufficient jurors ... is supported by the record," and that "[the voir dire process] is working to ferret out those jurors who should appropriately be excused for cause"), 26 (declaring that "the careful process employed by the district court ... ha[s] afforded [it] ‘a sturdy foundation to assess fitness for jury service.’ " (quoting Skilling v. United States, 561 U.S. 358, 395, 130 S.Ct. 2896, 177 L.Ed.2d 619 (2010) )), 26-28 (describing the district court's efforts to "explore, and eliminate, any prejudice" as "rigorous," "extensive," and "careful"). Today, this court reverses course and finds the district court's juror inquiry lacked adequate safeguards, a finding with which I fully agree.

The majority's reason for reaching that conclusion, however, misses the forest for the trees. Although I agree that jury selection in this case failed to comply with this court's mandate in Patriarca v. United States, 402 F.2d 314, 318 (1st Cir. 1968), the fact of the matter is that "[n]o amount of voir dire [could have] overcome th[e] pervasive prejudice" against Tsarnaev in the Eastern Division of the District of Massachusetts, "no matter how carefully it [was] conducted." Tsarnaev II, 780 F.3d at 30 (Torruella, J., dissenting). The district court's denials of Tsarnaev's motions for change of venue amount to an abuse of discretion and denied Tsarnaev the right to a fair trial and sentencing determination.

A. This Panel Should Address Venue

To decide this case on Patriarca grounds, the majority puts its weight on the mandamus majority's expectation that a searching voir dire would be conducted. See op. at 50, 61–62. Yet, this was not the only assurance that drove the mandamus majority's denial in Tsarnaev II. That venue-change-denial also heavily relied upon the assumption that, should Tsarnaev be convicted, he would "have the opportunity to raise a challenge based on a lack of a fair and impartial jury on direct appeal." 780 F.3d at 18. "Indeed, that is the customary mechanism by which such challenges are presented and assessed." Id. at 18-19 ; see also id. at 29 ("[M]ost importantly, ... the petitioner remains able to raise claims of lack of an impartial jury on direct appeal."). Because this "double layer of review is itself a guarantee of due process," the mandamus majority concluded, Tsarnaev could not make a showing of the "irreparable injury" necessary to warrant mandamus relief. Id. at 28-29.

The question of whether venue was proper in the Eastern Division is a preliminary matter that precedes all others raised in this case -- be it jury selection, the exclusion of mitigating evidence, or any evidentiary challenge. This is so because the impropriety of venue is a primordial prejudicial error; should a presumption of prejudice be warranted, the district court should not have embarked in jury selection (or made any evidentiary rulings) in the first place. Cf. Rideau v. Louisiana, 373 U.S. 723, 727, 83 S.Ct. 1417, 10 L.Ed.2d 663 (1963) (finding a presumption of prejudice "without pausing to examine a particularized transcript of the voir dire"); United States v. Casellas-Toro, 807 F.3d 380, 389 (1st Cir. 2015) (looking to jury selection only after assuming the presumption of prejudice is rebuttable). The administration of justice demands that the question of venue be resolved.

Despite claiming not to decide the venue issue, by one sidedly laying out the government's arguments as to venue and then proceeding to find Patriarca error, see op. at 54–62, the majority implicitly and effectively resolves the issue in the government's favor -- tacitly finding that venue would have been proper in the Eastern Division had the district court conducted an adequate voir dire. For the following reasons, I cannot agree. B. Venue in the Eastern Division was Improper

The physical and emotional wake of the Boston Marathon bombings, and the events of the following week, flooded the residents of the Eastern Division with sorrow, fear, and anger. Few crimes have been as offensive and devastating to an entire community than those committed by the Tsarnaev brothers. But for even the most heinous of offenses, our system of justice demands vigorous protection -- both in appearance and fact -- of a defendant's right to a fair trial and sentencing. "[T]he right to jury trial guarantees to the criminally accused a fair trial by a panel of impartial , indifferent jurors. The failure to accord an accused a fair hearing violates even the minimal standards of due process." Irvin v. Dowd, 366 U.S. 717, 722, 81 S.Ct. 1639, 6 L.Ed.2d 751 (1961) (emphasis added) (internal quotations marks omitted).

1. A Presumption of Prejudice was Warranted

Article III of the United States Constitution instructs that criminal trials "shall be held in the State where the said Crimes shall have been committed[.]" U.S. Const. art. III, § 2, cl. 3. The Sixth Amendment further directs that a criminal defendant be tried by a jury "of the State and district wherein the crime shall have been committed." U.S. Const. amend VI.

