dissenting:
The majority errs both in affirming the conviction and upholding the death sentence of Lezmond Mitchell, a young Navajo tribe member found guilty of committing a robbery and two horrific murders on Navajo land in Arizona.1 During the investigation of these crimes, federal agents convinced eagerly cooperative tribal authorities to arrest Mitchell, who was then held in tribal custody for twenty-five days without counsel or arraignment. Federal agents repeatedly interrogated Mitchell during his tribal detention, and unlawfully secured from him a series of confessions. The inculpatory statements, which were obtained in violation of Mitchell’s federal rights, were then used to convict him at his trial. The trial court’s failure to suppress the confessions warrants reversal of Mitchell’s conviction.
Another significant error occurred during jury selection. The trial court allowed the prosecution to strike the only African-American juror, accepting its pretextual explanation that the juror was removed not because of his race but because more than twenty years earlier he had served on a jury that voted to acquit a defendant of unknown charges. Removal of this juror violated the constitutional guarantees of equal protection and due process, and, like the first error, requires reversal of Mitchell’s conviction.
Finally, the sentencing phase of Mitchell’s trial was rife with errors. Contrary to the provisions of the Federal Rules, the trial court permitted Mitchell to waive his presence' at the sentencing phase; as a result, the jury heard this critical part of the proceedings, and voted for the death penalty, without facing or hearing from the individual whose life, or death, it was determining. This error in itself requires that the sentence of death that the jury imposed be vacated. Moreover, the government’s closing sentencing argument was a compendium of improper statements, exacerbating the prejudice resulting- from Mitchell’s absence from the sentencing phase. Finally, the court erroneously failed to instruct the jury that the government had the burden of proving that aggravating factors outweighed mitigating factors. Considered cumulatively, the sentencing errors were sufficiently prejudicial that Mitchell’s death sentence must be vacated.
Therefore, although I agree with the majority’s resolution of a number of Mitchell’s other claims, I respectfully dissent.
I. Conviction
A. Suppression of Post-Arrest Statements
Following a conference that led to an agreement between all the involved federal and tribal authorities, Mitchell was arrested on tribal charges on the morning of November 4, 2001. He was not presented to a federal magistrate until November 29, 2001. During this period of prolonged tribal detention, federal authorities inter*998rogated him and obtained a set of increasingly incriminating statements. Mitchell argues that the district court should have suppressed these statements because they were obtained in violation of his federal right to timely arraignment and to counsel.2 In order to determine whether these federal rights attached during Mitchell’s detention, it is necessary to decide whether he was effectively in federal custody during the period between his tribal arrest and his federal arrest. The majority concludes that Mitchell was not in federal custody until he was formally arrested by federal agents on November 29. See ante, at 962. I disagree.
Unlike the majority, see ante, at 959-61, I would not review this issue for plain error. The Federal Rules of Criminal Procedure provide that “[a] party may preserve a claim of error by informing the court — when the court ruling or order is made or sought — of the action the party wishes the court to take, or the party’s objection to the court’s action and the grounds for that objection.” Fed. R.Crim.P. 51(b). Before trial, Mitchell moved for a hearing to determine whether his custodial statements were voluntary. At this time, Mitchell was still being tried jointly with co-defendants Johnny Orsinger and Gregory Nakai, both of whom participated in the voluntariness hearing. During the hearing, Orsinger’s counsel argued that the FBI controlled the operation leading to the defendants’ tribal arrest. He stated that the defendants were effectively in federal custody from the time of their tribal arrest and sought leave to file a new motion to suppress on the ground that Orsinger was deprived of procedural rights for juveniles established by 18 U.S.C. § 5033. Mitchell’s counsel adopted this argument, stating:
I would like to join in the remarks of [Orsinger’s counsel] about what I think has happened in this case. I think this case is illustrative to all of us about FBI strategies and techniques in cases like this. Rather than file federal charges and implicate not only the juvenile rights that [Orsinger’s counsel] said were violated with respect to Mr. Or-singer, but also the obvious rights that would have attended the filing of charges including the right to the immediate appointment of counsel for Mr. Mitchell, essentially Mr. Mitchell was stashed in the tribal jail for more than three weeks on these tribal charges which no one connected with this case believed were ever going to proceed.
Thus, Mitchell not only requested that the court suppress the custodial statements, but he also identified the failure to afford him his pretrial rights as one ground for suppression. This complies with the requirements of Rule 51(b).
Turning to the merits of Mitchell’s claim, Rule 5(a) of the Federal Rules of Criminal Procedure provides that a person who is arrested must be arraigned “without unnecessary delay.” Rule 5(a) both reflects and is reinforced by Supreme Court decisions requiring exclusion of pre-arraignment statements obtained in violation of the prompt presentment requirement. See McNabb v. United States, 318 U.S. 332, 341, 63 S.Ct. 608, 87 L.Ed. 819 (1943); Mallory v. United States, 354 U.S. 449, 455, 77 S.Ct. 1356, 1 L.Ed.2d 1479 (1957). This judicially developed exclusionary rule is limited by 18 U.S.C. § 3501(c), which created a “safe harbor” for statements obtained within six hours of detention.
In United States v. Alvarez-Sanchez, 511 U.S. 350, 114 S.Ct. 1599, 128 L.Ed.2d *999319 (1994), the Supreme Court held that § 3501(c) authorizes suppression of a custodial statement only when a person has been held in federal custody for more than six hours before giving the statement. Id. at 358, 114 S.Ct. 1599. However, the Court pointed to one circumstance in which delay would be viewed differently, “namely, the situation that would arise if state or local authorities, acting in collusion with federal officers, were to arrest and detain someone in order to allow the federal agents to interrogate him in violation of his right to a prompt federal presentment.” Id. at 359, 114 S.Ct. 1599. Mitchell argues that his arrest and interrogations fall within this exception and that the district court was therefore required to suppress his confessions.
To establish collusion, a defendant must show a collaborative effort by the two authorities involved and “proof of a deliberate intent to deprive a defendant of [his] federal procedural rights.” United States v. Michaud, 268 F.3d 728, 735 (9th Cir.2001) (citing United States v. Doe, 155 F.3d 1070, 1078 (9th Cir.1998)). Here, there is no question as to the former. As for the “deliberate intent” element, it is not necessary for the defendant to extract admissions from the federal and tribal authorities that such was their intent. Intent may be shown from the objective circumstances surrounding the decision-making process and from subsequent occurrences. Undisputed facts in the record show that an Assistant U.S. Attorney and FBI agents persuaded tribal authorities to arrest Mitchell and then used the tribal detention to interrogate him. Two agents testified that tribal arrest was a way to get suspects into custody quickly so that they could interrogate them.3 They also acknowledged that they were aware that taking a suspect into tribal custody would mean that he would not be entitled to certain federal rights, such as the right to appointed counsel and the right to a prompt arraignment. The agents offered no plausible explanation for why it was necessary to use a tribal arrest rather than simply to arrest Mitchell on federal charges. Instead, the circumstances surrounding Mitchell’s arrest and interrogation compel the conclusion that the delay in filing federal charges was the result of a deliberate decision to have the federal agents question him in violation of his rights.
