This appeal comes from a trial of five defendants involving two related, though not overlapping, conspiracies.1 The only issue on appeal arose out of the jury deliberations, so we forego recitation of .the evidence presented at trial. For the reasons that follow, we reverse and remand for a new trial.
I.
On April 11, 2011, the jury sent a note to the trial court indicating that it had reached a verdict on some of the charges against several defendants but were deadlocked on other charges. The following day, the court asked the jury to complete the verdict form for the defendants as to whom it had reached verdicts and promised “further instructions” as to the remaining defendants. The jury complied, announcing guilty verdicts against appellant Marcus Brown on five counts while acquitting him on a sixth.2 The jury also announced verdicts acquitting two other defendants, Joshua Benton and Christian Benton, on all charges.
After the foreman read the verdicts, defense counsel for Brown requested a jury poll. The court informed the jury that “we’re going to have a poll with respect to defendant Marcus Brown as to him individually.” The court then instructed the jurors that if “your verdict agrees with that as announced by your foreperson,” each juror should say “yes.” If a juror disagreed, then he or she should say “no.” The court instructed the jury not to “say anything other than ‘yes’ or ‘no’ and do not say anything unless and until your seat number is called.” The court then asked each juror the following question: “[D]oes your individual verdict agree with that as announced by your foreperson?” Each juror answered “yes” until the court reached the eleventh juror, who answered “no.”
At this point, the court stopped the poll and asked counsel to approach. After denying motions for a mistrial, the court informed the assembled attorneys that it was “going to ask [the jury] to return and continue the deliberations with regard to Mr. Brown and the two remaining defen*970dants.” The court instructed the jury as follows:
Ladies and gentlemen, I’m going to ask you to return and continue deliberations with regard to Mr. McCorkle and Mr. McAllister and with regard to Mr. Brown since it’s not indicated that the jury has reached a unanimous verdict, so I’m going to ask you to return. And I’m going to give this — actually, we’ll give this form back to you and — with regard to Mr. Brown. Since you have not reached a unanimous verdict, continue deliberations. Thank you.
After sending the jury back, the court again rejected mistrial motions by counsel for the defendants who had not yet received adverse verdicts. Both counsel based their motions on concerns that the aborted jury poll would have a coercive effect on continuing deliberations. Although the court rejected the motions, it did agree to the government’s request that the court provide the jury with a copy of the first paragraph of Jury Instruction 2.608, applicable to jury polls after verdict.3 Brown’s counsel asked the court to include the additional bracketed language that followed this initial instruction.4 The court reviewed the language and ruled that “the other language, that seems to me that that would be a language if the jury was deadlocked.” The court worried that “if I give that to them[,] I would be foreclosed in giving any anti-deadlock instruction if I give that to them at this point in time.” The court then declined to “give the bracketed language.” The court agreed, however, to specify that the instruction it did give — Instruction 2.6085 — was given with “regard to Mr. Brown.”
Shortly after the court gave the instruction, counsel for Brown called the trial court’s attention to this court’s opinion in Crowder v. United States,6 where we recommended the language now contained in the bracketed portion of Instruction 2.6037 in situations with a high potential for juror coercion. Counsel also moved for a mistrial. The government responded that Harris v. United States,8 a more recent decision of this court, suggested that the trial court’s more limited instruction, as given, was the appropriate response to the jury poll. The trial court denied the motion for mistrial without addressing either Crowder or Harris.
*971The jury quickly returned to the courtroom. The foreman announced the same verdicts against Brown, and the jury was polled a second time. The court began with the juror in seat number eight.9 This time, all the jurors agreed with the announced verdict. Brown again moved for a mistrial; the motion was denied, and the court instructed the jury to continue its deliberations with regard to the two remaining defendants.10
II.