Sometimes in tension with this directive is a criminal defendant's Sixth Amendment right to an impartial jury and Fifth Amendment promise of fundamental fairness. One such circumstance is when "extraordinary local prejudice" will prevent a fair trial in the judicial district in which the crime was committed. Skilling, 561 U.S. at 378, 130 S.Ct. 2896. Where these constitutional provisions collide, Article III's venue dictate must give way. Accordingly, "[u]pon the defendant's motion, the court must transfer the proceeding against that defendant to another district if the court is satisfied that so great a prejudice against the defendant exists in the transferring district that the defendant cannot obtain a fair and impartial trial there." Fed. R. Crim. P. 21(a).

"The theory of our [trial] system is that the conclusions to be reached in a case will be induced only by evidence and argument in open court, and not by any outside influence, whether of private talk or public print." Skilling, 561 U.S. at 378, 130 S.Ct. 2896 (alteration in original) (quoting Patterson v. Colorado ex rel. Att'y Gen. of Colo., 205 U.S. 454, 462, 27 S.Ct. 556, 51 L.Ed. 879 (1907) (opinion for the Court by Holmes, J.)). Where "pervasive pretrial publicity has inflamed passions in the host community past the breaking point" and "permeat[es] the trial setting ... [such] that a defendant cannot possibly receive an impartial trial," the district court must presume local prejudice and transfer the proceeding. United States v. Quiles-Olivo, 684 F.3d 177, 182 (1st Cir. 2012) ; see also Sheppard v. Maxwell, 384 U.S. 333, 362, 86 S.Ct. 1507, 16 L.Ed.2d 600 (1966) ("Due process requires that the accused receive a trial by an impartial jury free from outside influences."). A presumption of prejudice is "reserved for those extreme cases where publicity is both extensive and sensational in nature." Quiles-Olivo, 684 F.3d at 182 (internal quotation marks omitted) (quoting United States v. Misla-Aldarondo, 478 F.3d 52, 58 (1st Cir. 2007) ).

We have once again been tasked with determining whether the effects of these tragic events, coupled with the unrelenting pretrial publicity, caused such extraordinary local prejudice that Tsarnaev could not receive a fair trial and sentencing determination. In more than forty-five years on the bench at both the trial and appellate levels, and in my years of practice before that, I have never borne witness to a case with pretrial publicity more "extreme" or "extraordinary" than this one -- with so great a potential for jury determinations induced by "outside influence." A presumption of prejudice was warranted.

a. The residents of the Eastern Division were neither impartial nor indifferent

The impact on the residents of the Eastern Division of the defendant and his brother's week-long reign of terror, and the extraordinary outpouring of unity and resilience that followed, quite understandably left the residents of the Eastern Division neither impartial nor indifferent. The majority opinion has detailed but a fraction of heart-wrenching destruction, pain, and suffering inflicted by the Tsarnaev brothers through their crimes. I do not believe it necessary to further elaborate on these harrowing details. Suffice it to say, a detailed read of the record touches even the most detached of readers.

Although the impact of the defendant's crimes was felt nationally and internationally, the destruction was acutely felt by the residents of the Eastern Division. The prospective jurors and their loved ones, and the communities themselves, were all victims of these disturbing acts of terror. In addition to those killed and maimed by the bombings, millions in Greater Boston witnessed firsthand the carnage at the finish-line, knew someone directly impacted by the bombings, were ordered to shelter in place, had their houses searched by law enforcement with weapons drawn, saw their neighborhoods occupied by military personnel, or were otherwise affected by the events. The physical, psychological, and emotional trauma of these events was long felt locally. While others around the country may have viewed the marathon bombings "as an attack on all of America," op. at 56, to the residents of the Eastern Division, the Boston Marathon bombings were an attack on them. In the wake of that distressing April week, the residents of Greater Boston rallied together as never before to support each other. Immediately after the bombings, residents watching the marathon worked alongside first responders to treat the injured. Runners that had finished the race and countless local citizens rushed to the hospitals to donate blood -- so much so that hospitals had to turn people away. Others cared for the injured for months and years following the bombings. And to a previously unparalleled extent, the community participated in the identification and capture of the two bombing suspects.

Memorials, commemorations, and fundraisers to support the victims began soon after the finish-line attacks. Amongst many others, all four of Boston's major sports teams played host to these events. A week later, iconic Red Sox designated hitter David Ortiz exclaimed to a sold-out Fenway Park, "[t]his is our fucking city, and nobody is going to dictate our freedom. Stay strong."