Although the record does not include Mitchell’s tribal arrest warrant, review of the trial transcript indicates that Mitchell was most likely arrested for robbery.4 Criminal conduct that would permit a defendant to be charged with robbery under the Navajo Nation Code would also allow the defendant to be charged with that offense under federal law.5 The most significant difference in “robbery,” as defined *1000in the two codes, is the Navajo Nation Code’s requirement that the threat or use of force be immediate, an element not described in the federal definition, creating a higher burden than under federal law. The statements made by FBI Agents Raymond Duncan and Bradley Purscell, that federal agents had determined that there was insufficient evidence to obtain a federal search warrant, but sufficient evidence to obtain a tribal arrest warrant, are clearly inconsistent with the governing law. Agent Purscell acknowledged that robbery on Indian land is subject to federal jurisdiction and could not offer any explanation for suggesting the existence of any difference in probable cause needed for a tribal versus a federal arrest warrant.6 Agent Duncan said that they discussed trying to obtain a federal warrant for robbery but rejected that approach because they could not connect the robbery to a house. However, as described above, federal robbery does not include any element requiring that the offense take place in a house. Moreover, that the arrest for a tribal offense was a ruse arranged to allow federal officers to interrogate Mitchell without affording him the opportunity to consult counsel becomes evident from a comparison of the possible penalties under tribal and federal law. The maximum penalty for any offense under tribal law, including criminal homicide, is one year. See 17 N.N.C. §§ 223, 303(b). The maximum under federal law for robbery and homicide is fifteen years and death respectively. Accordingly, there can be little doubt that the authorities always intended that Mitchell and his co-defendants would be tried on federal rather than on tribal charges. Even if the district court’s finding that the agents’ testimony was credible were not clearly erroneous, and I think that it was, the circumstances surrounding the decision to take Mitchell into tribal custody establish collusion with the requisite intent, despite the claim that officials never explicitly discussed depriving defendant of his rights. Both FBI agents made it quite clear in their testimony that a significant motivation underlying the arrest was to be able to interrogate the defendant and that if he were in tribal rather than federal custody, the interrogation could be conducted without affording him substantial rights designed to protect him against self-incrimination.
Moreover, if federal agents did not have probable cause to obtain a federal warrant at the time of the November 3 meeting (in which case the tribal officers also lacked probable cause for the tribal arrest), the *1001agents clearly had probable cause to arrest Mitchell after November 4, when he confessed to participating in the robbery. However, he remained in tribal custody and was questioned further. A federal indictment was not filed until November 21. On November 29, FBI agents conducted at least one more extensive pre-arraignment interrogation precisely because they knew that a lawyer would be appointed shortly thereafter, when Mitchell would appear before a magistrate. This evidence establishes that Mitchell was effectively in federal custody from the time of his tribal arrest. Therefore, federal rights attached and the violation of these rights requires suppression of statements obtained during this period.
The majority errs in the analysis that leads it to a contrary conclusion. First, the majority fails to consider the federal agents’ conduct after Mitchell’s tribal arrest. This court has made clear that the post-arrest activity of federal agents is directly relevant to a finding of collusion or lack thereof. See Michaud, 268 F.3d at 734-35. Critical to the Michaud court’s conclusion that the defendant offered “no evidence of actual collusion ... to deny her federal right to appear before a magistrate judge,” Id. at 735, was the timeliness with which federal charges were brought against her. Although Michaud had first been arrested by state authorities for kidnap and sexual assault, “[a]s soon as the federal agents gathered sufficient evidence against Michaud from the search of her van, they obtained an arrest warrant and took the steps necessary to prosecute her in federal court.” Id. at 734-35. An initial search of Michaud’s hotel room on December 2 led to her arrest by state officials, but a search of her van later that day revealed additional evidence that allowed the federal agents to obtain a federal warrant on December 5. Four days after that, on December 9, the federal agents executed the warrant.
Here, the federal agents made no similar effort to obtain a federal search warrant as “soon as [they] gathered sufficient evidence against [Mitchell].” To the contrary, the record suggests that the agents deliberately delayed bringing Mitchell under federal jurisdiction. ■ On November 4, the FBI and tribal officials executed the tribal warrant, and Mitchell was taken into custody. After agreeing to take a polygraph test and failing, Mitchell made a statement to a federal agent inculpating himself in the robbery. Although this confession should have given the FBI agents sufficient evidence to establish probable cause to obtain a federal warrant to arrest Mitchell, two hours later FBI agents successfully sought to have Mitchell record that confession. In this recorded statement, Mitchell admitted for the first time that he was present when “things happened” to the victims.' Then, the next day, Mitchell led tribal officers to the crime scene, where he confessed to participating in both murders and provided graphic details about the murders and post-mortem dismemberment. Even after this last statement, the FBI agents did not seek a federal warrant, but instead waited sixteen days to file an indictment and eighteen days to have a warrant issued. They then interrogated Mitchell one last time while he was in federal custody and shortly before they presented him to a federal magistrate — a full twenty-four days after he had taken federal agents to the crime scene and confessed to committing the crime of which he was convicted.7
*1002The actions of the federal agents in taking one final inculpatory statement from Mitchell before his arraignment provides especially strong evidence of deliberate intent to deprive him of his rights. On November 29, FBI agents arrested Mitchell on a federal warrant and drove him to the courthouse in Flagstaff. At the courthouse, they took him to a conference room and obtained another statement from him before bringing him before a magistrate judge. In this final and most incriminating statement, Mitchell gave a detailed confession describing the robbery scheme, carjacking, and murders. See ante, at 945. One of the agents admitted at trial that they talked to Mitchell this one last time because they knew that a lawyer would be appointed at his upcoming appearance and that they would not have another opportunity to interrogate him without counsel. This is a direct admission of deliberate intent to deprive Mitchell of his federal rights.
To conclude that these facts do not satisfy the “deliberate intent” requirement would turn an already difficult evidentiary burden into a free pass for FBI agents who know full well that the rights under federal and tribal law differ substantially, and who take purposeful actions to deny defendants their federal rights. Under the majority’s approach, federal agents could meet with tribal agents, as they did here, and encourage a tribal arrest with the express purpose of interrogating defendants and with the knowledge that these interrogations will occur without counsel and before arraignment. Yet so long as the agents do not explicitly state that they sought to avoid affording the defendants their federal protections and as long as they also testify that a purpose of the arrest was to protect the public, they would be free to interrogate repeatedly a defendant held indefinitely without counsel or any arraignment. I do not believe that this court intended such a result when it held that the defendant must show “deliberate intent” to deprive a defendant of procedural rights.