Before addressing Brown’s appellate claim — ie., the court’s alleged error in failing to instruct with the bracketed language at the end of instruction 2.60311— we pause to discuss the case law in this jurisdiction that has developed around what we have referred to as a “breakdown in a poll”12 — a situation that differs from a jury “deadlock.”13
We begin with Crowder, where we reversed when the trial court had instructed the jury to continue its deliberations after a poll breakdown.14 The twelfth juror polled had replied “not guilty” because of “the lack of evidence.”15 Given the “inescapable element of coercion” inherent in every jury poll,16 we concluded in Crowder that the trial court had erred in merely asking the jury to continue to deliberate pursuant to the first paragraph of Instruction 2.603 (as augmented).17 We cited several factors that influenced our decision, including (1) the fact that “the numerical split of the jury and the identity of the only dissenter have been revealed in open court,” (2) the “degree of assurance with which the single juror dissented,” and (3) the absence of any credible basis for assuming the juror was merely confused.18 Noting the “obvious danger in such a situation,” where “the lone recalcitrant juror” might “conclude that the trial judge is requiring further deliberations in order to eliminate his dissent,” we suggested that, in the future, trial courts faced with a poll breakdown may want to give an instruction with language similar to that contained in *972the second paragraph of the mild, ABA-approved, anti-deadlock charge.19
We returned to this issue in Harris,20 where we confronted another poll breakdown. There, the twelfth juror polled, when asked about agreement with the verdict announced by the foreperson, replied “part of it and not all of it.”21 The court instructed the jury to continue deliberating, and soon thereafter the jury sent a second note stating, “We have reached another verdict. We all agree we cannot reach a unanimous verdict.”22 The court was unsure what this note meant and decided to give an instruction like the one we recommended in Crowder, incorporating the bracketed language of Instruction 2.608 drawn from the ABA Thomas anti-deadlock charge.23 In our review, after surveying this court’s precedent, we first noted the variety of factors that can “establish the existence or degree of inherent coercive potential.”24 And although we found less “coercive potential” in Harris than in Crowder, we noted that the potential for coercion in Harris was still “especially high.”25 We determined, however, that the judge had dispelled this coercive potential with the supplementary instruction recommended by Crowder, and thus we affirmed the conviction. Specifically, we noted that the trial judge “did not give an ‘anti-deadlock’ instruction nor did he single the dissenting juror out in any way.”26 Rather, the judge had given an instruction that reduced the elements of coercion through “statements that (1) deliberations should aim toward agreement, but not at the expense of individual judgment, (2) each juror must decide the case for himself or herself, but only after impartial consideration of others and (3) a juror should not surrender his or her honest conviction merely to return a verdict.” 27
Finally, we have also confronted potentially coercive poll-breakdown situations— unlike those in Crowder and Harris— where the full Crowder instruction (or its equivalent) was not required because the potential for coercion was not strong enough to warrant such cautionary language. First, in Elliott v. United States,28 we analyzed a jury poll where the seventh juror revealed disagreement. The trial court halted the poll, excused the jurors, *973and instructed them “not to deliberate until the court instructed them further.”29 The court instructed the jury -with the same language used by the trial court in Crowder, complete with a reminder that “each juror is free to change his or her vote until the jury is discharged.”30 Because the “positions of the remaining jurors were never revealed,” this Court found only a “minimal” potential for coercion of the announced dissenter.31 Further, the trial court’s actions “carefully alleviat[ed] any potential coercive effect inherent in the situation.”32 As a result, we rejected defendant’s request for additional remedial measures.