Perhaps enhanced by Ortiz's comments, the ubiquitous and inspiring "BOSTON STRONG" campaign grew rapidly as an impressive expression of "defiance, solidarity, and caring." The blue background with yellow lettering (borrowing from the colors of the Boston Athletic Association, the organizers of the Boston Marathon race) was emblazoned on buildings, fences, fields, and bodies all over the Greater Boston metropolitan area. The campaign reflected a sense of compassion, unity, and recovery much needed in a community reeling from its upheaval. See Tsarnaev II, 780 F.3d at 25 n.13 ("[T]he Boston Strong theme [was] about civic resilience and recovery."). "BOSTON STRONG" reflected "all of us coming together as a city," one member of the venire aptly noted. As then-Boston Police Commissioner Edward F. Davis, III, told Congress:

These two terrorists tried to break us. What they accomplished was exactly the opposite. They strengthened our resolve, causing us to band together as a city and a Nation in time of crisis, to help

one another during life changing moments, to allow heroes to emerge and to prove to Bostonians and to the world, that our city is, indeed Boston Strong.

Prospective jurors (including those in the venire) purchased "BOSTON STRONG" merchandise, attended fundraisers and concerts to raise money for the victims, or donated directly to the One Fund Boston. The slogan, and what it stood for, became forever ingrained in the community psyche.

Underlying this awesome showing of resilience was its root cause: "deep[ ] personal grief, [and] a sense of loss forged by years of Patriots Day celebrations and the cherished ritual of cheering the runners on." "From Hopkinton to Boston, ... the bombings hit wrenchingly close to home and left many forlorn and adrift." Residents struggled to make sense of what had happened, to themselves and their neighbors, loved ones, and communities. They knew that things would never be the same. Widely shared amongst the Eastern Division was a feeling of sorrow, and that each day "[w]e are all just doing the best we can."

Just as the victims of other crimes (and their loved ones) cannot be "indifferent" or "impartial" for purposes of their wrongdoer's trial, despite any declarations to the contrary, neither here were the residents of the Eastern Division. Thus, the Fifth and Sixth Amendments required that they not be seated on Tsarnaev's jury.

b. Local pretrial publicity was both extensive and sensational

In a district still suffering physical and emotional trauma, the pretrial publicity enhanced its effects. Indisputably the volume, depth, and duration of the media coverage, from the bombings to Tsarnaev's capture, and well beyond, was nothing short of extraordinary. Nor can one challenge the sensational nature of the pretrial publicity -- including the horrific sights and sounds at the marathon finish line, the ensuing manhunt and lockdown of a million people, and the removal of a bloodied Tsarnaev from a boat in Watertown. This coverage included photographic and video footage of the crime being committed, maimed victims with bones protruding from their bodies, and the marathon finish line covered in blood. Every moment of the search for the suspects was live streamed. And newspapers and magazines documented Tsarnaev's confessions (both written in the boat and at the hospital, the latter of which he gave without the benefits of Miranda warnings, despite his request for a lawyer, and which thus would not be offered into evidence).

Coverage of the marathon bombings and its aftermath received international attention -- the who, what, where, and when of the crimes themselves widely covered. For anyone exposed to the media spectacle (as 99.7% of the venire was), or anyone with a smartphone and social media account, the pretrial publicity was "in a very real sense" the guilt-phase of Tsarnaev's trial. See Rideau, 373 U.S. at 726, 83 S.Ct. 1417 . For those in the Eastern Division, however, the media coverage was amplified. The disturbing images of maimed victims were broadcast on repeat. Many local residents were confined to their homes during the Governor's lockdown order, watching live footage of law enforcement scouring the city for the Tsarnaevs. After viewing this news coverage, "[a]ny subsequent court proceedings [about Tsarnaev's guilt] ... could be but a hollow formality." See id. Indeed, approximately two-thirds of prospective jurors admitted in court to having concluded that Tsarnaev was guilty of the charged crimes before seeing a single piece of evidence.

The majority adopts the government's view -- unsupported by the record -- that "this is not a case where almost everybody locally knows something and very few elsewhere know of it." op. at 55. Tsarnaev submitted volumes of articles from local newspapers that belie this assertion. Despite its recognition of these articles in its recantation of the facts, see op. 42–43, the majority seemingly ignores this detail in its venue analysis.

Whereas nationwide coverage of the bombing gradually waned over the following weeks and months, the record reflects that local media coverage did not. In Greater Boston, the scope of that reporting shifted from the facts surrounding the bombing to a focus on "the city as a whole[,] ... includ[ing] stories of the victims and their family and friends, those who bravely risked their lives to help the victims, and how the entire community came together." Tsarnaev II, 780 F.3d at 31 (Torruella, J., dissenting) (footnote omitted). News sources humanized the local victims and their families, describing in heart-wrenching and gruesome detail the emotional and physical struggles of the wounded surviving victims. Of the first responders, local newspapers (befittingly) wrote that "what every firefighter in the city[,] ... every cop, every EMS worker did[ ] ... was nothing short of heroic." Other journalists described how the Greater Boston community came together to mourn the deceased, honor the injured, and begin the collective healing process. Many of the articles, (rightfully) pointing to the defendant as the cause of the community's suffering, took to the use of negative descriptors -- including repeatedly calling him a "monster", a "terrorist," "depraved," "callous," "vile," "revile," and the "devil."