Because Mitchell has shown collusion, his right to federal presentment attached at the time of his tribal arrest, and statements made during the period of unnecessary delay before arraignment should be suppressed. Mitchell’s first statement may have fallen. within the six hour safe harbor established by 18 U.S.C. § 3501(c), but his subsequent confessions did not.8 Although this court does not require exclusion of all non-safe harbor statements, the delay in Mitchell’s arraignment is not justified by either of the reasons we recognize for admitting such statements. See United States v. Mendoza, 157 F.3d 730, 731 (9th Cir.1998) (“We will admit a statement made outside of the safe harbor if the delay was reasonable or if public policy concerns weigh in favor of admission.”) (citing United States v. Van Poyck, 77 F.3d 285, 289 (9th Cir.1996)). We have held that an overnight or weekend delay due to the unavailability of a magistrate is reasonable. Van Poyck, 77 F.3d at 289. However, Mitchell’s arraignment was de*1003layed twenty-five days, not one or two. Furthermore, as there was no intent to take Mitchell before a magistrate when he was arrested, the delay in his arraignment does not fall within that exception. Additionally, given the deliberate delay in arraignment, public policy concerns weigh in favor of suppression, not admission.9 Cf. Van Poyck, 77 F.3d at 290 (finding public policy did not require suppression and explaining “[t]his is not a case where the officers intentionally postponed arraignment so they could interrogate the defendant”).
In sum, the district court erred in not suppressing Mitchell’s inculpatory statements. The confessions' were highly incriminating and highly prejudicial. Thus, this error warrants reversal of Mitchell’s conviction.
B. Jury Selection
Mitchell also contends that the trial court erred in allowing the prosecution to strike Juror # 30, the only African American juror on the venire at the time he was struck. As the Supreme Court recognized in Batson v. Kentucky, 476 U.S. 79, 106 S.Ct. 1712, 90 L.Ed.2d 69 (1986), and sub, sequent cases, the constitutional guarantee of equal protection forbids a prosecutor from excluding jurors because of their race.10 To determine whether a' prosecutor violated this guarantee by using peremptory strikes in a racially discriminatory manner, courts apply the three prong analysis announced in Batson. First, a defendant must establish a prima facie case by showing that relevant circumstances support an inference that the prosecutor excluded venirepersons because of their race. Id. at 96, 106 S.Ct. 1712. At the second step, the burden shifts to the prosecution to justify the strikes by articulating a “neutral explanation related to the particular case to be tried.” Id. at 98, 106 S.Ct. 1712. Finally, the trial court must determine whether the defendant established purposeful discrimination. Id. At this third stage, “[t]he trial court rtiust not simply accept the proffered reasons at face value; it has a duty to evaluate ‘meaningfully the persuasiveness of the prosecutor’s race-neutral explanation to discern whether it is a mere pretext for discrimination.” Williams v. Rhoades, 354 F.3d 1101, 1108 (9th Cir.2004) (internal punctuation omitted) (quoting United States v. Alanis, 335 F.3d 965, 969 (9th Cir.2003)).
In this case, the prosecution struck the only Native American and African-American jurors surviving to the peremptory stage of jury selection. Mitchell raised Batson challenges to both strikes, and the court disallowed the strike of the Native American juror (# 29) but overruled the objection to the strike of the African-American (# 30). In permitting the prosecution to strike juror # 30, the court found that Mitchell did not make out a prima facie case of discrimination and that the government presented a persuasive nondiscriminatory reason for the strike that was not pretextual. Mitchell contends that these findings are clearly erroneous *1004and that they warrant reversal of his conviction. He is correct.
In assessing the prima facie case of discrimination with respect to juror # 30, the trial court gave no weight to its previous finding that the prosecution violated Bat-son by striking juror # 29, the only Native American remaining on the panel.11 However, as this Court recognized in Fernandez v. Roe, 286 F.3d 1073, 1079 (9th Cir.2002), the prosecution’s strikes of jurors of one race are relevant in assessing strikes of jurors of another race. This is particularly true when the absolute number of jurors of a particular racial group is small, and the use of challenges against that group may be insufficient to support an inference of discrimination. Id at 1078. In Fernandez, we held that peremptory strikes against Hispanic jurors helped establish a prima facie case of discrimination against African American jurors. Similarly, in this case, the prosecution’s unlawful strike of the only Native American juror is a relevant circumstance giving rise to an inference of discrimination in the strike of the only African-American juror. The trial court erred in failing to consider this evidence and in concluding that Mitchell did not establish a prima facie case of discrimination.
Turning to the second step of the Bat-son analysis, the government may have offered a facially “legitimate” reason for its decision to dismiss juror # 30. See Miller-El v. Dretke, 545 U.S. 231, 239, 125 S.Ct. 2317, 162 L.Ed.2d 196 (2005) (quoting Batson, 476 U.S. at 98, 106 S.Ct. 1712) (“[T]he prosecutor must give a clear and reasonably specific explanation of his legitimate reasons for exercising the challenge.”) (emphasis added). The government stated that it struck juror # 30 because he had served on a past jury that acquitted a criminal defendant. The record indicates that on the jury questionnaire, juror #30 stated that over twenty years earlier he had participated in ' a state criminal trial in which the verdict was not guilty. During voir dire, the prosecution asked him no questions about this prior jury service. In contrast, defense counsel asked juror #30 whether he remembered the nature of the offense or anything about the prior case. Juror # 30 responded: “I think it occurred like two or three years earlier [prior to the trial]. And we just had a little information. I guess I really don’t.” However, despite the prosecution’s apparent lack of interest in the issue during voir dire, and despite juror # 30’s statement that he did not recall anything of significance about the case, the prosecution declared that it struck the one African-American juror because of his prior jury service. The district court held that this was “clearly allowed.”
*1005The district court and the majority accept without question the proposition that the government can permanently prevent someone from serving on all juries simply because, several decades previously, that individual once voted to acquit a person accused of a crime, possibly in a ease involving a minor non-violent offense, even a misdemeanor, in which the defendant established his innocence beyond question. Such a proposition, if accepted generally, would place a substantial burden on.the exercise by any citizen of the right to serve on a jury. As the Supreme Court recognized in Powers v. Ohio, 499 U.S. 400, 111 S.Ct. 1364, 113 L.Ed.2d 411 (1991), “[j]ury service is an exercise of responsible citizenship by all members of the community, including those who otherwise might not have the opportunity to contribute to our civic life.” Id. at 402, 111 S.Ct. 1364. Like voting, jury service is both an opportunity to participate in the affairs of government and a means by which our government maintains its legitimacy:
Jury service preserves the democratic element of the law, as it guards the rights of the parties and ensures continued acceptance of the laws by all of the people. See Green v. United States, 356 U.S. 165, 215, 78 S.Ct. 632, 2 L.Ed.2d 672[] (1958) (Black, J., dissenting). It “affords ordinary citizens a valuable opportunity to participate in a process of government, an experience fostering, one hopes, a respect for law.” Duncan, supra, 391 U.S. at 187[], 88 S.Ct. 1444 (Harlan, J., dissenting). Indeed, with the exception of voting, for most citizens the honor and privilege of jury duty is their most significant opportunity to participate in the democratic process.