Similarly, in Green v. United States,33 the eighth juror polled expressed disagreement with the announced verdict.34 The trial court asked the jurors to continue deliberations and, as in this case, instructed them only with the language of the first paragraph of Instruction 2.603.35 The court declined to give the Crowder language contained in the bracketed portion of that instruction, reasoning that “the situation did not present unusual coercive circumstances different from a typical jury poll breakdown.”36 Moreover, the court “noted that the jury had been reminded of its obligation to heed honest convictions before deliberations had begun.”37 In affirming the trial court’s exercise of discretion, we noted that the “problem of a deadlocked or ‘hungjury ... is not necessarily the same as a jury that, in polling, simply reveals a split.”38 Put more succinctly, a jury split does not necessarily indicate a deadlock. We, therefore, announced a “baseline assumption that at least some, if not the majority, of poll breakdowns do not indicate such a high potential for undue coercion that additional instruction is required.”39 We then approved the trial court’s omission of the Crowder language from the jury instruction after the poll breakdown because (1) the “exact numerical division of the jury” was not known; (2) no one juror was identified “as the sole obstacle to unanimity,” and (3) the trial court “had, in its initial instructions, effectively given the Crowder instruction with its language on remaining faithful to a juror’s honest convictions.”40 For these reasons, we perceived no high potential for coercion and thus concluded that no additional instruction to dispel potential coercion was required.
While distinguishing, in Green, the coercive potential of a deadlock from a poll breakdown, we did not overlook that the source of Crowder’s suggested language for a poll breakdown had “anti-deadlock origins,” albeit “far milder” origins than those in other anti-deadlock charges.41 On *974the facts of Green, therefore, we recognized “another course of action, namely, silence,” when a poll breakdown does not suggest the potential for a coerced verdict.42 At the same time, however, we did not retreat from our rulings in Crowder and Harris.
Nor did we back away from our decisions in Davis v. United States43 and Benlamine v. United States,44 where we reversed convictions because, “after learning through a jury poll of the existence of a minority for acquittal,” the trial court gave the severe “anti-deadlock Winters charge.”45 In these latter two cases, “the Winters instruction put too much targeted pressure on the revealed minority juror or jurors,” and “the court created undue coercion, thus abusing its discretion.”46
Accordingly, the state of the law we confront now is this: In some cases, there will be no heightened risk of coercion following a jury poll breakdown, so the trial court may either remain silent or issue a simple instruction to continue deliberating, as the court chose to do in this case.47 In other poll breakdown cases, however, there may be a high enough potential for coercion that the trial court must give a more comprehensive instruction.48 The best instruction to give in this situation will be one, justified by Crowder, that closely resembles the Thomas anti-deadlock instruction with its “coercion reducing” elements.49 To the contrary, however, if the trial court uses language from the Winters anti-deadlock instruction for a poll breakdown, it will increase the risk of coercion unacceptably.50 With this background in mind, we turn to Brown’s appellate contentions.
III.
In Harris, this court adopted a two-step process for determining “whether a jury verdict was coerced” following a poll breakdown.51 First, we evaluate the “existence or degree of inherent coercive potential” resulting from the poll breakdown.52 Second, we analyze “how the judge reacted to the situation” to determine whether the judge increased, decreased, or did not affect the “coercive potential.”53 This inquiry is made “from the perspective of the jurors,” and “[a]ny alleged coercion must be evaluated in context and with regard to all the circumstances of the case.”54
A. Existence or Degree of Coercive Potential
As we have noted, “[ejvery jury poll has an inescapable element of coerc*975ion.”55 When dissent is revealed in open court and the jury is simply instructed to continue deliberations, “[t]he most obvious danger” is that the dissenting jurors “will conclude that the trial judge is requiring further deliberations in order to eliminate [their] dissent.”56 We evaluate this risk on a continuum, according to the list of factors described in Harris. In the present case, the eleventh juror, who answered “no,” was either completely isolated or one of two dissenters from the announced verdict (the twelfth juror was not asked and did not answer the poll).57 Either way, that eleventh juror faced a high degree of isolation. While “the exact numerical division” of the jury was not established, only two were possible: 11-1 or 10-2.58 Furthermore, that juror’s identity was “revealed in open court”;59 the judge “[knew] the identity” of at least one dissenting juror, and this juror knew that the judge knew.60 Finally, although the judge instructed the jury to continue deliberations on the charges against the remaining defendants as well as those against Brown, the non-dissenting jurors were not told they were permitted to change their announced votes on Brown.61 Therefore, while the dissenting juror may have felt pressure to change his vote, the other jurors may have felt “bound” to stick with their original votes.62 As a result, the potential for coercion, while not the highest we have encountered, was still “especially high”63 for the dissenting juror and • high as well even for the others.