Tsarnaev's guilt preordained, reporters soon focused on whether Tsarnaev should be put to death -- prior even to the government's announcement of its intention to seek this outcome. This despite the fact that Massachusetts abolished capital punishment in its state courts in 1984, had not executed a criminal defendant for nearly forty years prior, and that the majority of residents of the Eastern Division had previously expressed general opposition to the death penalty. In this case, the media reported, even those who had previously opposed capital punishment admitted to being conflicted. Krystle Campbell's mother, Patricia Campbell, told the Boston Globe that she had been rethinking her longtime opposition to the death penalty because "an eye for an eye feels appropriate." Some of the amputees and their families were reported to have expressed similar sentiments. First responders and victims told reporters that a death sentence would "help everyone in their recovery." Mayor Menino, a proclaimed opponent of the death penalty, exclaimed, "in this one, I might think it's time ... that this individual serves his time and [gets] the death penalty." After he assumed office in January 2014, Boston Mayor Martin J. Walsh -- who had opposed the death penalty as a state representative -- expressed his support of Attorney General Eric Holder's "process that ... brought him to [the] decision" to seek capital punishment. Other politicians did the same, including both United States Senators from Massachusetts. In contrast to these more restrained endorsements, some expressed less hesitation about their support for Tsarnaev's execution. Former Boston Police Commissioners Edward F. Davis and William Evans, and MIT Police Chief John DiFava, all expressed their approval. And, although it was defeated in the Massachusetts House of Representatives, a bipartisan group of lawmakers used the bombing as support for a bill seeking to reinstate the death penalty.

Perhaps because of the nature of the crime, or because of its impact on them and their communities, the residents of the Eastern Division were inundated with reporting about this case.

c. Pride begets prejudice

Jury selection began in early January 2015. "BOSTON STRONG" continued to be proudly displayed throughout Greater Boston up to and through Tsarnaev's trial. Merchandise bearing the slogan continued to be sold at Boston Logan's International Airport. A banner displaying "BOSTON STRONG" was hung from a hotel nearby the courthouse, high above the surrounding buildings. And the drum of a cement truck parked directly across from the courthouse's visitor's entrance was decorated with "BOSTON STRONG" on one side and "THIS IS OUR CITY" on the other. A local Teamsters union continued to distribute "BOSTON STRONG" t-shirts and jackets to its members. Fundraising for the victims continued, and local road races placed the "BOSTON STRONG" logo on shirts distributed to its participants.

The district court dismissed the prevalence of these displays in various ways: although the defense team took the photographs of the banner and cement truck while jury selection was ongoing, the district court found the logo's appearances inconsequential because the cement truck photograph was taken on a day on which no empaneled jurors attended court, and the hotel banner was not visible at the juror's entrance to the courthouse or from the jury room ; that the association of "BOSTON STRONG" "weakened somewhat over time through overuse"; and that (quoting Skilling, 561 U.S. at 361, 130 S.Ct. 2896 ), "the decibel level of media attention [had] diminished somewhat." As jury selection in this case began less than two years after the bombings, the district court ignored, however, that the continued displays of "BOSTON STRONG" reflected an enduring community sentiment which formed the base of the movement, and a well-deserved pride of accomplishment in the community's efforts to return to normalcy. As one member of the venire put it, "BOSTON STRONG" was "the spirit of Boston, that despite whatever happens, ... we will continue."

A coming together remarkably similar to this one emerged in the wake of the 1995 bombing of the Alfred P. Murrah Federal Office Building in Oklahoma City, Oklahoma, which killed 168 people, injured hundreds more, and damaged numerous federal buildings. See United States v. McVeigh, 918 F. Supp. 1467, 1471-72 (W.D. Okla. 1996). Like "BOSTON STRONG," "Oklahoma family" became "a common theme" amongst the Oklahoma media and political leaders, emphasizing "how the explosion shook the entire state, ... how the state has pulled together ... as a family," and that "the survival and recovery from this tragedy is ‘Oklahoma's story.’ " Id. at 1471. Finding that the values of due process and fairness required that the trial of the Oklahoma City bombing suspects be transferred to Denver, Chief Judge Matsch pertinently described the profound potential for prejudice in this situation:

Pride is defined as satisfaction in an achievement, and the people of Oklahoma are well deserving of it. But it is easy for those feeling pride to develop a

prejudice ... [t]he existence of [which] is difficult to prove. Indeed it may go unrecognized in those who are affected by it. The prejudice that may deny a fair trial is not limited to a bias or discriminatory attitude. It includes an impairment of the deliberative process of deductive reasoning from evidentiary facts resulting from an attribution to something not included in the evidence. That something has its most powerful effect if it generates strong emotional responses and fits into a pattern of normative values.

