We consider here the propriety of the district court’s dismissal of a juror allegedly engaged in “nullification” — the intentional disregard of the law as stated by the presiding judge — during the course of deliberations. We address, in turn, (1) whether such alleged misconduct constitutes “just cause” for dismissal of a deliberating juror under Rule 23(b) of the Federal Rules of Criminal Procedure (“Rule 23(b)”),1 so that a jury of only eleven persons may continue to deliberate and return a verdict, and (2) what eviden-tiary standard must be met to support a dismissal on this ground.
The appellants are two sets of defendants convicted of violations of federal narcotics laws after two separate trials in the United States District Court for the Northern District of New York (Thomas J. McAvoy, Chief Judge). We have decided the appeals of defendants convicted at the first of these trials in a summary order of this date. We write here to consider only the appeals of Grady Thomas, Ramse Thomas, Jason Thomas, Tracey Thomas, and Loray Thomas, all of whom were convicted in the second trial. On appeal, they argue, chiefly, that the district court abused its discretion when it ordered the dismissal of one of the jurors pursuant to Rule 23(b) during the course of jury deliberations. The court based its decision to remove the juror, in large part, on a finding that the juror was purposefully disregarding the court’s instructions on the law — in effect, that the juror intended to acquit the defendants regardless of the evidence of their guilt.
We consider below whether a juror’s intent to convict or acquit regardless of the evidence constitutes a basis for the juror’s removal during the course of deliberations under Rule 23(b). We also consider what constitutes sufficient evidence of that intent in light of the limitations on a presiding judge’s authority to investigate allegations of nullification required by the need to safeguard the secrecy of jury deliberations. We conclude, inter alia, that — as an obvious violation of a juror’s oath and duty — a refusal to apply the law as set forth by the court constitutes grounds for dismissal under Rule 23(b). We also conclude that the importance of safeguarding the secrecy of the jury deliberation room, coupled with the need to protect against the dismissal of a juror based on his doubts about the guilt of a criminal defendant, require that a juror be dismissed for a refusal to apply the law as instructed only where the record is clear beyond doubt that the juror is not, in fact, simply unpersuaded by the prosecution’s case. We hold that the district court erred in dismissing a juror, based largely on its finding that the juror was purposefully disregarding the court’s instructions on the law, where the record evidence raised the possibility that the juror’s view on the merits of the case was motivated by doubts about the *609defendants’ guilt, rather than by an intent to nullify the law. Accordingly, we vacate the judgments of the district court and remand for a new trial.
I.
We have before us the consolidated appeals of ten criminal defendants convicted of related conduct in two trials held in the Northern District of New York. The named defendants in this ease, including those whose appeals we consider here, were arrested on May 5, 1994. In an indictment returned on May 13, 1994, they were charged with conspiracy to possess and distribute cocaine and crack cocaine and actual possession and distribution of these substances. A 30-count, superseding indictment was returned on October 14, 1994, which added a series of forfeiture counts against the defendants.
Ceasare Thomas, Myron Thomas, Lamont Joseph, Santo Bolden, and Raymond Eaddy were tried on charges set forth in the superseding indictment beginning on November 22, 1994. After a Government witness apparently made certain prejudicial statements on the stand, a mistrial was declared on November 28,1994. A second trial of the same defendants began two days later, on November 30, 1994, and the jury returned verdicts of guilty for all defendants but Raymond Eaddy on December 14, 1994. We affirm these convictions in a summary order filed today. See United States v. Thomas et al, Nos. 96-1337 et al. (2d Cir. May 20, 1997).
The remaining appellants, Grady Thomas, Ramse Thomas, Jason Thomas, Tracey Thomas, and Loray Thomas, along with Terrence Thomas, Shawne Thomas, Carrie Thomas, Stephon Russell, and Robert Gibson, were the subject of a separate trial, which began on January 18, 1995.2 Grady Thomas, Ramse Thomas, Jason Thomas, Tracey Thomas, and Loray Thomas appeal from judgments of conviction entered against them following this trial, and we consider their appeals here.3 We confine our factual discussion of this trial to the events leading up to and including the ultimate dismissal of one of the jurors. These events provide the basis for the appellants’ primary challenge to the proceedings below.
During jury selection, the Government attempted to exercise a peremptory challenge to a juror who would later be empaneled as “Juror No. 5.” Because the juror was black — indeed, the only black person remaining as a potential juror in a case in which, as the record indicates, all of the defendants are black — defense counsel objected to the peremptory challenge under Batson v. Kentucky, 476 U.S. 79, 106 S.Ct. 1712, 90 L.Ed.2d 69 (1986), as racially motivated. The Government responded that it wished to exclude the juror based not on his race, but on the fact that he failed to make eye contact with the Government’s counsel during the voir dire. Although the district court explicitly found that the Government’s peremptory challenge was not motivated by race, the court, in a misapplication of Batson,4 nevertheless denied the challenge on the ground that the juror’s failure to make eye contact was an insufficient basis for his removal. The court would later explain that Juror No. 5’s status as the only black juror in a case involving black defendants had motivated its decision to deny the Government’s challenge.
Problems regarding Juror No. 5 did not end with his selection for the jury, however. During the course of defense summations on Friday, February 17, 1995, following several weeks of trial, a group of six jurors approached the courtroom clerk to express *610their concerns about the juror. The six jurors complained that Juror No. 5 was distracting them in court by squeaking his shoe against the floor, rustling cough drop wrappers in his pocket, and showing agreement with points made by defense counsel by slapping his leg and, occasionally during the defense summations, saying “[y]eah, yes.”
Chief Judge McAvoy met with counsel in chambers to discuss the complaints about Juror No. 5. The judge raised the possibility of conducting interviews with each member of the jury to determine the extent to which Juror No. 5 was distracting them from their duties. Alternatively, he considered dismissing Juror No. 5 in favor of an alternate juror pursuant to Fed.R.CRIM.P. 24(c).5 While the Government approved of the idea of interviewing the jurors, and dismissing Juror No. 5 if the interviews revealed that his behavior was disturbing other members of the jury, defense counsel generally opposed both options, preferring that the court permit summations to continue with only a general instruction to the jurors that they were not to form any opinions before starting their deliberations. In the face of conflicting recommendations from counsel on how to proceed, the judge dismissed the jurors for the day, requesting that counsel submit memoranda of law over the three-day weekend on an appropriate course of action. The court did, however, instruct the jurors before dismissing them that they should not yet “form any opinions or conclusions about the case.”
The court received letter briefs from the Government and from counsel for Ramse Thomas, dated February 17 and 20, 1995, respectively, on the proper course of action regarding Juror No. 5. The Government recommended in its brief that the court conduct in camera, on-the-record interviews with the jurors (outside the presence of counsel to promote candid responses) to determine whether there were grounds to dismiss Juror No. 5 — specifically, to determine whether the juror “had been disruptive to the point where the other jurors don’t believe that they can deliberate with him, ... has formed opinions about the case that he has communicated to the other jurors, or if there is some other misconduct found which establishes that he is unable to render a fair and impartial verdict.” In his brief, counsel for Ramse Thomas argued that the court should refrain from conducting any juror interviews, urging that such an inquiry might bias members of the jury against the defendants. He also objected to the dismissal of Juror No. 5.
When the jury returned from the three-day weekend on Tuesday, February 21, the court followed the procedure recommended by the Government.- Without counsel present, the court conducted in camera, on-the-record interviews with each juror to determine the extent of any distraction resulting from Juror No. 5’s behavior in the jury box. Quite appropriately, the court began each interview with a general inquiry as to whether anything had happened during the course of trial -that would interfere with the juror’s ability to deliberate and decide the case properly; the court did not ask about Juror No. 5’s behavior unless the interviewee first raised the issue. Although seven of the jurors indicated that Juror No. 5 was a source of some distraction, all but one, who “thought possibly” that she would experience problems during deliberations because of Juror No. 5, anticipated nothing that would interfere with their own ability to deliberate. For his part, Juror No. 5 explained to the court that he sometimes got “carried away” in listening to the attorneys’ argumente, but he stated that he would have no difficulty in applying the law as set out by the court to the evidence presented at trial. He also assured the court that he would “restrain [himjself” from engaging in any further distracting behavior.