Id. at 406-07, 111 S.Ct. 1364.
A strike based solely on a juror’s participation along with all of his fellow jurors in a prior acquittal in a trial of an undetermined nature, held many years earlier, threatens the institution of the jury by penalizing a juror for his past lawful and proper “exercise of responsible citizenship.” In addition to infringing upon the juror’s right to serve, such strikes create an incentive for jurors to vote to convict in order to retain the possibility of serving on other juries in the future. This undermines the jury’s role in “guarding] the rights of the parties” and encourages jurors to violate their constitutional duty to acquit when required by the law and the evidence. As a society, we benefit from having jurors who can exercise their responsibilities fully and fairly without the government seeking improperly to influence their decisions. Permitting prosecutors to strike jurors who have once voted to acquit an unknown defendant on an unknown charge authorizes them to prevent those who have previously stood between the state and the individual from ever again “being part of the judicial system of the country [and] preventing] its arbitrary use or abuse.” Id. at 406, 111 S.Ct. 1364 (quoting Balzac v. Porto Rico, 258 U.S. 298, 42 S.Ct. 343, 66 L.Ed. 627 (1922)). It is enough to disqualify from serving on death penalty juries those citizens who have reservations about capital punishment or who would reserve that penalty for the most extreme offenses. See Uttecht v. Brown, — U.S. -, 127 S.Ct. 2218, 2224-29, 167 L.Ed.2d 1014 (2007). To also disqualify all those who might have ever believed that a defendant could be innocent would seem to cross the line between providing a fair jury — a jury of one’s peers — and adopting a system that is inherently proprosecution and weighted overwhelmingly in favor of conviction.
Even though a strike for the reasons suggested by the prosecution is not truly legitimate in the ordinary sense of the word, it nonetheless survives the de min-imis burden placed on the prosecution at the second of the Batson analysis. Although Batson spoke of “legitimate” rea*1006sons, 476 U.S. at 98 n. 20, 106 S.Ct. 1712, and recent Supreme Court cases continue to use this language, see, e.g., Miller-El v. Dretke, 545 U.S. 231, 239, 125 S.Ct. 2317, 162 L.Ed.2d 196 (2005), courts applying this step have equated “legitimate” with “race-neutral.” See, e.g., Purkett v. Elem, 514, U.S. 765, 768, 115 S.Ct. 1769, 131 L.Ed.2d 834 (1995) (quoting Hernandez, 500 U.S. at 360, 111 S.Ct. 1859) (“At this [second] step of the inquiry, the issue is the facial validity of the prosecutor’s explanation. Unless a discriminatory intent is inherent in the prosecutor’s explanation, the reason offered will be deemed race neutral.”); Tolbert v. Page, 182 F.3d 677, 680 (9th Cir.1999) (en banc) (“Whether the justification offered by a prosecutor is an adequate race-neutral explanation is a question of law reviewed de novo.”) (quoting United States v. Bishop, 959 F.2d 820, 821 n. 1 (9th Cir.1992)) (internal quotation marks omitted). On its face, the explanation that a juror was removed because he once voted to acquit a defendant is race-neutral. Under that standard, the reason advanced here by the prosecution is “legitimate” for purposes of step two of Batson.
An absolute bar on jury service for a citizen who once voted; long ago, to acquit an unknown defendant on unknown charges is, however, so divorced from the individual consideration of jurors required by Batson and so damaging to the fairness and independence of the jury system that, when exercised to remove the sole African-American on the panel, it must, under the third step, be viewed as a pretext. So improper and arbitrary a justification cannot be presumed to be the prosecution’s true motive.
“Competence to serve as a juror ultimately depends on an assessment of individual qualifications and ability impartially to consider evidence presented at a trial.” Batson, 476 U.S. at 87, 106 S.Ct. 1712. When a juror has recently served on a jury that voted to acquit, individualized assessment of that juror may well lead the prosecution to conclude that he might be inclined to favor the defense. See, e.g., United States v. Power, 881 F.2d 733, 740 (9th Cir.1989) (noting the challenged juror “had recently completed service on another jury” and the prosecutor “feared the juror might be hostile to the government for calling him to serve again so soon after his service as a juror in a long prior trial”); United States v. Thompson, 827 F.2d 1254, 1256 n. 1 (9th Cir.1987) (quoting the prosecution’s explanation that it struck a juror “because he acquitted in a case just a couple of weeks ago”). However, when the past jury service was over two decades ago, and of such minimal consequence that the juror cannot recall the substance of the case, there is no basis for inferring that his prior vote to acquit would influence his vote in the present case. The prosecution’s questions during voir dire did not address the past acquittal at all, much less reveal information supporting a genuine concern that the experience would affect juror # 30’s present deliberations. Under these circumstances, the proffered explanation for striking juror # 30 must be deemed a pretext for what was, in fact, a race-based .strike. The district court’s decision to the contrary was clearly erroneous,' and this error as well requires reversal of Mitchell’s conviction. See Powers v. Ohio, 499 U.S. 400, 412, 111 S.Ct. 1364, 113 L.Ed.2d 411 (1991); Batson, 476 U.S. at 100, 106 S.Ct. 1712.
II. Sentencing
On appeal, Mitchell raises numerous errors to which he failed to object during trial. I now consider the effect of three of the most significant of these errors that occurred during the sentencing phase of this case. The three errors are as follows: First, the district court erred in permitting' Mitchell to be absent from the sentencing *1007phase, in direct violation of the Federal Rules. This is a structural error requiring automatic reversal of Mitchell’s sentence. However, were the error deemed not to be structural, it would then constitute a part of the cumulative error analysis. See Section D, infra at 1013-14. While individual errors “may not rise to the level of reversible error, their cumulative effect may nevertheless be so prejudicial to the appellants that reversal is warranted.” United States v. Wallace, 848 F.2d 1464, 1475 (9th Cir.1988) (citing United States v. Berry, 627 F.2d 193, 200-01 (9th Cir.1980)). In addition to Mitchell’s absence from the penalty phase, the two other errors on which I rely are (a) the prosecution’s numerous improper statements during closing arguments and (b) the court’s failure to instruct the jury on the correct standard of proof for finding that aggravating factors outweigh mitigating factors. I conclude that the cumulative effect of these three errors is prejudicial, and I would, accordingly, vacate Mitchell’s death sentence.
A. Defendant’s Absence From the Sentencing Phase
As the majority opinion describes, ante, at 985-86, Mitchell unequivocally stated that he did not wish to attend the sentencing phase of the proceedings. The district judge strongly urged him to exercise his right to be present on several occasions, and carefully explained the reasons why he should do so. Ultimately, with considerable misgivings, she reluctantly allowed him to be absent. Mitchell now argues that the Federal Rules of Criminal Procedure require a capital defendant’s presence during sentencing proceedings and that the district court erred in permitting him to waive his presence. He is correct, and notwithstanding the majority’s difficulty with the issue, the error is plain.