Id. at 1472 (internal quotation marks and citation omitted). Similarly, the Supreme Court has noted that "[t]he influence that lurks in an opinion once formed is so persistent that it unconsciously fights detachment from the mental processes of the average man." Irvin, 366 U.S. at 727, 81 S.Ct. 1639.

Such was the state of the Eastern Division. Amongst an entire community so deeply affected by these crimes, the "intensity of the humanization of the victims" by the media, McVeigh, 918 F. Supp. at 1472, the heavy emphasis on grief, and the powerful portrayals of people struggling to make sense of this calamity, imprinted a pervasive and insurmountable prejudice in the community psyche. In such circumstances, an individual juror "may have an interest in concealing his own bias ... [or] may be unaware of it." Smith v. Phillips, 455 U.S. 209, 221-22, 102 S.Ct. 940, 71 L.Ed.2d 78 (1982) (O'Connor, J., concurring). And the unconscious nature of this impairment makes juror questionnaires and voir dire a poor means for assessing juror impartiality. See William H. Farmer, Presumed Prejudiced, But Fair?, 63 Vand. L. Rev. En Banc 5, 8 (2010).

The majority's "serious points against [Tsarnaev's] venue-change arguments" lean heavily on the pretrial polling data. See op. at 54–55. The majority finds this data convincing because the pollster did not "ask respondents to judge for themselves whether they are biased[ ]" and "[i]nstead ... asked whether Tsarnaev should get the death penalty." op. at 55. But, even though the pollster did not explicitly ask respondents if they were biased, it is the court's (and the majority's) dependence on this polling data to decide the venue challenge that leaves prospective jurors as the judges of their own impartiality. This is the very problem that the majority takes issue with in its Patriarca analysis. See op. at 57–58, 58–59, 59–60, 61–62. Further, the data cannot be relied upon to accurately identify local prejudice. Despite the majority's conclusory claim otherwise, see op. at 55, the survey does not -- and cannot -- account for the fact that the people who are most acutely affected by trauma and persistent media coverage thereof often lack awareness of the impact this exposure has to their decision-making capacity in the jury deliberation room -- particularly when they are being asked to make a decision as high-stakes as the appropriate punishment for the individual that wreaked havoc on their lives and those of their neighbors and communities. See McVeigh, 918 F. Supp. at 1473. That over 92% of Boston residents admitted in the community poll that they believed Tsarnaev was "definitely" or "probably" guilty based on their exposure to pretrial publicity, whereas 25% less admitted prejudgment in their juror questionnaires, highlights the inadequacy of survey polling to determine local prejudice in circumstances such as existed in the Eastern Division, and the potential inability of jury selection to sufficiently weed out the prejudices in this venue.

This "impairment of the deliberative process," particularly during a moment such as jury deliberations in a case of such prominence, is not easily quantifiable. Again, I invoke Chief Judge Matsch:

The possible prejudicial impact of this type of publicity is not something measurable by any objective standards.... [S]urveys are but crude measures of opinion at the time of the interviews. Human behavior is far less knowable and predictable than chemical reactions or other subjects of study by scientific methodology. There is no laboratory experiment that can come close to duplicating the trial of criminal charges.

Id.

Given the unknowing nature of the prejudicial effect of inflammatory pretrial publicity on severely afflicted people, see Irvin, 366 U.S. at 727, 81 S.Ct. 1639 ; McVeigh, 918 F. Supp. at 1472-73, and the great potential for equivocation by individual jurors impacted by these outside influences, see Smith, 455 U.S. at 221-22, 102 S.Ct. 940, it must be given little weight that polling data showed somewhat similar (but still lower) numbers regarding a proclivity for the death penalty in Springfield and New York than in Boston.

The majority next declares, again without record support, that "most of the publicity was true." op. at 55. Much of the local publicity included not only factual narrations of the events that transpired but also commentary and opinions, such as the aforesaid references to him as a "monster," "terrorist," "evil," the "devil," and other similar derogations. I have no doubt that other media sources either did the same or quoted others who did. Even if these are appropriate adjectives to describe Tsarnaev, this does not mean the descriptors did not have a prejudicial impact on the venire's ability to make decisions based solely on what was presented in court. See, e.g., Rideau, 373 U.S. at 725-26, 83 S.Ct. 1417 (televised confession was factual but prejudicial). The same is true of the reports that public officials believed Tsarnaev should die, and the detailed chronicling of the pain of the survivors and decedents' families. These reports may reflect reality, but the media's emphasis of these topics carried a significant risk of disturbing potential jurors' impartiality. This publicity was anything but "trivial."