After completing his in camera interviews, Chief Judge McAvoy met with counsel in chambers. He explained that he had interviewed the jurors, briefly summarized the testimony he had received in the interviews, and expressed his intention, based on these interviews, to remove Juror No. 5 in favor of one of the alternate jurors. The judge indicated that he was concerned that Juror No. *6115’s behavior, especially in light of the court’s own inquiries of the jurors, might place him in an adversarial relationship with his fellow jurors as they began deliberations. The judge then sought comments from each of the parties’ counsel. The Government agreed with the court’s proposal to remove Juror No. 5, but the proposed dismissal met with unanimous opposition from defense counsel. Apparently persuaded by the defense’s vigorous objections, the judge reconsidered the matter and decided to retain Juror No. 5 on the panel. Following the court’s meeting with counsel, summations concluded, and the court charged the jury that same day.
The jury deliberated throughout the day on February 22. On February 23, the courtroom clerk reported to the court, and then on the record to all counsel, that she had been approached on two separate occasions earlier in the day by jurors expressing concern over the course of their deliberations. Juror No. 1 reportedly had indicated to the clerk that deliberations were likely to continue beyond February 23 because of a “problem with an unnamed juror.” That same morning, Juror No. 12 had also reported to the clerk “that there was a problem ... in the jury room [with] one of their number, and specifically ... indicated [that] juror number five, had, at each time a vote was taken, voted not guilty and had indicated verbally that he would not change his mind.” The court concluded, after hearing argument from counsel for the parties, that no action was immediately necessary; the court would “give it a little more time to see what develops.”
Troubles in the jury room seemed to escalate rapidly, however. On the following morning, February 24, the court received a note from Juror No. 6, apparently written only on his own behalf. The note indicated that, due to Juror No. 5’s “predisposed disposition,” the jury was unable to reach a verdict. Following an off-the-record conversation with counsel for the parties, the court again conducted in camera, on-the-record interviews with each of the jurors outside the presence of counsel. This time, jurors focused their comments more directly on Juror No. 5. Several mentioned the disruptive effect he was having on the deliberations. One juror described him “hollering” at fellow jurors, another said he had called his fellow jurors racists, and two jurors told the court that Juror No. 5 had come close to striking a fellow juror. The judge was also informed by a juror that, at one point, Juror No. 5 pretended to vomit in the bathroom while other jurors were eating lunch outside the bathroom door. The jurors, however, were not unanimous in identifying Juror No. 5 as a source of disruption in the jury room. One juror informed the judge that friction among the jurors had been “pretty well ironed out,” and another indicated that the other jurors were in fact “picking on” Juror No. 5.
Although the district court did not specifically inquire into any juror’s position on the merits of the case, at least five of the jurors indicated that Juror No. 5 was unyieldingly in favor of acquittal for all of the defendants. The accounts differed, however, regarding the basis for Juror No. 5’s position. On the one hand, one juror described Juror No. 5 as favoring acquittal because the defendants were his “people,” another suggested that it was because Juror No. 5 thought the defendants were good people, two others stated that Juror No. 5 simply believed that drug dealing is commonplace, and another two jurors indicated that Juror No. 5 favored acquittal because he thought that the defendants had engaged in the alleged criminal activity out of economic necessity. On the other hand, several jurors recounted Juror No. 5 couching his position in terms of the evidence — one juror indicated specifically that Juror No. 5 was discussing the evidence, and four recalled him saying that the evidence, including the testimony of the prosecution’s witnesses, was insufficient or unreliable. As for Juror No. 5, he said nothing in his interview with the court to suggest that he was not making a good faith effort to apply the law as instructed to the facts of the ease. On the contrary, he informed the court that he needed “substantive evidence” establishing guilt “beyond a reasonable doubt” in order to convict.
After interviewing the jurors, the judge met in chambers with counsel for the parties. He had the record of the interviews read *612aloud and permitted counsel to comment on the appropriate course of action. The Government argued that the jurors’ responses indicated that there was “almost a jury nullification issue pattern with [Juror No. 5],” and urged the court to order the juror’s dismissal, while defense counsel unanimously opposed his removal. Having heard argument from counsel, the judge rendered his decision to remove Juror No. 5. He believed that Juror No. 5 had become a “distraction” and a “focal point” for the jury’s attention, and that his removal might “allow [the jury] to deliberate in a full and a fair fashion.” The court cited Juror No. 5’s failure to live up to his assurances regarding proper conduct, referring in particular to the allegation that he nearly struck another juror and to his feigned vomiting. Most importantly, however, the court found that Juror No. 5 was ignoring the evidence in favor of his own, preconceived ideas about the case:
I believe after hearing everything that [Juror No. 5’s] motives are immoral, that he believes that these folks have a right to deal drugs, because they don’t have any money, they are in a disadvantaged situation and probably that’s the thing to do. And I don’t think he would convict them no matter what the evidence was.
The court found that Juror No. 5 was refusing to convict “because of preconceived, fixed, cultural, economic, [or] social ... reasons that are totally improper and impermissible.”
The court then called Juror No. 5 into chambers to inform him of his dismissal and, that afternoon, announced the dismissal to the remaining jurors. Jurors were instructed that they were “to draw no inferences or conclusions whatsoever” from the removal and told that they were to start over in their deliberations.
On the afternoon of the following Monday, February 27, 1995, the remaining eleven jurors returned a verdict. They found the defendants Grady, Ramse, Tracy, and Terrence Thomas guilty on all counts, Jason Thomas guilty on three of the four counts against him, and Carrie and Loray Thomas each guilty on a conspiracy count. The jury deadlocked on the fourth count against Jason Thomas and acquitted Carrie and Loray Thomas of possession with intent to distribute a controlled substance. Stephon Russell was acquitted of conspiracy to distribute and to possess with intent to distribute, the only count with which he had been charged.
Ramse, Tracey, Loray, Grady, and Jason Thomas here appeal from the judgment of conviction. As their chief argument on appeal, each of these defendants challenges the dismissal of Juror No. 5.
II.
The district court dismissed Juror No. 5 pursuant to Fed.R.Ceim.P. 23(b), which provides, in pertinent part, that where “the court finds it necessary to excuse a juror for just cause after the jury has retired to consider its verdict, in the discretion of the court a valid verdict may be returned by the remaining 11 jurors.” We review the district court’s exercise of this authority for abuse of discretion. See, e.g., United States v. Reese, 33 F.3d 166, 173 (2d Cir.1994), cert. denied, 513 U.S. 1092, 115 S.Ct. 756, 130 L.Ed.2d 655 (1995); United States v. Casamento, 887 F.2d 1141, 1187 (2d Cir.1989), cert. denied, 493 U.S. 1081, 110 S.Ct. 1138, 107 L.Ed.2d 1043 (1990).
To determine whether the court erred in dismissing Juror No. 5, we must first decide whether the district court’s primary basis for the dismissal — the juror’s intention to disregard the applicable criminal laws — constitutes “just cause” for his removal under Rule 23(b). In holding that a presiding judge has a duty to dismiss a juror who purposefully disregards the court’s instructions on the law, we briefly review the factors that courts have traditionally considered to be “just cause” for dismissal pursuant to Rule 23(b), and discuss the dangers inherent in so-called “nullification.” Having concluded that a deliberating juror’s intent to nullify constitutes “just cause” for dismissal, we next consider whether the district court in this case had a sufficient evidentiary basis for concluding that Juror No. 5 was purposefully disregarding the court’s instructions on the law.
*613A. Dismissal of a Juror During Deliberations: Rule 23(b) and Factors that Traditionally Constitute “Just Cause”
In evaluating the district court’s decision to remove Juror No. 5 pursuant to Rule 23(b), we must first decide whether the reasons that the court cited as grounds for the removal constitute “just cause” as that term is employed in the rule. We consider, in particular, the district court’s primary ground for dismissal — that Juror No. 5 refused to apply the law as set out in the court’s instructions.6 Whether a juror’s defiance of the court’s instructions on the law constitutes “just cause” for that juror’s removal under Rule 23(b) is apparently a question of first impression in this Circuit.