Rule 43(a)(3) provides that a defendant must be present at sentencing. Although a non-capital defendant may waive this right, a capital defendant may not.12 The majority makes the perplexing assertion that Rule 43(a)(3) does not apply because the sentencing phase of a capital trial is not part of “sentencing.” See ante, at 988. This contradicts both the common understanding of the nature of the sentencing phase and the rule’s purposes in requiring a capital defendant’s presence at sentencing, as well as the plain meaning of the rule itself.
In a capital trial, the term sentencing phase is synonymous with the term penalty phase. The two terms are used interchangeably, along with the statutory term sentencing hearing. See, e.g., Rice v. Wood, 77 F.3d 1138 (9th Cir.1996) (en banc) (observing that a capital defendant’s trial “was bifurcated into separate guilt and sentencing phases, with the same jury sitting in both”). Cf. Black’s Law Dictionary (8th ed.2004) (noting that “penalty phase” is “[ajlso termed sentencing phase”). Indeed, the Federal Death Penalty Act specifically provides that if a defendant is found guilty of a death eligible offense, the trial judge “shall conduct a separate sentencing hearing to determine *1008the punishment to be imposed.” 18 U.S.C. § 3593(b) (emphasis added). It is at the sentencing hearing (a.k.a. the sentencing phase or the penalty phase) that the evidence is adduced, the arguments are made by counsel, the defendant has the opportunity to address the jury — the body that determines what his punishment — and the ultimate decision is made as to the defendant’s sentence.
The conclusion that “sentencing” includes the sentencing hearing is also required by an examination of the reasons for requiring a defendant’s presence at sentencing. In non-capital cases, the requirement that a defendant be present at sentencing, inter alia, “serves the defendant’s interest by facilitating allocution .... ” United States v. Curtis, 523 F.2d 1134, 1135 (D.C.Cir.1975). It is at sentencing that the defendant has “the opportunity to address any issues relevant to his sentence.” United States v. Robinson, 390 F.3d 853, 887 (6th Cir.2004). Additionally, the mere presence of the defendant in the courtroom during sentencing proceedings exerts a psychological influence on the jury. See, e.g., United States v. Canady, 126 F.3d 352, 362 (2d Cir.1997) (“In the jury context, several courts, in rejecting the argument that the defendant’s presence is useless, have pointed to the fact that the defendant’s mere presence exerts a ‘psychological influence upon the jury.’ ”) (quoting United States v. Santiago, 977 F.2d 517, 523 n. 6 (10th Cir.1992)).
In a federal capital case, the reasons for requiring a defendant’s presence at sentencing are applicable primarily to the sentencing hearing rather than to the formal pronouncement of sentence by the judge, although the defendant’s presence is required at both.13 As noted, it is during the sentencing phase of the proceedings — the sentencing hearing — that a capital defendant has the opportunity to present any mitigating evidence, and to ask the jury— the body deciding on his sentence — for mercy. Additionally, the psychological influence of the defendant’s presence on the jury is most significant during the time when it is hearing the argument as to whether he should live or die and when it is receiving the information that serves as the basis for its decision. Finally, to the extent that a defendant’s presence legitimizes a sentence, this legitimizing function is most critical in the sentencing phase, when the State argues that there is cause to exercise the full measure of its power by taking the life of a defendant and the defendant seeks the alternative of life imprisonment. In short, the process of determining a capital defendant’s sentence occurs during the sentencing or penalty phase. The Federal Death Penalty Act provides that at the conclusion of this phase, the jury “shall recommend whether the defendant should be sentenced to death, to life imprisonment without possibility of release or some other lesser sentence.” 18 U.S.C. § 3593(e). If the jury recommends death and, the judge is obligated to impose that sentence and has no discretion to impose a lesser penalty. See 18 U.S.C. § 3594 (“Upon a recommendation under section 3593(e) that the defendant should be sentenced to death or life imprisonment without possibility of release, the court shall sentence the defendant accordingly.”). Thus, although Mitchell was present when the jury returned its verdict and, months later, when the judge entered the death sentence, he was no more than a passive audience on these occasions. The decision-making process was over. The time when Mitchell’s presence was important, when he could have been an active participant in the sen-*1009fencing proceedings and influenced the determination as to his sentence, was during the sentencing or penalty phase. Any remarks he may have been permitted to make to the judge subsequent to the jury’s decision were, of no legal or practical consequence. His absence during the sentencing hearing violated the plain language as well as the purposes of Rule 43(a).14
There is little room for doubt that Rule 43(a)(3)’s requirement that a defendant be present at sentencing applies to the sentencing or penalty phase of a capital trial: to the sentencing hearing. Because Rule 43(a) states that only a noncapital defendant may waive his presence during sentencing, the district court erred in allowing Mitchell to be voluntarily absent during the penalty phase. Furthermore, this error is a structural error that requires reversal even without a specific showing of prejudice. Although we generally consider structural error within the context of constitutional errors, “numerous errors are subject to automatic reversal even though they do not violate constitutional rights.” United States v. Annigoni, 96 F.3d 1132, 1144 (9th Cir.1996) (en banc). See, e.g., Gomez v. United States, 490 U.S. 858, 874-76, 109 S.Ct. 2237, 104 L.Ed.2d 923 (1989) (holding defendant did not need to show prejudice resulting from a magistrate supervising voir dire in violation of the Federal Magistrate’s Act because the defendant’s right “to have all critical stages of a criminal trial conducted by a person with jurisdiction to preside” is “among the basic fair trial rights that can never be treated as harmless”). In determining whether an error requires automatic reversal, we look at whether the error is a “structural defect affecting the framework within which the trial proceeds, rather than simply an error in the trial process itself.” Arizona v. Fulminante, 499 U.S. 279, 310, 111 S.Ct. 1246, 113 L.Ed.2d 302 (1991). Here, Mitchell’s absence throughout the penalty phase affected the very nature of the proceedings. The absence of a defendant when his presence is required is as fundamental as the absence of his lawyer, see Gideon v. Wainwright, 372 U.S. 335, 344, 83 S.Ct. 792, 9 L.Ed.2d 799 (1963), or even of the judge. See Tumey v. Ohio, 273 U.S. 510, 535, 47 S.Ct. 437, 71 L.Ed. 749 (1927). The defendant’s presence is required so *1010that he may assist his counsel and plead for mercy from the jury. Additionally, his presence could well have significantly affected the course of the proceedings, the presentation of the mitigating and aggravating evidence. In contrast to cases in which the defendant was absent' only during the pronouncement of the verdict, see, e.g., Rice v. Wood, 77 F.3d 1138 (9th Cir.1996), a conversation between a judge and a juror, see, e.g., Rushen v. Spain, 464 U.S. 114, 119, 104 S.Ct. 453, 78 L.Ed.2d 267 (1983), or the readback of testimony, see, e.g., Hegler v. Borg, 50 F.3d 1472, 1474-75 (9th Cir.1995), a capital defendant has an “active role to play” during the pénalty phase of his trial. Cf. Rice, 77 F.3d at 1142. Because Mitchell’s absence was an error that permeated the entire penalty phase, it may not be “quantitatively assessed in the context of other evidence presented.... ” Fulminante, 499 U.S. at 308, 111 S.Ct. 1246. It is impossible to determine from a review of a record whether the penalty imposed would have been different if an absent defendant had been present during the sentencing phrase.