I do not mean to imply that every juror was being disingenuous or deceitful in their self-declared impartiality. But our "[t]rust in their ability" to disregard this prejudicial information "diminishes when the prior exposure ... evokes strong emotional responses or such an identification with those directly affected by the conduct at issue that the jurors feel a personal stake in the outcome." McVeigh, 918 F. Supp. at 1473. Our own biases often go unrecognized or ignored. The risk of implicit biases in this case was impermissibly high, particularly with the defendant's life on the line.

d. The "Skilling" factors

The majority suggests that an application of the eponymous Skilling factors counsels against an abuse-of-discretion finding. See op. at 55–56. As I noted in my mandamus dissent, I find the comparison between this case and Skilling to be inapposite. See Tsarnaev II, 780 F.3d at 42 (Torruella, J., dissenting). Skilling involved "neither heinous nor sensational" facts, but rather white-collar economic crimes which impacted only a minority of potential jurors in the Eastern District of Texas. 561 U.S. at 370, 384, 130 S.Ct. 2896 (noting that the "jurors' links to Enron were either nonexistent or attenuated"); see also United States v. Skilling, 554 F.3d 529, 560 n.47 (5th Cir. 2009), aff'd in part, vacated in part, 561 U.S. 358, 130 S.Ct. 2896, 177 L.Ed.2d 619 (2010) (noting an opinion poll that found that one in three Houstonians knew someone harmed by what happened at Enron). Contrast that with the facts in this case: a terrorist attack with explosives at the iconic Boston Marathon; a widespread and crowdsourced search for the suspects; the execution of a police officer; a carjacking; a shootout on a suburban street; a shelter-in-place order; a televised manhunt; a standoff around the Watertown boat; the televised removal of a bloodied defendant from that boat; and a traumatized Eastern Division.

Even assuming arguendo the applicability of the Skilling factors to this terrorism case, I would find that they weigh in Tsarnaev's favor. First, the Skilling Court looked to the "size and characteristics of the community in which the crime occurred," noting that Houston was the fourth most populous city in the United States with a "large, diverse pool of potential jurors." 561 U.S. at 382, 130 S.Ct. 2896. Boston is not even in the top twenty in terms of population size. See United States Census Bureau, Population Division, Annual Estimates of City and Town Population Totals: 2010-2019 (May 21, 2020).

Equally if not more important are the characteristics of the community. See Casellas-Toro, 807 F.3d at 387 (commenting that, although Puerto Rico has a population of approximately three million, it is an "insular community that is highly susceptible to the impact of local media" and "seem[s] to be a small island" (internal quotation marks and citation omitted)). The Court in Skilling noted that only 12.3% of Houstonians were able to name Skilling as an Enron executive they believed guilty of crimes, and 43% had never heard of him. This case is again easily distinguishable. Here, "Tsarnaev and the Boston Marathon bombings [were] one and the same." Tsarnaev II, 780 F.3d at 42 (Torruella, J., dissenting); see also Def.'s Reply to Opp'n to Mot. to Change Venue at 4, United States v. Tsarnaev, No. 13-cr-10200-GAO (D. Mass. filed Aug. 7, 2014), ECF No. 461-23 (showing that approximately 90% of survey participants in Boston recognized the name Dzhokhar Tsarnaev, as opposed to 58% in Springfield, MA, 44% in Manhattan, and 34% in Washington, D.C.). For instance, 99.7% of the voir dire had been exposed to pretrial publicity about the case as every local news sources reported about it, and two thirds admitted in their juror questionnaires that they believed Tsarnaev to be guilty. Moreover, terrorism targets the very fabric of the community, seeking to tear it apart. We should thus be especially sensitive to the community response here as being indicative of its sense of communal victimhood -- a marker perhaps of its tightness as a judicial division. The Greater Boston area "band[ed] together" in response to this crisis, a "close-knit place" where "we grieve for [our neighbors]." At that moment, the Eastern Division was "BOSTON STRONG."

Second, the news stories in Skilling "contained no confession or other blatantly prejudicial information." 561 U.S. at 382, 130 S.Ct. 2896. This case involved both. Not only did the media print Tsarnaev's message in the boat admitting his crimes, it also reported information about his non-Mirandized hospital confession to the FBI. The news stories also contained blatantly prejudicial opinions that Tsarnaev should die.