Language was added in 1983 to Rule 23(b) to provide a court with the unilateral authority to remove jurors during the course of deliberations. Prior to that time, Rule 23(b) required the consent of all parties in order for the trial court to dismiss one or more jurors, and to allow the remaining jurors to proceed to a verdict.7 As explained in the Note of the Advisory Committee on the Federal Rules of Criminal Procedure (“Advisory Committee”), the 1983 amendment was a response to cases in which, after a trial of significant length and involving substantial expense, a juror became “seriously incapacitated or [was] otherwise found to be unable to continue service upon the jury.” Fed. R.CRIM.P. 23(b) Advisory Committee’s Note (“Advisory Committee Note”); see United States v. Gabay, 923 F.2d 1536, 1543 (11th Cir.1991) (noting that, in that case, “a mistrial would have necessitated a second expenditure of substantial prosecution, defense and court resources ... the outcome Rule 23(b) was designed to alleviate”). The amendment provides an alternative short of mistrial in such cases, and it does so without calling for the use of alternate jurors once deliberations have begun, an option that the Advisory Committee expressly rejected. Advisory Committee Note; see also United States v. Gambino, 788 F.2d 938, 948-49 (3d Cir.), cert. denied, 479 U.S. 825, 107 S.Ct. 98, 93 L.Ed.2d 49 (1986).
Since the adoption of the 1983 amendment, federal courts have exercised their unilateral authority to dismiss jurors during the course of deliberations for a variety of reasons. Often, courts employ Rule 23(b) in cases, like those described in the Advisory Committee Note, where a juror is incapacitated or has otherwise become unavailable during the course of deliberations. See, e.g., Reese, 33 F.3d at 172-73 (juror leaving for business trip); United States v. Wilson, 894 F.2d 1245, 1249-51 (11th Cir.) (juror became ill), cert. denied, 497 U.S. 1029, 110 S.Ct. 3284, 111 L.Ed.2d 792 (1990); United States v. Armijo, 834 F.2d 132, 134 (8th Cir.1987) (juror in car accident), cert. denied, 485 U.S. 990, 108 S.Ct. 1297, 99 L.Ed.2d 507 (1988); United States v. Molinares Charris, 822 F.2d 1213, 1222-23 (1st Cir.1987) (juror was “nervous and upset,” had been crying during deliberations, and had taken tranquilizer); United States v. Stratton, 779 F.2d 820, 830-31 (2d Cir.1985) (juror unable to deliberate on religious holiday), cert. denied, 476 U.S. 1162, 106 S.Ct. 2285, 90 L.Ed.2d 726 (1986).
“Just cause” is not limited to instances of juror illness or unavailability, however. Courts have also found “just cause” to dismiss jurors who, although available and physically capable of serving, are nonetheless found to be unable to perform their duties properly. In particular, Rule 23(b) dismissals have been upheld repeatedly in cases where the trial court found that a juror was no longer capable of rendering an impartial verdict. These cases have involved instances of jurors who felt threatened by one of the parties, see United States v. Ruggiero, 928 F.2d 1289, 1300 (2d Cir.) (juror “disabled by fear” after receiving what he thought was threat from defendant), cert, denied sub nom. Gotti v. United States, 502 U.S. 938, *614112 S.Ct. 372, 116 L.Ed.2d 824 (1991); Casamento, 887 F.2d at 1186-87 (juror fearful after daughter received threatening phone call), who are discovered to have a relationship with one of the parties, see United States v. Barone, 846 F.Supp. 1016, 1018-19 (D.Mass.1994) (juror who was informed that defense attorney had represented his cousin deemed unable to “render a fair and impartial verdict”), or whose life circumstances otherwise change during the course of deliberations in such a way that they are no longer considered capable of rendering an impartial verdict, see United States v. Egbuniwe, 969 F.2d 757, 762-63 (9th Cir.1992) (juror “might not be able to be fair to both parties” after learning that girlfriend had been arrested and mistreated by police).
B. Nullification as “Just Cause” for Dismissal
Here, Chief Judge McAvoy identified a different form of bias as the primary ground for dismissing Juror No. 5 — one arising not from an external event or from a relationship between a juror and a party, but rather, from a more general opposition to the application of the criminal narcotics laws to the defendants’ conduct. In the court’s view, Juror No. 5 believed that the defendants had “a right to deal drugs.” Based on what the court described as the juror’s “preconceived, fixed, cultural, economic, [or] social ... reasons that are totally improper and impermissible,” the court concluded that Juror No. 5 was unlikely to convict the defendants “no matter what the evidence was.” Essentially, the judge found that Juror No. 5 intended to engage in a form of “nullification,” a practice whereby a juror votes in purposeful disregard of the evidence, defying the court’s instructions on the law.
We take this occasion to restate some basic principles regarding the character of our jury system. Nullification is, by definition, a violation of a juror’s oath to apply the law as instructed by the court — in the words of the standard oath administered to jurors in the federal courts, to “render a true verdict according to the law and the evidence.” FEDERAL Judicial Center, Benchbook for U.S. DistriCT Court Judges 225 (4th ed. 1996) (emphasis supplied).8 We categorically reject the idea that, in a society committed to the rule of law, jury nullification is desirable or that courts may permit it to occur when it is within their authority to prevent. Accordingly, we conclude that a juror who intends to nullify the applicable law is no less subject to dismissal than is a juror who disregards the court’s instructions due to an event or relationship that renders him biased or otherwise unable to render a fair and impartial verdict.
We are mindful that the term “nullification” can cover a number of distinct, though related, phenomena, encompassing in one word conduct that takes place for a variety of different reasons; jurors may nullify, for example, because of the identity of a party, a disapprobation of the particular prosecution at issue, or a more general opposition to the applicable criminal law or laws. We recognize, too, that nullification may at times manifest itself as a form of civil disobedience that some may regard as tolerable. The case of John Peter Zenger, the publisher of the New York Weekly Journal acquitted of criminal libel in 1735, and the nineteenth-century acquittals in prosecutions under the fugitive slave laws, are perhaps our country’s most renowned examples of “benevolent” nullification. See United States v. Dougherty, 473 F.2d 1113, 1130 (D.C.Cir.1972) (Leventhal, J.); see also Shannon C, Stimson, The American Revolution In The Law: Anglo-American JURISPRUDENCE BEFORE JOHN MARSHALL 52-55 (1990) (describing Zenger trial). We are also aware of the long and complicated history of juries acting as judges of the law as well as the evidence, see, e.g., John D. Gordan III, Juries as Judges of the Law: The American Experience, 108 Law Q. Rev. 272 (1992); Mark De Wolfe Howe, Juries as Judges of Criminal Law, 52 Harv. L. Rev. 582 (1939), and of the theoretical underpinnings of this practice in the United States, in which legal decisions by juries were sometimes regarded as an expression of faithfulness to the law (regardless of the authority of institutions or officeholders), rather than defiance of the law or “nullification.”
*615More generally, the very institution of trial by jury in a criminal case, as Judge Learned Hand observed, “introduces a slack into the enforcement of law, tempering its rigor by the mollifying influence of current ethical conventions.” U.S. ex rel. McCann v. Adams, 126 F.2d 774, 776 (2d Cir.), rev’d on other grounds, 317 U.S. 269, 68 S.Ct. 236, 87 L.Ed. 268 (1942). This is so because, as Judge Hand explained, “[t]he individual can forfeit his liberty — to say nothing of his life— only at the hands of those who, unlike any official, are in no wise accountable, directly or indirectly, for what they do, and who at once separate and melt anonymously in the community from which they came.... [S]inee if they acquit their verdict is final, no one is likely to suffer of whose conduct they do not morally disapprove-” Id. at 775-76.