I would reverse the death sentence on the basis of the court’s fundamental error in allowing Mitchell to be absent from the sentencing hearing. See Annigoni, 96 F.3d at 1147 (holding that a violation of Rule 24(b) of the Federal Rules of Criminal Procedure, regarding peremptory challenges, is an error requiring automatic reversal). In the alternative, I will consider the Rule 43 error below as if it were mere trial error, in which case it would constitute part of the cumulative error analysis. See section D at pp. 1013-14.
B. Closing Arguments
As the majority opinion acknowledges, ante, at 995, the government made many improper statements in its closing arguments. In determining whether these improper statements alone warrant reversal, “[t]he critical inquiry is whether, in the circumstances of the trial as a whole, the remarks were so prejudicial that they likely influenced the jury adversely to the defendant and deprived the defendant of a fair trial.” United States v. Patel, 762 F.2d 784, 795 (9th Cir.1985).
Much of the argument consisted of prohibited statements intended to arouse the passion of the jury. See United States v. Leon-Reyes, 177 F.3d 816, 822 (9th Cir.1999) (“Prosecutors may not make comments calculated to arouse the passions or prejudices of the jury.”). For instance, the prosecution commented: “What I like to call mitigating factors are excuses for murder because we have free will.” Regarding defense witness who offered mitigation testimony, the prosecution stated: “By the way, the defendant’s using these people, too. Manipulating them. They are up there begging for his life. Do you think [Jane Doe] begged when she was told to lay down and die?” In arguing a death sentence would be a “message” to Mitchell, the prosecution suggested the jury could consider its own trauma in hearing about the murders: “It is a message all right, it is a message to the defendant, that you have challenged us, you have come before us with things that are going to cause us nightmares as long as we are alive.”
In a similar vein, the prosecution suggested Mitchell had already received mercy by virtue of the trial itself: “Perhaps years ago, Tombstone, he would have been taken out back, strung up. He would have gotten a trial, nothing like this. We have been at this for seven weeks. We’ve gone to great lengths to choose the jury. We have presented a trial. You have made your findings. And yet still he gets to come before you and say, ‘Spare my life.’ ”
Other statements improperly shifted the burden of proof to Mitchell to show that *1011death was not warranted. The prosecutor asked: “What has Lezmond Mitchell done to earn a minimum sentence for the slaughter of two people? Nothing.” Similarly, he remarked: “The defendant has done nothing to earn an opportunity to be a message to someone in prison. He has earned no opportunity to live in a cell whether he likes the cell or not. He has earned no opportunity to get a college degree while in prison or to Internet chat with someone or to work in the shop of prison. He has earned nothing.” In contrast with cases where the prosecution clarified its comments by reminding the jury that the government had the burden of proof, see, e.g., United States v. Cabrera, 201 F.3d 1243, 1249 (9th Cir.2000), the prosecution in this case offered no such reminder.
However, in the context of the entire trial, I cannot conclude that these statements alone warrant reversal. It is not clear that by themselves, they were so prejudicial that they violated due process and “deprived the defendant of a fair trial.” Patel, 762 F.2d at 795. Nonetheless, the prosecutor’s numerous improper statements Were contrary to applicable law and constituted sentencing phase error. The effect of these errors must therefore be considered as a part of our cumulative error analysis.
C. Failure to Include “Beyond a Reasonable Doubt” in Weighing Instruction
Finally, Mitchell argues that the district court erred by failing to instruct the jury that the government bore the burden of proving beyond a reasonable doubt that aggravating factors outweighed mitigating factors. Instead, the instructions stated:
This weighing process asks whether you are unanimously persuaded that the aggravating factors sufficiently outweigh any mitigating factors or, in the absence of any mitigating factors, that the aggravating factors are themselves sufficient to justify a sentence of death.
The “beyond a reasonable doubt” standard applies if the finding that aggravating factors outweigh mitigating factors is a finding of fact that serves to increase the maximum sentence a judge may impose on a defendant. “Other than the fact of a prior conviction, any fact that increases the penalty for a crime beyond the prescribed statutory maximum must be submitted to a jury, and proved beyond a reasonable doubt.” Apprendi v. New Jersey, 530 U.S. 466, 490, 120 S.Ct. 2348, 147 L.Ed.2d 435 (2000). In Ring v. Arizona, the Supreme Court held that Apprendi applies to the sentencing phase of capital trials. Ring v. Arizona, 536 U.S. 584, 588, 122 S.Ct. 2428, 153 L.Ed.2d 556(2002) (“Capital defendants, no less than noncapi-tal defendants, we conclude, are entitled to a jury determination of any fact on which the legislature conditions an increase in their maximum punishment.”). In so holding, it overruled Walton v. Arizona, 497 U.S. 639, 110 S.Ct. 3047, 111 L.Ed.2d 511 (1990), “to the extent that it allows a sentencing judge, sitting without a jury, to find an aggravating circumstance necessary for imposition of the death penalty.” Ring, 536 U.S. at 609, 122 S.Ct. 2428.
There is no doubt that the finding that aggravating factors outweigh mitigating factors increased Mitchell’s maximum punishment. The Federal Death Penalty Act provides that as the final step in determining whether death is justified, the jury must consider:
whether all the aggravating factor or factors found to exist sufficiently outweigh all the mitigating factor or factors found to exist to justify a sentence of death, or, in the absence of a mitigating factor, whether the aggravating factor or factors alone1 are sufficient to justify a sentence of death.
*101218 U.S.C. § 3593(e). Absent this finding, the maximum sentence the court could have imposed would have been life imprisonment without the possibility of release. See 18 U.S.C. § 3594. Thus, the only question is whether this finding is a finding of fact. The majority suggests that it is not, although it concludes that it is unnecessary to resolve the question on plain error review. I conclude that it is necessary to make the determination if only because the failure to instruct the jury on the burden of proof is a serious error that must be considered as part of our cumulative error analysis.
The majority suggests that the weighing process merely channels the jury’s discretion and that it does not result in a finding of essential fact. Ante, at 993. The majority supports this suggestion with two observations. First, it cites dicta from Kansas v. Marsh, — U.S. -, 126 S.Ct. 2516, 2526, 165 L.Ed.2d 429 (2006), to the effect that the weighing step simply channels jury discretion by giving jurors criteria to use in deciding whether or not to recommend a death sentence. While the majority is correct that this step channels juror discretion, the same can be said of every other step in the Federal Death Penalty Act’s sentencing process. Channeling discretion is not incompatible with finding essential facts. Rather, it is precisely by instructing juries which facts are essential that the Act channels juror discretion.