Third, the Skilling Court looked to the media attention surrounding Skilling's crime and trial, noting that "the decibel level ... diminished somewhat" over the four years between Skilling's crime and his trial. Here, although there was a slight diminution of pretrial publicity over the twenty-one months between the bombings and the commencement of jury selection, the reporting continued to be omnipresent. Finally, the Skilling Court looked to the jury verdict, finding that the jury's not-guilty findings on nine of the twenty-eight counts in the case "yielded no overwhelming victory for the government." 561 U.S. at 375, 383, 130 S.Ct. 2896. Here, because Tsarnaev's counsel admitted Tsarnaev's guilt during opening and closing statements, the jury verdict finding Tsarnaev guilty on all thirty counts neither supports nor refutes a presumption of impartiality. See, e.g., Luong v. State, 199 So. 3d 139, 148 (Ala. 2014) ("[I]n light of the facts of this case, in particular Luong's admission that he threw each of his children off the bridge, the fact that Luong was not acquitted of any of the charged offenses does not either support or rebut a presumption of jury bias or impartiality.").

The majority's comparison of the nine acquittals in Skilling to Tsarnaev's jury's decision to recommend death for six of the seventeen death-eligible counts is mind-boggling. See op. at 55–56. There is a monumental distinction between the full acquittals in Skilling -- resulting in no punishment for that defendant, and this jury's decision about which of the two most extreme punishments in our criminal justice system to recommend for each count. The jury's decision to recommend that Tsarnaev receive six death sentences and serve eleven life sentences, instead of recommending that Tsarnaev be killed on seventeen separate counts, does not indicate a lack of prejudice. Far from it. A criminal defendant can only be put to death once. The Supreme Court has noted that the decision of whether to recommend a death sentence "is mostly a question of mercy." Kansas v. Carr, 577 U.S. 108, 136 S. Ct. 633, 642, 193 L.Ed.2d 535 (2016). As Tsarnaev points out, "six separate death sentences can hardly be considered an act of mercy such as to establish that jurors were either unaffected by the pretrial publicity or willing to ignore the community sentiment."

Tsarnaev was entitled to a presumption of prejudice.

2. The Government Cannot Overcome the Presumption of Prejudice

The parties quarrel over whether the presumption of prejudice is rebuttable. In Casellas-Toro, this court assumed without deciding that the presumption was rebuttable, and I follow the same track. 807 F.3d at 388-90. Yet, even under this assumption, the government cannot prevail.

The government argues that it can rebut that presumption by showing that the district court was able to ascertain the effects of the potential jurors' exposure to extensive pretrial publicity and excuse those that were incapable of setting any prejudice aside. Not so. As today's majority has explained, by refusing to ask prospective jurors content-specific questions about what they had read and heard, the district court was unable to identify biases or prejudices that may have resulted from that exposure. See Patriarca, 402 F.3d at 318. The district court relied on the venire's self-declarations of impartiality, an error of law and an abuse of discretion.

The government also cannot show that the district court seated an impartial jury because the district court failed to investigate Tsarnaev's "colorable" and "plausible" claims of juror-misconduct. When a defendant raises such a claim, regardless of timing, the district court has the "unflagging duty" to investigate. See United States v. French, 904 F.3d 111, 117 (1st Cir. 2018), cert. denied sub nom. Russell v. United States, ––– U.S. ––––, 139 S. Ct. 949, 203 L.Ed.2d 137 (2019) (quoting United States v. Zimny, 846 F.3d 458, 464 (1st Cir. 2017) ). The district court need not hold a full evidentiary hearing, but it must fashion and "even-handedly implement ... a sensible procedure reasonably calculated to determine whether something untoward had occurred." United States v. Paniagua-Ramos, 251 F.3d 242, 249-50 (1st Cir. 2001). The court's procedural discretion does not include a refusal to conduct any inquiry whatsoever. Zimny, 846 F.3d at 465.

But that is precisely what the district court did. Tsarnaev presented a colorable claim that Juror 286 knowingly withheld from the court the fact that she posted twenty-two online comments mourning the death of Martin Richard, praising law enforcement officers (three of whom would later testify at trial), expressing "BOSTON STRONG" civic pride, and calling Tsarnaev a "piece of garbage." Tsarnaev further showed that Juror 286 may have lied on her juror questionnaire and during voir dire about sheltering in place with her family. And Tsarnaev presented a second plausible claim that Juror 138 both refused to follow simple but important court rules and intentionally withheld from the court his participation in social media conversations. In the comments thread on his Facebook page, on which he continued to engage, his friend urged him to "[p]lay the part so u get on the jury then send [Tsarnaev] to jail where he will be taken care of." The district court's refusal to inquire -- at the government's behest -- left much to be desired in the way of certainty about these jurors' impartiality.