As courts have long recognized, several features of our jury trial system act to protect the jury’s power to acquit, regardless of the evidence, when the prosecution’s case meets with the jury’s “moral[ ] disap-prov[al].” Since the emergence of the general verdict in criminal eases and the famous opinion in Bushell’s Case, 124 Eng.Rep. 1006 (C.P.1670), freeing a member of the jury arrested for voting to acquit William Penn against the weight of the evidence, nullifying jurors have been protected from being called to account for their verdicts. Moreover, and in addition to the courts’ duty to safeguard the secrecy of the jury deliberation room (discussed in greater detail below), the several rules protecting the unassailability of jury verdicts of acquittal — even where these verdicts are inconsistent with other verdicts rendered by the same jury in the same case, United States v. Carbone, 378 F.2d 420, 423 (2d Cir.) (Friendly, J.) (recognizing link between upholding inconsistent verdicts and protecting juries’ power of lenity), cert. denied, 389 U.S. 914, 88 S.Ct. 242, 19 L.Ed.2d 262 (1967) — serve to “permit[ ] juries to acquit out of compassion or compromise or because of their assumption of a power which they had no right to exercise, but to which they were disposed through lenity.” Standefer v. United States, 447 U.S. 10, 22, 100 S.Ct. 1999, 2007, 64 L.Ed.2d 689 (1980) (internal quotation marks omitted).
But as the quotation from the Supreme Court’s opinion in Standefer indicates, in language originally employed by Judge Learned Hand, the power of juries to “nullify” or exercise a power of lenity is just that — a power; it is by no means a right or something that a judge should encourage or permit if it is within his authority to prevent. It is true that nullification has a long history in the Anglo-American legal system, see Dougherty, 473 F.2d at 1130-33, and that the federal courts have long noted the de facto power of a jury to render general verdicts “in the teeth of both law and facts,” Horning v. District of Columbia, 254 U.S. 135, 138, 41 S.Ct. 53, 54, 65 L.Ed. 185 (1920); see, e.g., United States v. Trujillo, 714 F.2d 102, 105-06 (11th Cir.1983). However, at least since the Supreme Court’s decision in Sparf v. United States, 156 U.S. 51, 102 (1895) (holding that, while juries are finders of fact, “it is the duty of juries in criminal cases to take the law from the court and apply that law to the facts as they find them”), courts have consistently recognized that jurors have no right to nullify. See Gordan, supra, at 272, 277 (noting that, with Sparf, the Supreme Court “fixed the law where Lord Mansfield had left it” in King v. Shipley (“The Dean of St. Asaph’s Case”), 4 Doug. 73 (K.B. 1784), in which Mansfield had written that jurors have the power, but not the right, to decide the law); Howe, supra, at 589 (referring to Sparf as “the Supreme Court’s final and authoritative denial of the .[jury’s] right” to serve as judges of the law); see, e.g., United States v. Kerley, 838 F.2d 932, 938 (7th Cir.1988) (“[J]ury nullification is just a power, not also a right-”). As a panel of the Court of Appeals for the District of Columbia Circuit — composed of Chief Judge Spottswood W. Robinson, III, Judge George E. Mac-Kinnon, and then-Judge Ruth Bader Ginsburg — explained:
A jury has no more “right” to find a “guilty” defendant “not guilty” than it has to find a “not guilty” defendant guilty, and the fact that the former cannot be corrected by a court, while the latter can be, does not create a right out of the power to misapply the law. Such verdicts are law*616less, a denial of due process and constitute an exercise of erroneously seized power.
United States v. Washington, 705 F.2d 489, 494 (D.C.Cir.1983) (per curiam) (emphasis in original). Indeed, as we noted above, the exercise of this de facto power is a violation of a juror’s sworn duty to “apply the law as interpreted by the court.” United States v. Boardman, 419 F.2d 110, 116 (1st Cir.1969), cert. denied, 397 U.S. 991, 90 S.Ct. 1124, 25 L.Ed.2d 398 (1970).9
Moreover, although the early history of our country includes the occasional Zenger trial or acquittals in fugitive slave eases, more recent history presents numerous and notorious examples of jurors nullifying— cases that reveal the destructive potential of a practice Professor Randall Kennedy of the Harvard Law School has rightly termed a “sabotage of justice.” Randall Kennedy, The Angry Juror, Wall St. J., Sept. 30, 1994, at A12. Consider, for example, the two hung juries in the 1964 trials of Byron De La Beckwith in Mississippi for the murder of NAACP field secretary Medgar Evers, or the 1955 acquittal of J.W. Millam and Roy Bryant for the murder of fourteen-year-old Emmett Till, see David HalbeRSTam, The Fifties 431-41 (1993); Randall Kennedy, Race, Crime and the Law 60-63, 250 (1997); Juan WilliaMS, Eyes on the Prize: Amer-ioa’s Civil Rights Years, 1954-1965, at 38-57, 221-25 (1987) — shameful examples of how nullification has been used to sanction murder and lynching.
Inasmuch as no juror has a right to engage in nullification — and, on the contrary, it is a violation of a juror’s sworn duty to follow the law as instructed by the court— trial courts have the duty to forestall or prevent such conduct, whether by firm instruction or admonition or, where it does not interfere with guaranteed rights or the need to protect the secrecy of jury deliberations, see infra Section II.C, by dismissal of an offending juror from the venire or the jury. If it is true that the jury’s “prerogative of lenity,” Dougherty, 473 F.2d at 1133, introduces “a slack into the enforcement of law, tempering its rigor by the mollifying influence of current ethical conventions,” Adams, 126 F.2d at 776, then, as part and parcel of the system of checks and balances embedded in the very structure of the American criminal trial, there is a countervailing duty and authority of the judge to assure that jurors follow the law. Although nullification may sometimes succeed — because, among other things, it does not come to the attention of a presiding judge before the completion of a jury’s work, and jurors are not answerable for nullification after the verdict has been reached — it would be a dereliction of duty for a judge to remain indifferent to reports that a juror is intent on violating his oath. This is true regardless of the juror’s motivation for “nullification,” including race, ethnicity or similar considerations. A federal judge, whose own oath of office requires the judge to “faithfully and impartially discharge and perform all the duties incumbent upon [the judge] ... under the Constitution and laws of the United States,” 28 U.S.C. § 453 (1994), may not ignore colorable claims that a juror is acting on the basis of such improper considerations.
Accordingly, every day in courtrooms across the length and breadth of this country, jurors are dismissed from the venire “for cause” precisely because they are unwilling or unable to follow the applicable law.10 In*617deed, one of the principal purposes of voir dire is to ensure that the jurors ultimately selected for service are unbiased and willing and able to apply the law as instructed by the court to the evidence presented by the parties.
So also, a presiding judge possesses both the responsibility and the authority to dismiss a juror whose refusal or unwillingness to follow the applicable law becomes known to the judge during the course of trial. Rule 24(c) of the Federal Rules of Criminal Procedure provides for the substitution of alternates for “jurors who, prior to the time the jury retires to consider its verdict, become or are found to be unable or disqualified to perform their duties.” Surely a juror is “unable or disqualified,” for purposes of this rule, who is intent on nullifying the applicable law and thereby violating his oath to “render a true verdict according to the law and the evidence.”
Similarly, we conclude that a juror who is determined to ignore his duty, who refuses to follow the court’s instructions on the law and who thus threatens to “undermine! 1 the impartial determination of justice based on law,” Krzyske, 836 F.2d at 1021, is subject to dismissal during the course of deliberations under Rule 23(b). This conclusion reinforces the court’s inherent authority to conduct inquiries in response to reports of improper juror conduct and to determine whether a juror is unwilling to carry out his duties faithfully and impartially. The rule we adopt applies with equal force whether the juror’s refusal to follow the court’s instructions results from a desire to “nullify” the applicable law or, for example, as in the cases described above, see supra pp. 613-14, from a perceived physical threat or from a relationship with one of the parties.
Our position in this respect is in accord with that of the Eleventh Circuit. See United States v. Geffrard, 87 F.3d 448, 450-52 (11th Cir.), cert. denied, - U.S.-, 117 S.Ct. 442, 136 L.Ed.2d 339 (1996). In Geffrard, a juror submitted a letter to the court during the course of deliberations in which she stated that she adhered to the Christian teachings of Emanuel Swedenborg. Under Swedenborg’s theology, the juror explained, she could not “ ‘live with a verdict of guilty for any of the accused on any of the charges, as [she] believe[d] deep within [her] heart and soul and mind that [the defendants] were unjustly led into this so called transaction by a more intelligent and powerful figure.’ ” Id. at 451. The juror was convinced, assertedly as a result of her religious beliefs, that the defendants were the victims of governmental entrapment, notwithstanding the fact that the court had earlier instructed the jury that entrapment was not at issue in the case. Id. The juror in Geffrard thus was prepared purposefully to disregard or ignore — to “nullify” — the law as set forth in the court’s instructions to the jury. The district court “saw in the letter an inability of the juror to follow the court’s instructions on the law.” Id. Accordingly, the court dismissed the juror pursuant to Rule 23(b), and the remaining eleven jurors convicted the defendants. The Court of Appeals upheld the dismissal, reasoning that the juror’s letter “[made] it a certainty that this particular juror could not reach a verdict following the judge’s instructions as applied to the facts.” Id. at 452.