We recognize that for those who have never participated on a jury, the suggestion that there may be pressure of any kind from a trial judge’s unelaborated instruction — as in this case — to resume deliberations after a jury poll breakdown may seem fanciful. But the reality of participating on a jury is quite different from thinking about it in the abstract. As our case law makes clear, for the juror exposed in open court as a dissenter from an announced unanimous verdict, the pressure to conform is real when the judge requires further deliberations with virtually unanimous jurors of a contrary mind— unless the judge assures the dissenter, *976indeed all jurors, that none should surrender honest conviction and that each is free to change his or her mind. The concern here, therefore, is for all jurors. Without a clearer, more specific instruction, the other jurors, not just the announced dissenter, may be unsure about what options are open to them if the dissenter stands pat.
The government advances a different perspective, citing several factors that it believes substantially lessened the potential for coercion. The government notes, first, that in contrast with Crowder and Harris, “the precise split of the jury was not revealed in open court” because the poll stopped after the eleventh juror had dissented. While this is trae, we do not believe that the presence of a single un-polled juror lessened potential coercion to a legally significant degree. We recognize that our case law has yet to address the specific factual situation we consider here. However, “our evaluation of jury coercion focuses on probabilities, not certainties,”64 and thus our case law provides no basis for the proposition that the presence of a single unpolled juror alone would be enough to eliminate the risk of coercion. Here, there was at least a fifty-fifty chance that only one juror was, in fact, a holdout, unlike the situation in Green where four jurors remained unpolled and thus five numerical combinations were possible.66 Again: while the risk of coercion was not the highest we have assessed, it was still quite high.
The government also argues that the record suggests that the dissenting juror was merely confused by the poll, and thus that his or her response did not reflect an obstacle to unanimity. The government points out that the original poll was conducted shortly after 10:51 a.m. Then, after receiving the trial court’s instructions following the poll breakdown, the jurors submitted a second note — indicating that they had reached a verdict on Brown — only seven minutes later, at 10:58 a.m. The government maintains that this quick response, in the context of a complex, five-defendant, two-conspiracy, multiple charge trial, permitted the trial court, and thus this court, to infer that the dissenting juror was merely confused by the poll. We cannot agree.
The speed of the jury’s response, in itself, offers no new information about the degree of coercion.67 And, in any event, we see no record basis for concluding that the juror was confused. In the first place, the trial court informed the jury that the poll was conducted with regard only to the verdict for Brown; there is no indication that the eleventh juror misunderstood that instruction and, in responding “no,” was referring not to Brown but to one or more of the two verdicts of acquittal that preceded the poll. Furthermore, as in Crow-der, the dissenting juror had just heard most of the other jurors respond with the expected answer (“yes”) to the judge’s question, which makes it less likely that the juror simply did not know which an*977swer to give to affirm the verdict. Finally, because Brown was the only defendant convicted that morning among the three the jury was still considering after the first poll, we are not persuaded that the juror simply lost track of which defendant the court was referencing.68
Finally, it is true, as the government notes, that, during the poll, the dissenting juror did not explain the vote as a view of the merits, rather than a reflection of confusion. To be sure, we have no statement, such as those each dissenter made in Crowder69 and in Harris,70 that provides insight into this juror’s thought process or the basis for the answer. However, this does not help the government. We can only speculate about whether the juror had a minority view about Brown’s guilt or was confused in his or her answer; the fact that he followed the trial court’s instruction and did not reveal the reason behind his answer does not add heft to the government’s assertion that “no” must have meant mere contusion, not considered judgment.