Second, the majority suggests that Mitchell’s argument is inconsistent with the Fifth Amendment requirement that facts increasing the maximum penalty be found by the grand jury and charged in the indictment. See Jones v. United States, 526 U.S. 227, 243 n. 6, 119 S.Ct. 1215, 143 L.Ed.2d 311 (1999). According to the majority, since the grand jury cannot know which mitigators the defendant will urge, it cannot find that aggravators outweigh mitigators. Thus, the majority concludes that Mitchell’s position is flawed because it characterizes as an essential fact something that cannot be found by the grand jury. Ante, at 993. However, the majority creates conflict where none exists. The weighing process mandated by the Act applies regardless of whether jurors have found mitigating factors. As the Act explains, “in the absence of a mitigating factor,[the jury shall consider] whether the aggravating factor or factors alone are sufficient to justify a sentence of death.” 18 U.S.C. § 3593(e). The grand jury, working without knowledge of mitigating facts, is capable of making a similar determination based upon the aggravating factors before it. In fact, grand juries regularly make factual determinations, for the purposes of that body’s limited function, without having heard the defendant’s opposing facts that might have caused it to reach an opposite conclusion. Hearing only one-side of the story and rendering a judgment is a hallmark of the grand jury system.
This court has not yet decided whether the weighing of aggravators and mitigators results in a finding of fact. It is apparent to me that it does. Pre-Apprendi decisions in other circuits rejected the Eighth Amendment claim that the “beyond a reasonable doubt” standard applies to the weighing of aggravators and mitigators, but these decisions did not consider whether the weighing decision was a “fact.” See United States v. Flores, 63 F.3d 1342, 1376 (5th Cir.1995); United States v. Chandler, 996 F.2d 1073, 1091-93 (11th Cir.1993). The one circuit court that has considered this question since Apprendi held, in the context, of a Fifth Amendment claim, that the Federal Death Penalty Act’s weighing process did not result in a finding of fact that must be charged in . an indictment. See United States v. Purkey, 428 F.3d 738, 750 (8th Cir.2005) (noting the weighing process is not an elemental fact but in*1013stead merely “the lens through which the jury must focus the facts that it has found to produce an individualized determination”). However, this reasoning is faulty. As noted, the weighing process does have a discretionary aspect, but it also requires the jury to make an ultimate factual determination about whether aggravating factors sufficiently outweigh mitigating factors so as to justify a sentence of death. Unless it so determines, the death penalty cannot be imposed.
Notwithstanding the discretionary elements of the weighing step, three state supreme courts have held that Apprendi applies to similar weighing provisions in state death penalty statutes. See Missouri v. Whitfield, 107 S.W.3d 253 (Mo.2003) (en banc) (noting that weighing “require[s] factual findings that are prerequisites to the trier of fact’s determination that a defendant is death eligible”); Woldt v. People, 64 P.3d 256, 265 (Colo.2003) (en banc) (invalidating a superseded Colorado death penalty statute because three steps of the sentencing process, including weighing, “required judges to make findings of fact that render a defendant eligible for death”); Johnson v. State, 118 Nev. 787, 802-03, 59 P.3d 450 (Nev.2002) (holding Apprendi and Ring apply to the finding that mitigation does not outweigh aggravation). As stated succinctly by the Nevada Supreme Court, the jury’s finding on this issue “is in part a factual determination, not merely discretionary weighing.” Johnson, 118 Nev. at 802-03, 59 P.3d 450.
Ultimately, the majority’s attempt to characterize the weighing step as resulting in something other than an Apprendi finding of fact reflects a level of formalism rejected by both Apprendi and Ring. As Apprendi emphasized, “the relevant inquiry is one not of form, but of effect-does the required finding expose the defendant to a greater punishment than that authorized by the jury’s guilty verdict?” 530 U.S. at 494, 120 S.Ct. 2348. In Ring v. Arizona, the Court relied on this approach in holding that aggravating factors must be proven to a jury beyond a reasonable doubt. 536 U.S. at 609, 122 S.Ct. 2428 (“Because Arizona’s enumerated aggravating factors operate as -the functional equivalent of an element of a greater offense,’ -the Sixth Amendment requires that they be found by a jury.”) (internal citation omitted). From this functional perspective, there is no practical difference between' the increase in punishment due to the finding of an aggravating factor and the increase due to the finding that aggravators outweigh mitigators. Because the Federal Death Penalty Act requires both findings in order for a judge to sentence a defendant to, death, the Sixth Amendment requires a jury to make these findings beyond a reasonable doubt. The district court erred in concluding otherwise.
D. Cumulative Error
Considered cumulatively, each of the errors discussed above combined to preju: dice Mitchell with respect to the jury’s final determination to impose the death penalty. Mitchell’s absence from penalty phase proceedings meant that he was not present to participate in the sentencing hearing, a critical point in the determination of his fate. His absence also meant that the jury was not required to face him in the immediate period before it decided that he should die. The government’s improper closing arguments appealed to the jury’s emotions and undoubtedly exacerbated their revulsion at the offenses. Finally, the jurors were not required to decide beyond a reasonable doubt whether aggravating factors outweighed mitigating factors, and thus were allowed to opt for death without having to make the more difficult and stringent decision required by law.
*1014Given the nature of the jury’s function in determining whether to impose capital punishment, it seems apparent that the cumulative effect of these errors was not harmless. Each affected in an important way a significant aspect of the jury’s decision-making process. Cumulative error may include violations that fail the plain error test, but are nevertheless errors. See United States v. Fernandez, 388 F.3d 1199, 1256-57 (9th Cir.2004) (considering errors that did not rise to the level of plain error in a cumulative error analysis); United States v. Wallace, 848 F.2d 1464, 1476 n. 21 (9th Cir.1988) (holding that an error that was not objected to at trial, and may - not have amounted to plain error, .should be considered in the cumulative error analysis). Multiple errors, even if harmless when considered individually, may have a cumulative prejudicial effect that deprives the defendant of the due process right to a fair trial. See Karis v. Calderon, 283 F.3d 1117, 1132 (9th Cir.2002); Ceja v. Stewart, 97 F.3d 1246, 1254 (9th Cir.1996). Here, the combined effect of multiple sentencing phase errors may well have caused at least one juror to vote for death when he or she would otherwise have decided not to do so. In light of this probability, I would vacate Mitchell’s death sentence.
III. Conclusion
I respectfully dissent. For the reasons set forth above, I would reverse Mitchell’s conviction and remand for a new trial. Should his conviction stand, however, I would, for the additional reasons explained in this opinion, vacate Mitchell’s sentence and remand for a new sentencing hearing.
. Mitchell was subject to the death penalty under the Federal Death Penalty Act, because file murder resulted from an armed carjacking. See 18 U.S.C. §§ 3591-98.
. As the majority states, ante, at 960-61 n. 4, the Sixth Amendment right to counsel does not attach to defendants in tribal custody. See United States v. Percy, 250 F.3d 720, 725 (9th Cir.2001).