We agree that a juror’s purposeful refusal to apply the law as set forth in a jury charge constitutes an appropriate basis for that juror’s removal. Nor is this conclusion any less valid, in the instant case, in light of Juror No. 5’s race. The rule authorizing dismissal of a juror who disregards the law does not include an exception for jurors who violate their sworn duty on the basis of racial or ethnic interests or affinities. Accordingly, *618the district court’s finding that Juror No. 5 was unlikely to convict the defendants “no matter what the evidence” was a proper basis for the exercise of the court’s dismissal authority, provided that the court had a sufficient evidentiary basis for this finding. As we explain below, however, the need to safeguard the secrecy of jury deliberations requires the use of a high evidentiary standard for the dismissal of a deliberating juror for purposeful disobedience of a court’s instructions, a standard that the record in the instant ease does not meet.
C. Jury Secrecy and the Investigation of Alleged Juror Impropriety
The extent to which a presiding judge may investigate alleged juror bias or misconduct differs depending on when the investigation takes place. In particular, and as we explain below, a district court’s authority to investigate allegations of juror impropriety necessarily becomes more limited once the jury has begun to deliberate. Once a jury retires to the deliberation room, the presiding judge’s duty to dismiss jurors for misconduct comes into conflict with a duty that is equally, if not more, important — safeguarding the secrecy of jury deliberations. This conflict is especially pronounced here, where the alleged misbehavior is a purposeful disregard of the law, a particularly difficult allegation to prove and one for which an effort to act in good faith may easily be mistaken.
In the instant case, the defendants do not directly challenge the proposition that a juror’s refusal to follow the court’s instructions on the law may be a proper basis for removal under Rule 23(b), but they do contest the court’s finding of fact that Juror No. 5 was unwilling to apply the law as instructed. The defendants contend that the record provides ample evidence to demonstrate that Juror No. 5 simply remained unpersuaded by the Government’s case. See Brief for Defendant-Appellant Grady Thomas at 13, 16-17; Brief for Defendant-Appellant Loray Thomas at 18, 20; Brief for Defendant-Appellant Ramse Thomas at 26, 28; Brief for Defendant Appellant Tracey Thomas at 11; Brief for Defendant-Appellant Jason Thomas at 4— 5. In addition, two of the defendants argue that the very procedures employed by the district court to investigate reports of Juror No. 5’s misconduct — the two sets of in camera interviews with members of the jury (in the absence of counsel and, on at least the first occasion, over the objection of defense counsel) — are themselves grounds for reversal insofar as they unduly intruded into the jury’s deliberative process. See Brief for Defendant-Appellant Grady Thomas at 18-21; Brief for Defendant-Appellant Ramse Thomas at 20-23. As these claims are inextricably related, we consider them together below. First, however, it is important to restate the importance of secrecy in jury deliberations.
1. Safeguarding the Secrecy of Jury Deliberations
Courts face a delicate and complex task whenever they undertake to investigate reports of juror misconduct or bias during the course of a trial. This undertaking is particularly sensitive where, as here, the court endeavors to investigate allegations of juror misconduct during deliberations. As a general rule, no one — including the judge presiding at a trial — has a “right to know” how a jury, or any individual juror, has deliberated or how a decision was reached by a jury or juror. The secrecy of deliberations is the cornerstone of the modem Anglo-American jury system. While jury secrecy, like rules insulating jury verdicts of acquittal from prosecutorial challenge, necessarily serves to permit nullification to occur, see supra p. 615, its primary contemporary purpose lies elsewhere. Indeed, courts and commentators alike recognize that the secrecy of deliberations is essential to the proper functioning of juries. It is well understood, for example, that disclosure of the substance of jury deliberations may undermine public confidence in the jury system, see Note, Public Disclosures of Jury Deliberations, 96 Haev. L.Rev. 886, 889 (1983) (“Public Disclosures”), and poses a threat to adjudicatory finality. Especially troublesome is the danger that such disclosure presents to the operation of the deliberative process itself. As one commentator has observed:
Juror privacy is a prerequisite of free debate, without which the decisionmaking *619process would be crippled. The precise value of throwing together in a jury room a representative cross-section of the community is that a just consensus is reached through a thoroughgoing exchange of ideas and impressions. For the process to work according to theory, the participants must feel completely free to dissect the credibility, motivations, and just deserts of other people. Sensitive jurors will not engage in such a dialogue without some assurance that it will never reach a larger audience.
Id. (footnotes omitted). “Freedom of debate,” as Justice Cardozo wrote, “might be stifled and independence of thought cheeked if jurors were made to feel that their arguments and ballots were to be freely published to the world.” Clark v. United States, 289 U.S. 1, 13, 53 S.Ct. 465, 468-69, 77 L.Ed. 993 (1933); see also United States v. Antar, 38 F.3d 1348, 1367 (3d Cir.1994) (Rosenn, J, concurring) (“We must bear in mind that the confidentiality of the thought processes of jurors, their privileged exchange of views, and the freedom to be candid in their deliberations are the soul of the jury system.”); In re Globe Newspaper Co., 920 F.2d 88, 94 (1st Cir.1990) (“It is undisputed that the secrecy of jury deliberations fosters free, open and candid debate in reaching a decision.”); Abraham S. Goldstein, Jury Secrecy and the Media: The Problem of Postverdict Interviews, 1993 U.Ill.L.Rev. 295, 295 (“[Jjurors must deliberate in secret so that they may communicate freely with one another, secure in the knowledge that what they say will not be passed along to others.”); Benjamin S. DuVal, Jr., The Occasions of Secrecy, 47 U.Pitt.L.Rev. 579, 646 (1986) (“The secrecy of the jury room, like that of the Supreme Court conference, is designed to promote the free and candid interchange of views.”).
The history of Anglo-American law and common experience in our own time lend little or no support to the occasional suggestion, see, e.g., Marcy Strauss, Juror Journalism, 12 Yale L. & Pol’y Rev. 389, 404-05 (1994), that public scrutiny of how jurors reach their decisions may improve the quality of jury deliberations. In cases that generate much attention or passion in the community, or that involve allegedly dangerous persons or organizations, the mere suggestion that the views of jurors may be conveyed to the parties and the public, even after the trial is over, understandably may cause anxiety and fear in jurors, and distort the process by which a verdict is reached; actually making such information available to the public might invite the retribution that jurors would rightly fear.
The jury system incorporated in our Constitution by the Framers was not intended to satisfy yearnings for perfect knowledge of how a verdict is reached, nor to provide assurances to the public of the primacy of logic in human affairs. Nor was it subordinated to a “right to know” found in the First Amendment. The jury as we know it is supposed to reach its decisions in the mystery and security of secrecy; objections to the secrecy of jury deliberations are nothing less than objections to the jury system itself.
Because the rule of secrecy is fundamental to the effective operation of the jury system, it is not surprising that courts have been concerned to maintain the confidentiality of the process even after a verdict has been returned and the jury has been formally discharged. It is the historic duty of a trial judge to safeguard the secrecy of the deliberative process that lies at the heart of our system of justice, even in the face of relentless, and sometimes inappropriate, demands by the news media and the public for post-verdict disclosure of what went on behind the closed door of the jury room. See, e.g., Antar, 38 F.3d at 1364 (Rosenn, J., concurring) (noting “the historic efforts of the courts to protect the confidentiality of a jury’s deliberative process”); id. at 1367 (stating that juror deliberations “must be zealously guarded from any impermissible encroachment if the [jury] system is to survive”); see also Fed. R. Evid. 606(b) (discussed infra p. 623 & n.13).