Finally, a review of the seated jurors' questionnaires and voir dire transcripts confirms that the government cannot rebut the presumption in this case. The government admits that there is a statistically significant correlation between the prospective jurors' media exposure and their opinions to guilt, and ten of the twelve seated jurors had been exposed to "a moderate amount" or "a lot" of publicity. Prior to trial, three of the twelve voting jurors admitted to having predetermined Tsarnaev's guilt, and another two stated that they believed Tsarnaev participated in the bombings. See Juror 83: "obviously he was involved in something"; Juror 229: "[f]rom the media" "I suppose that we knew that he was involved"; Juror 349: marked on juror questionnaire belief that Tsarnaev was guilty, and stated on voir dire that from "so much media coverage" "anybody would think that [Tsarnaev was involved]"; Juror 395: marked on juror questionnaire belief that Tsarnaev was guilty; Juror 487: marked on juror questionnaire belief that Tsarnaev was guilty, and explained on voir dire that this belief came from what he had seen on the news. Likewise, three of the six alternate jurors stated that they believed that Tsarnaev was involved. See Juror 552: "[F]rom the videos I saw, it appeared that he was part and parcel of perhaps depositing those devices"; Juror 567: "I do believe that he was somewhat involved"; Juror 588: "I believe there's some involvement somewhere, but I don't know what it is." And, as just discussed, two seated jurors were at least inconsistent -- if not deliberately untruthful -- in their responses during jury selection, calling into question their declarations that they had not yet formed an opinion about Tsarnaev's guilt. See Sampson v. United States, 724 F.3d 150, 164 (1st Cir. 2013) ("The voir dire process ... is frustrated when a prospective juror is dishonest. Both the juror's dishonesty and her motivation for that dishonesty may cast doubt upon her impartiality.").

Although all the seated jurors declared that they could be fair and impartial and decide the case solely on the evidence presented in court, little weight can be given to such declarations by community members so impacted by the crimes and the subsequent pretrial publicity. "Natural human pride would suggest a negative answer to [a question of] whether there was a reason the juror could not be fair and impartial." United States v. Dellinger, 472 F.2d 340, 375 (7th Cir. 1972) ; see also Irvin, 366 U.S. at 728, 81 S.Ct. 1639 ("[The] psychological impact requiring such a declaration before one's fellows is often its father."). In sum, the government cannot show that the jury that convicted Tsarnaev and recommended that he be put to death was impartial. The government has therefore failed to rebut the presumption of prejudice.

C. Harmless as to Guilt, Not as to Sentencing

I agree with the majority that the district court's abuse of discretion was harmless as to Tsarnaev's guilt but not as to his sentence. See op. at 61–62, 62 n.33. Tsarnaev's counsel admitted his participation in these crimes during both opening and closing statements, and he does not contend on appeal that he would not have made these admissions had his trial taken place elsewhere. Sitting in its expanded role under the Federal Death Penalty Act, 18 U.S.C. §§ 3591 - 99, a sentencing jury was required "to make a moral judgment ... after consideration of aggravating and mitigating circumstances" about whether Tsarnaev deserved to live out his natural life in custody or be killed by the government. See McVeigh, 918 F. Supp. at 1474. The question of whether these aggravating circumstances outweigh the mitigating factors is one of " mercy" for the sentencing jury. See Carr, 136 S. Ct. at 642.

With a jury so intensely impacted by the charged crimes, and so exposed to inflammatory pretrial publicity -- including reports detailing the extreme anguish of their neighbors and repeated calls for Tsarnaev to be sentenced to death -- I cannot say with any degree of certainty that the jurors did not possess a "predilection toward that penalty." McVeigh, 918 F. Supp. at 1474.

"With his life at stake, it is not requiring too much that [a criminal defendant] be tried in an atmosphere undisturbed by so huge a wave of public passion...." Irvin, 366 U.S. at 728, 81 S.Ct. 1639. The government cannot show that the district court's abuse of discretion was harmless beyond a reasonable doubt as to Tsarnaev's sentence. See 18 U.S.C. § 3595(c)(2), (c)(2)(C). For this reason, I agree that Tsarnaev's death sentences must be vacated and the case remanded to the district court for a sentencing retrial.

II. Closing Remarks

In dissent on Tsarnaev's second mandamus petition, I expressed my concern that -- in a case having this magnitude of press coverage and widespread dissemination of information -- a subsequent jury on retrial would have been exposed to the evidence (and results) of the first trial and would know that the new trial was the result of a post-conviction reversal. Tsarnaev II, 780 F.3d at 46 (Torruella, J., dissenting). My concern has and very likely will come to fruition. Because the majority rules in the manner in which it does on the issue of venue, I also maintain these concerns for future cases. The majority's reasoning cripples Rule 21(a) of the Federal Rules of Criminal Procedure and undermines the due process and impartiality principles of the Fifth and Sixth Amendments. I asked a simple question in 2015 that is still fitting, and I repeat it today: "If not here, when?" Id. at 45.