.The agents also testified that they wanted to get defendant into custody quickly to protect the public safety. Public safety is a concern whenever violent crimes are involved. The pertinent question is not whether there was a legitimate reason to detain the defendant, but rather whether the manner in which the authorities detained him — having him arrested by tribal rather than federal authorities — reflected a deliberate effort to avoid affording him his federal rights. See Alvarez-Sanchez, 511 U.S. at 359, 114 S.Ct. 1599.
. Agent Purscell, one of the FBI agents involved in Mitchell’s arrest and interrogation, stated that he believed the tribal arrest warrant was for robbery or assault. According to Orsinger, Mitchell’s co-defendant, after the tribal arrest federal agents told him that he was arrested on an armed robbery charge. Orsinger also testified that three FBI agents, and no tribal agents, attended his first post-arrest interrogation.
. The offense of robbery is similarly defined under the Navajo Nation Code and federal law. The Navajo Nation Code describes the *1000elements of robbery and armed robbery as follows:
17 N.N.C. § 491(A) A person commits robbery if in the course of committing theft, he or she threatens or uses immediate force against any person with intent either to coerce surrender of property or to forestall resistance to his or her taking or retaining of property.
17 N.N.C. § 492(A) A person commits armed robbery if in the course of committing robbery as defined in 17 N.N.C. § 491, he or she or an accomplice:
(1) Is armed with a deadly weapon; or
(2) Uses or threatens to use a deadly weapon or dangerous instrument.
The Federal Criminal Code describes the elements of robbery:
18 U.S.C.A. § 2111 Whoever, within the special maritime and territorial jurisdiction of the United States, by force and violence, or by intimidation, takes or attempts to take from the person or presence of another anything of value, shall be imprisoned not more than fifteen years.
Armed robbery is encompassed by § 2111, see United States v. Burns, 701 F.2d 840, 841, 842 (9th Cir.1983).
. As in the case of robbery, if the arrest were for assault the type of assault involved here would constitute a federal crime as well as a tribal crime, compare 18 U.S.C. §§ 113, 1153 with 17 N.N.C. § 314-15, and the probable cause requirement would be the same.
. The majority’s reliance on U.S. v. Percy, 250 F.3d 720 (9th Cir.2001), for the proposition that this delay does not compel a finding of collusion is unavailing. First, in Percy, the delay occurred between the time defendant was detained and the time federal agents questioned him. Id. at 724. The defendant was transferred into federal custody the day after the interview. Id. Here, Mitchell re*1002mained in tribal custody long after the first interrogation by federal officials, and through successive rounds of additional questioning. Clearly, there is a difference between delaying the initial interrogation and delaying the transfer into custody once the interrogation has taken place. Second, the defendant in Percy did not challenge the district court’s finding that there was no collusion, so the court was bound by that finding on appeal and did not conduct its own analysis of the question. Id. at 727. Thus, Percy is of no assistance to the majority here.
. During his first round of questioning, Mitchell denied involvement in the disappearances and the robbery, and an inability to suppress this statement would therefore be of little consequence.
. With respect to public policy concerns, some of the federal agents' statements emphasized the importance of arresting Mitchell and his coparticipants in order to remove them from the streets for public safety reasons. However, this goes to the reason for the arrest, not the reason for the subsequent retention of Mitchell in tribal custody or the inordinate delay in his arraignment after probable cause unquestionably existed.
. Although the prima facie case described in Batson required a defendant to show that the prosecutor used peremptories to remove, veni-repersons of the same race as the defendant, 476 U.S. at 96, 106 S.Ct. 1712, in Powers v. Ohio, 499 U.S. 400, 402, 111 S.Ct. 1364, 113 L.Ed.2d 411 (1991), the Supreme Court removed the requirement that the defendant and the stricken jurors be of the same race.
. The majority suggests it cannot determine whether in ruling on juror #30 the district court factored in its ruling on juror # 29. However, the record makes quite clear that the trial court found no relevant circumstances other than juror # 30’s being African-American. The Court stated:
I do not find that a prima facie case has been made strictly because # 30 is an African-American male. I see nothing else in the facts and circumstances that have been presented or that the Court is aware of surrounding the exercise of the peremptory challenge that would raise an inference of discrimination.
Although the Court did discuss its previous ruling on juror # 29, it did so only to contrast the two strikes, and not as a circumstance relevant to the strike of juror #30:
# 30 is different from # 29, although both are members of a recognizably protected group. For the reasons I stated previously with # 29 because of the totality of the circumstances of this particular case involving Native Americans and # 29 being the only Native American venireperson, and the other reasons stated I think distinguish # 29 from # 30 and my consideration that a prima facie case has been made.
. Rule 43(a)(3) states “Unless this rule ... provides otherwise, the defendant must be present at ... sentencing.” The rule is limited by 43(c)(1)(B), which permits waiver of that provision "in a noncapital case, when the defendant is voluntarily absent during sentencing.” As the majority concedes, Rule 43(c)(1)(B) by negative implication does not permit a capital defendant to waive his presence during sentencing by being voluntarily absent. Because Rule 43(c)(1)(B) does not apply to capital defendants, and the facts do not support any of the other reasons for waiver, See Fed.R.Crim.P. 43(c)(1)(A) and (B), we are presented with a straightforward question about the meaning of “sentencing” as the term is used in Rule 43(a)(3).
. In contrast, in a non-capital case the hearing regarding the sentence and the pronouncement of the sentence occur at a single proceeding.
. The majority argues that Fed R.Crim. P. 35(c), which defines the word "sentencing” as "the oral announcement of the sentence,” should govern the interpretation of Rule 43. Rule 35(c) specifically states that the definition applies to the term sentencing "[a]s used in this rule.” Id. (emphasis added). Even aside from that express limitation, there is no reason to believe that the definition should carry over to other rules. Indeed, there is every reason to believe that it should be limited to Rule 35, because that rule contains several provisions that set deadlines based on the date of sentencing. Rule 35(c) simply clarifies how those deadlines should be calculated. Rule 43 serves a completely different purpose. The fact that it explains in one section that "sentence correction” is a proceeding under Rule 35 is of no relevance whatsoever to the question before us. The majority also relies on the legislative history of the amendments to Rule 43. I find this history inconclusive. Although the earlier version allowed a defendant in a noncapital case to be absent from the "imposition of sentence” while the current version allows a noncapital defendant to be absent from "sentencing,” there is no practical difference between the terms as applied to a noncapital defendant. There is one sentencing proceeding in noncapital cases and it is at that proceeding that impo.sition of the sentence occurs. The word "sentencing” and the term "sentencing proceeding” have far broader meanings with respect to a capital defendant. A negative implication under such circumstances is also broader. Thus, general committee notes, which are ambiguous as applied, are hardly conclusive of critical questions that otherwise would be resolved by looking to the plain meaning or ordinaiy construction of the terms in question.