Today, it is common — and entirely appropriate — for a conscientious trial judge to advise jurors against disclosing the substance of their deliberations after the end of a trial. See Globe Newspaper Co., 920 F.2d at 94-95 (“It has ... been a common and, we believe, wise custom for trial judges to advise jurors ... that they are not only free to *620refuse to disclose what went on in the jury-room, but that they may well think it better and more prudent to decline to discuss what has occurred.”)- At times, courts quite properly go further than this to protect the secrecy of deliberations, imposing strict limitations on what jurors are permitted to disclose. See, e.g., United States v. Harrelson, 713 F.2d 1114, 1118 (5th Cir.1983) (upholding trial court restriction on juror interviews, including requirement that “[n]o interviewer may inquire into the specific vote of any juror other than the juror being interviewed”), cert, denied sub nom. El Paso Times, Inc. v. United States Dist. Ct. for the Western Dist. of Tx., 465 U.S. 1041, 104 S.Ct. 1318, 79 L.Ed.2d 714 (1984). In addition, many federal judicial districts have enacted rules that subject post-verdict juror interviews to judicial supervision. See Public Disclosures, supra, at 901 & n. 93 (citing rules from 26 federal districts). While these rules often apply only to parties and their counsel, some federal districts have adopted rules that extend the court’s supervisory authority to any post-verdict interviews of jurors. Id. at 901 & n. 95; see, e.g., D.Conn. R.Civ.P. 12(e)(1) (incorporated by reference into criminal rules pursuant to D.Conn. R.Crim.P. 1) (“No juror shall respond to any inquiry as to the deliberations or vote of the jury or of any other individual juror, except on leave of Court which shall be granted only upon the showing of good cause.” (emphasis supplied)). None of this is to suggest that we cannot do more to protect the secrecy of deliberations. One eminent authority, for example, has proposed a statute that would impose criminal sanctions on jurors who disclose information about their deliberations, as well as on anyone who seeks such disclosure, without permission of the court. See Goldstein, supra, at 308-10.
2. Investigating Alleged Juror Misconduct During Deliberations
Protecting the deliberative process requires not only a vigilant watch against external threats to juror secrecy, but also strict limitations on intrusions from those who participate in the trial process itself, including counsel and the presiding judge. A court must limit its own inquiries of jurors once deliberations have begun. As the district court observed in the instant case, the very act of judicial investigation can at times be expected to foment discord among jurors. See supra pp. 610-11. In particular, a presiding judge is limited in the extent to which he may investigate the reasons underlying a juror’s position on the merits of a case. United States v. Brown, 823 F.2d 591, 596 (D.C.Cir.1987) (“[A] court may not delve deeply into a juror’s motivations because it may not intrude on the secrecy of the jury’s deliberations.”). The mental processes of a deliberating juror with respect to the merits of the case at hand must remain largely beyond examination and second-guessing, shielded from scrutiny by the court as much as from the eyes and ears of the parties and the public. Were a district judge permitted to conduct intrusive inquiries into — and make extensive findings of fact concerning — the reasoning behind a juror’s view of the case, or the particulars of a juror’s (likely imperfect) understanding or interpretation of the law as stated by the judge, this would not only seriously breach the principle of the secrecy of jury deliberations, but it would invite trial judges to second-guess and influence the work of the jury.
In many cases, a presiding judge is able to determine whether there is “just cause” to dismiss a deliberating juror without any inquiry into the juror’s thoughts on the merits of the case. Evidence of the nature and extent of a juror’s unavailability, see Reese, 33 F.3d at 172-73 (business trip); Stratton, 779 F.2d at 830-31 (religious holiday), or incapacitation, see Wilson, 894 F.2d at 1249-51 (illness), for example, is ordinarily available without inquiring into the substance of deliberations. In such instances, the judge is free to conduct a thorough examination of the basis for removal — a basis that is itself unlikely to be confused with a juror’s views of the sufficiency of the evidence — and to make appropriate findings of fact, including determinations of the credibility of the juror in question.
The need to protect the secrecy of jury deliberations begins to limit the court’s investigatory powers where the asserted basis for a deliberating juror’s possible dismissal is the *621juror’s alleged bias or partiality in joining or not joining the views of his colleagues. As the examples set forth in Section II.A reveal, however, claims of partiality or bias often arise from some event, or from a relationship between a juror and a party, that is both easily identifiable and subject to investigation and findings without intrusion into the deliberative process. In Ruggiero, for example, the juror in question testified that he had received what he took to be a physical threat from one of the defendants. 928 F.2d at 1300. Likewise, in Casamento, a juror’s daughter had received a threatening phone call. 887 F.2d at 1186-87. Although the jurors dismissed for bias in Barone and Eg-buniwe had not been the subject of threats, the source of their possible bias was similarly subject to ready identification. See Barone, 846 F.Supp. at 1018-19 (defense attorney had represented juror’s cousin); Egbuniwe, 969 F.2d at 762-63 (during course of deliberations, juror learned that girlfriend had been arrested and mistreated by police).
In cases such as these, the presiding judge can make appropriate findings and establish whether a juror is biased or otherwise unable to serve without delving into the reasons underlying the juror’s views on the merits of the case. Rather, an event or relationship itself becomes the subject of investigation, with the trial court considering the likelihood that it will prejudice or otherwise disable the juror or jurors in question. Moreover, to the extent that in some of these eases the court is investigating whether a juror has become distracted or agitated following a particular incident, the court is able to rely on its assessment of the demeanor of the juror in question. Such cases may thus come to resemble those involving juror incapacitation. See Ruggiero, 928 F.2d at 1300 (juror “disabled by fear”). In Casamento, for example, Judge Leval, then sitting in the district court, quite appropriately weighed the juror’s assurance that the threatening phone call received by her daughter had not “affected [her] ability to think or to judge in any way,” against both the nature of the incident and the juror’s observable agitation. United States v. Badalamenti, 663 F.Supp. 1539, 1540-41 (S.D.N.Y.1987), aff'd, Casamento, 887 F.2d at 1187. Judge Leval grounded his decision to dismiss her from the jury on the finding that “[a] threat to the safety of one’s child ... cannot be put out of mind or disregarded,” and the fact that the juror “was obviously worried, troubled and upset, and repeatedly said so.” Id.
Where, however, as here, a presiding judge receives reports that a deliberating juror is intent on defying the court’s instructions on the law, the judge may well have no means of investigating the allegation without unduly breaching the secrecy of deliberations. There is no allegedly prejudicial event or relationship at issue, nor is the court being asked to assess whether a juror is so upset or otherwise distracted that he is unable to carry out his duties. Rather, to determine whether a juror is bent on defiant disregard of the applicable law, the court would generally need to intrude into the juror’s thought processes. Such an investigation must be subject to strict limitations. Without such an inquiry, however, the court will have little evidence with which to make the often difficult distinction between the juror who favors acquittal because he is purposefully disregarding the court’s instructions on the law, and the juror who is simply unpersuaded by the Government’s evidence. Yet this distinction is a critical one, for to remove a juror because he is unpersuaded by the Government’s case is to deny the defendant his right to a unanimous verdict. See Brown, 823 F.2d at 596. In a ease involving a juror’s own request to be dismissed from duty because of what the prosecution interpreted to be an unwillingness to apply the law as instructed, Judge Mikva, in an opinion joined by Judge Bork and Judge Douglas H. Ginsburg, observed:
[A] court may not delve deeply into a juror’s motivations because it may not intrude on the secrecy of the jury’s deliberations. Thus, unless the initial request for [a juror’s] dismissal is transparent, the court will likely prove unable to establish conclusively the reasons underlying it. Given these circumstances, we must hold that if the record evidence discloses any ‘possibility that the request to discharge stems from the juror’s view of the suffi*622ciency of the government’s evidence, the court must deny the request.
Id. (emphasis supplied).
We adopt the Brown rule as an appropriate limitation on a juror’s dismissal in any case where the juror allegedly refuses to follow the law — whether the juror himself requests to be discharged from duty or, as in the instant case, fellow jurors raise allegations of this form of misconduct. Given the necessary limitations on a court’s investigatory authority in cases involving a juror’s alleged refusal to follow the law, a lower evidentiary standard could lead to the removal of jurors on the basis of their view of the sufficiency of the prosecution’s evidence.
Consider a case where, for example, a strong majority of the jury favors conviction, but a small set of jurors — perhaps just one— disagrees. The group of jurors favoring conviction may well come to view the “holdout” or “holdouts” not only as unreasonable, but as unwilling to follow the court’s instructions on the law. The evidentiary standard that we endorse today — that “if the record evidence discloses any possibility that” a complaint about a juror’s conduct “stems from the juror’s view of the sufficiency of the government’s evidence, the court must deny the request” — serves to protect these holdouts from fellow jurors who have come to the conclusion that the holdouts are acting lawlessly.11
This evidentiary standard protects not only against the wrongful removal of jurors; it also serves to protect against overly intrusive judicial inquiries, into the substance of the jury’s deliberations. A presiding judge faced with anything but unambiguous evidence that a juror refuses to apply the law as instructed need go no further in his investigation of the alleged nullification; in such circumstances, the juror is not subject to dismissal on the basis of his alleged refusal to follow the court’s instructions. A lower evidentiary' standard would encourage the court faced with ambiguous evidence of such impropriety to investigate further, eliciting testimony from jurors until enough evidence surfaced to affirm or reject allegations of juror nullification. One unavoidable consequence of imposing a lower evidentiary standard would thus be to open up the possibility that judges, in response to demands of counsel or otherwise, would wind up taking sides in disputes between jurors on allegations of juror nullification — in effect, to permit judicial interference with, if not usurpation of, the fact-finding role of the jury.12
We recognize that this standard — buttressing the core principle of the secrecy of jury deliberations — leaves open the possibility that jurors will engage in irresponsible activity that will remain outside the court’s powers to investigate or correct. It is an imperfect rule, no doubt, but one fully consistent with our history and traditions, in which the judge’s duty and authority to prevent nullification and the need for jury secrecy co-exist uneasily. We recall, too, that “[t]he jury system has worked out reasonably well overall, providing ‘play in the joints’ that imparts flexibility and avoid[s] undue rigidity .... acting as a ‘safety valve’ for exceptional cases, without being a wildcat or runaway institution,” Dougherty, 473 F.2d at 1134, and that our ultimate goal at trial is fairness and substantial justice — not perfection. See, e.g., Delaware v. Van Arsdall, 475 U.S. 673, 681, 106 S.Ct. 1431, 1436, 89 L.Ed.2d 674 (1986) (“As we have stressed on more than one *623occasion, the Constitution entitled a criminal defendant to a fair trial, not a perfect one.”). Where the duty and authority to prevent defiant disregard of the law or evidence comes into conflict with the principle of secret jury deliberations, we are compelled to err in favor of the lesser of two evils— protecting the secrecy of jury deliberations at the expense of possibly allowing irresponsible juror activity. Achieving a more perfect system for monitoring the conduct of jurors in the intense environment of a jury deliberation room entails an unacceptable breach of the secrecy that is essential to the work of juries in the American system of justice. To open the door to the deliberation room any more widely and provide opportunities for broad-ranging judicial inquisitions into the thought processes of jurors would, in our view, destroy the jury system itself.
A similar choice — to protect deliberative secrecy at the risk of leaving some juror misconduct beyond the court’s power to remedy — underlies the long-standing common law rule that “a juror may not impeach his own verdict,” McDonald v. Pless, 238 U.S. 264, 267-68, 35 S.Ct. 783, 784-85, 59 L.Ed. 1300 (1915), attributed originally to Lord Mansfield’s decision in Vaise v. Delaval, 99 Eng.Rep. 944 (K.B.1785); see 8 John H. WigmoRE, Evidence 2352 (J. McNaughton rev. ed. 1961), and now embodied, with qualifications, in Rule 606(b) of the Federal Rules of Evidence (Rule 606(b)).13 The Supreme Court has expressly recognized the choice that underlies the rule:
The rule is based upon controlling considerations of a public policy which in these cases chooses the lesser of two evils. When the affidavit of a juror, as to the misconduct of himself or the other members of the jury, is made the basis of a motion for a new trial, the court must choose between redressing the injury of the private litigant and inflicting the public injury which would result if jurors were permitted to testify as to what had happened in the jury room.
McDonald, 238 U.S. at 267, 35 S.Ct. at 784. Specifically, if post-verdict juror testimony could be used to impeach a verdict, “the result would be to make what was intended to be a private deliberation, the constant subject of public investigation to the destruction of all frankness and freedom of discussion and conference.” Id. at 267-68, 35 S.Ct. at 784-85; see also Attridge v. Cencorp Div. of Dover Technologies Int'l, Inc., 836 F.2d 113, 116 (2d Cir.1987) (listing as one of the rule’s three purposes, “to promote free and uninhibited discourse during deliberations”); S.Rep. No. 93-1277 (1974), reprinted in 1974 U.S.C.C.A.N. 7051, 7060 (“[C]ommon fairness requires that absolute privacy be preserved for jurors to engage in the full and free debate necessary to the attainment of just verdicts. Jurors will not be able to function effectively if their deliberations are to be scrutinized in post-trial litigation.”). The standard that we adopt here with respect to inquiries of deliberating jurors likewise recognizes the basic necessity of protecting the secrecy of the jury room, even when this protection places some instances of willful disregard of the applicable law beyond the reach of the court’s corrective powers.
We recognize that the standard that we endorse today had not been articulated as part of the law of this Circuit when the district court in the instant case ordered Juror No. 5’s dismissal. But in view of the risks, described above, inherent in the application of a lower standard, we cannot uphold the dismissal. Juror No. 5 said nothing to the court to indicate that he was unwilling to follow the court’s instructions. On the contrary, he assured the court that his vote was based on his view of the . evidence: “I want substantive evidence against them ... and I *624want to know that it’s clear in my mind beyond a reasonable doubt.” Nor was this statement without corroboration from fellow jurors. Several of the jurors indicated in their interviews with the court that Juror No. 5 justified his position during deliberations in terms of the evidence — that he found the Government’s evidence, including its witness testimony, insufficient or unreliable. On this record, we cannot say that it is beyond doubt that Juror No. 5’s position during deliberations was the result of his defiant unwillingness to apply the law, as opposed to his reservations about the sufficiency of the Government’s case against the defendants.
In reaching the conclusion that the district court erred in dismissing Juror No. 5, we note that Chief Judge MeAvoy proceeded with painstaking care and caution throughout this trial, with a clear respect for the rights of the defendants to a fair trial. It bears recalling that the judge on two earlier occasions leaned over backward to retain Juror No. 5 on this jury — indeed, Juror No. 5 was not removed at jury selection by a peremptory challenge of the Government only because the judge, apparently hesitant to permit a challenge to the single remaining black member of the venire, erred in the defendants’ favor in applying the teachings of Batson,14 We are required to vacate these judgments because the court dismissed Juror No. 5 largely on the ground that the juror was acting in purposeful disregard of the court’s instructions on the law, when the record evidence raises a possibility that the juror was simply unpersuaded by the Government’s case against the defendants.
We need not reach the question of whether the court’s inquiries were themselves sufficiently intrusive to constitute reversible error. Moreover, we do not decide here whether it would have been within the district court’s discretion to dismiss Juror No. 5 for his distracting behavior, pursuant to Rule 24(c),15 after the first round of in camera interviews but before the jury began its deliberations. Finally, we do not suggest, much less hold, that a juror’s disruptive behavior— his reported “hollering,” threatening to strike a fellow juror, or feigned vomiting— could not serve as grounds for dismissal, but we conclude that, in the circumstances presented here, the juror was removed largely for his allegedly nullifying behavior.
D. Removal to Break Jury Deadlock
Several of the defendants urge that there was a separate source of error in the district court’s decision to dismiss Juror No. 5. They argue, specifically, that the court removed the juror as a means of achieving a unanimous verdict. See Brief for Defendanb-Appellant Grady Thomas at 17; Brief for Defendant-Appellant Loray Thomas at 17-18, 21; Brief for Defendant-Appellant Ramse Thomas at 23-24; Brief for Defendant-Appellant Tracey Thomas at 9. Having concluded that the district court erred by dismissing Juror No. 5 without the requisite evidentiary basis, we need not decide whether the record reveals an intention on the part of the court to remove Juror No. 5 as a means of achieving jury unanimity. We address this issue briefly, however, merely to note that, as the law of our Circuit makes clear, a district court may under no circumstances remove a juror in an effort to break a deadlock. See United States v. Hernandez, 862 F.2d 17, 23 (2d Cir.1988), cert, denied sub nom. Quinones v. United States, 489 U.S. 1032, 109 S.Ct. 1170, 103 L.Ed.2d 228 (1989) (finding error in decision to dismiss juror where, inter alia, [t]he record ... seem[ed] to reflect that the cause of the removal was as much to avoid a mistrial because of a hung jury as to excuse an incompetent juror); see also Wilson, 894 F.2d at 1250 (noting that record lacked even “the slightest basis to believe that [the removed] juror was a holdout juror or that the jury had reached any sort of impasse in its deliberations”); Stratton, 779 F.2d at 832 (finding that “the record [did] not present even the slightest basis to believe that [the dismissed juror] was excused on a pretext to remove an obstacle to reaching a unanimous verdict”). In fact, we subject a Rule 23(b) dismissal to “meticulous” scrutiny in any case *625where the removed juror was known to be the sole holdout for acquittal. Hernandez, 862 F.2d at 23.
Finally, in order to clarify the nature of the new trial permitted by our decision,16 we comment briefly on the defendants’ remaining claims, including Ramse Thomas’s arguments that he was prejudiced by being forced to wear leg irons during trial and that the court erred in failing to suppress certain telephone communications and Jason Thomas’s claim that the Government presented the grand jury with perjured testimony. We have reviewed all of the defendants’ claims and have found them to be without merit.
III.
For the reasons stated above, we conclude that:
(1) The district court properly determined that a juror’s purposeful disregard of the law as set forth in the court’s instruction may constitute just cause for that juror’s removal under Rule 23(b).
(2) A court must not, however, remove a juror for an alleged refusal to follow the law as instructed unless the record leaves no doubt that the juror was in fact engaged in deliberate misconduct — that he was not simply unpersuaded by the Government’s case against the defendants.
(3) The court in the instant case thus erred by dismissing Juror No. 5, and permitting the jury of eleven to continue its deliberations, based largely on Juror No. 5’s alleged refusal to follow the court’s instructions on the law, where the record evidence raises the possibility that the juror was attempting to follow the law as instructed, but that he simply remained unpersuaded of the defendants’ guilt.
Accordingly, we vacate the judgments of the district court and remand for a new trial.
. Fed.R.Crim.P. 23(b) provides:
Jury of Less Than Twelve. Juries shall be of 12 but at any time before verdict the parties may stipulate in writing with the approval of the court that the jury shall consist of any number less than 12 or that a valid verdict may be returned by a jury of less than 12 should the court find it necessary to excuse one or more jurors for any just cause after trial commences. Even absent such stipulation, if the court finds it necessary to excuse a juror for just cause after the jury has retired to consider its verdict, in the discretion of the court a valid verdict may be returned by the remaining 11 jurors. (Emphasis supplied.)
. The court subsequently declared a mistrial with respect to Robert Gibson and granted Shawne Thomas’s Rule 29 motion for a judgment of acquittal.
. Terrence Thomas and Carrie Thomas have not appealed their convictions.
. Under Batson, the Government’s burden is to "come forward with a neutral explanation for challenging black jurors.” 476 U.S. at 97, 106 S.Ct. at 1723. The court’s finding that the Government's peremptory challenge was not motivated by the race of the challenged juror should have been sufficient to sustain the Government's request to exercise its peremptory challenge of the juror. Juror No. 5 thus became a member of this jury as a result of the district court's erroneous decision in favor of the defendants.
. Fed.R.Civ.P. 24(c) provides, in pertinent part, that "[a]ltemate jurors ... shall replace jurors who, prior to the time the jury retires to consider its verdict, become or are found to be unable or disqualified to perform their duties.”
. We wish to make clear that nothing in this opinion is intended to suggest that jurors who deliberate under a good faith misinterpretation of the law as instructed by the court are subject to dismissal. In this case we address only the applicability of Rule 23(b) where a juror is alleged to be acting in purposeful disregard of the court's instructions.
. Rule 23(b) continues to provide for the elimination of one or more jurors by stipulation. See supra note 1.
. This same language was included in the oath administered to the jurors in the instant case.
. Accordingly, criminal defendants have no right to a jury instruction alerting jurors to this power to act in contravention of their duty. See United States v. Edwards, 101 F.3d 17, 19-20 (2d Cir.1996) (citing cases); see also United States v. Sepulveda, 15 F.3d 1161, 1190 (1st Cir.1993), cert. denied, 512 U.S. 1223, 114 S.Ct. 2714, 129 L.Ed.2d 840 (1994); Dougherty, 473 F.2d at 1136-37. As the Court of Appeals for the Sixth Circuit recognized, to instruct on nullification "would ... undermine[ ] the impartial determination of justice based on law.” United States v. Krzyske, 836 F.2d 1013, 1021 (6th Cir.) (finding no error in court's response to jury inquiry on nullification that included the admonition to the jury: "You would violate your oath and the law if you willfully brought in a verdict contrary to the law given you in this case."), cert. denied, 488 U.S. 832, 109 S.Ct. 89, 102 L.Ed.2d 65 (1988).
. The Federal Judicial Center’s Benchboolt for U.S. District Court Judges includes the following among its list of standard voir dire questions to prospective jurors:
If you are selected to sit on this case, will you be able to render a verdict solely on the evidence presented at the trial and in the context of the law as I will give it to you in my *617instructions, disregarding any other ideas, notions, or beliefs about the law that you may have encountered in reaching your verdict?
Benchbook for U.S. District Court Judges, supra, at 93 (emphasis supplied).
A widely discussed example of the use of voir dire to eliminate jurors likely to nullify the law is the selection of “death qualified” juries in capital cases. Jurors are excluded whose responses at voir dire indicate that their views on capital punishment “would prevent or substantially impair the performance of [their] duties as ... juror[s] in accordance with [their] instructions and [their] oath.” Adams v. Texas, 448 U.S. 38, 45, 100 S.Ct. 2521, 2526, 65 L.Ed.2d 581 (1980); see also Witherspoon v. Illinois, 391 U.S. 510, 513-14, 88 S.Ct. 1770, 1772-73, 20 L.Ed.2d 776 (1968).
. We do not intend to confine the limitation set out in Brown to criminal cases or cases in which the record evidence raises the possibility that the juror's request to be discharged, or the allegations of his fellow jurors, stems from the juror's doubts about the sufficiency of the prosecution's case. The courts must in all cases guard against the removal of a juror — who aims to follow the court's instructions — based on his view on the merits of a case. See United States v. Hernandez, 862 F.2d 17, 23 (2d Cir.1988) ("That a juror may not be removed because he or she disagrees with the other jurors as to the merits of a case requires no citation.”). Accordingly, if the record raises any possibility that the juror's views on the merits of the case, rather than a purposeful intent to disregard the court’s instructions, underlay the request that he be discharged, the juror must not be dismissed.
. As we note below, see infra p. 624, we do not decide whether, in this case, the court's inquiries of the jurors regarding Juror No. 5's behavior were themselves sufficiently intrusive to constitute reversible error. We decide this case, instead, on the question of the adequacy of the evidentiary basis for Juror No. 5’s removal.
. Fed.R.Evid. 606(b) provides, in pertinent part: Upon an inquiry into the validity of a verdict or indictment, a juror may not testify as to any matter or statement occurring during the course of the jury's deliberations or to the effect of anything upon that or any other juror’s mind or emotions as influencing the juror to assent to or dissent from the verdict or indictment or concerning the juror's mental processes in connection therewith, except that a juror may testify on the question whether extraneous prejudicial information was improperly brought to the juiy’s attention or whether any outside influence was improperly brought to bear upon any juror.
. See supra note 4, and accompanying text.
. See supra note 5.
. "The principle that [the Double Jeopardy Clause] does not preclude the Government’s retrying a defendant whose conviction is set aside because of an error in the proceedings leading to conviction is a well-established part of our constitutional jurisprudence." United States v. Tateo, 377 U.S. 463, 465, 84 S.Ct. 1587, 1588-89, 12 L.Ed.2d 448 (1964).