The complexity of a five-defendant, two-conspiracy, multi-charge trial does not change the rules applicable to a jury poll breakdown. Thus, when — as in this case — there is the same “especially high” risk of coercion that this court found in Harris, the evidence of mere, confusion, not coercion, must be reasonably clear, not weakly inferential. Neither argument the government proffers to establish confusion here, nor even both arguments taken together, can be said to meet that test of clarity. For the government to prevail, therefore, the trial court must have acted in a way that dispelled “any coercive potential.” 71
B. Trial Judge’s Response
We noted in Harris that a trial judge’s response to a jury poll breakdown can either “dispel any coercive potential,” be “neutral,” or “compound the problém by ... effectively adding to juror pressure” and thus “independently” creating a “situation of coercive potential.”72 If an atmosphere of coercion exists, the trial judge must reduce it by giving an instruction that reminds the jury that “(1) deliberations should aim toward agreement, but not at the expense of individual judgment, (2) each juror must decide the case for himself or herself, but only after impartial consideration of the views of others[,] and (8) a juror should not surrender his or her honest conviction merely to return, a verdict.” 73
Here, as we have observed, a heightened potential for coercion existed. As a result, the trial court was obligated to give some sort of instruction to reduce that potential. The instruction given by the trial court, however, did not contain any of the coercion-reducing elements the trial judge relied on in Harris, especially the reminder — crucial in the context here— that “a juror should not surrender his or her honest conviction merely to return a *978verdict.”74 Instead, the instruction was even more limited than the instruction we deemed inadequate in Crowder, where the jurors were at least informed that they were permitted to change their votes after the poll.75 As a result, the trial court’s instruction did not diminish the potential for coercion.
We note that the trial court rejected Brown’s request for the additional Thomas language in the bracketed portion of Instruction 2.603 because it believed that such language was inappropriate when the jury was not deadlocked. Our case law has held otherwise. Although we acknowledge the anti-deadlock origins of the Thomas language in Instruction 2.603, we have required courts to instruct juries with that language even in the absence of a true “deadlock” when circumstances indicate such language is necessary to mitigate the potential for coercion after a polling breakdown that reflects a split, not necessarily a deadlock.76 We recognize that a future case may present precisely the situation the trial court worried about here, namely a case where this court is asked to decide whether the anti-deadlock origins of the bracketed language of Instruction 2.603 preclude a subsequent anti-deadlock instruction drawn from those listed in Instruction 2.601(III).77 But this is not that case. Because Brown requested an instruction that should have been given, and the trial court declined to give it without legal justification, we must reverse and remand for a new trial.
So ordered.
. According to the government’s theory of the case, after an altercation at a nightclub, a group of individuals from the Kenilworth neighborhood, including two of the defendants at this trial, committed a series of crimes in the Trinidad neighborhood. In retaliation, a number of individuals from the Trinidad neighborhood committed crimes against residents of Kenilworth. Appellant Marcus Brown was charged with providing a gun to members of the Trinidad conspiracy. He was convicted of one count of Conspiracy, D.C.Code §§ 22-1805a, -401, -2101 (2001), two counts of Assault with Intent to Kill While Armed, D.C.Code §§ 22-401, -4502 (2001), and two counts of Possession of a Firearm During A Crime of Violence, D.C.Code § 22-4504(b) (2001). He was acquitted on one count of Carrying a Pistol Without a License, D.C.Code § 22-4504(a) (2001).
. See supra note 1.
. That instruction reads as follows:
In the polling of the jury, it has become apparent that you may not have reached a unanimous verdict. For this reason, I am asking you to return to the jury room for further consideration of your verdict. If you are unanimous, your foreperson should send me a note indicating that, and I will poll you again. If you are not unanimous, please resume deliberations and see if you can reach a unanimous verdict.
Criminal Jury Instructions for the District of Columbia, No. 2.603 (5th ed. 2012)("Return of the Jury After Polling”).
. That language reads as follows:
It is your duty as jurors to consult with one another and to deliberate with a view to reaching an agreement, if you can do so without violence to individual judgment. Each of you must decide the case for yourself but do so only after an impartial consideration of the evidence with your fellow jurors.
In the course of your deliberations do not hesitate to reexamine your own views and change your opinion if convinced it is erroneous. But do not surrender your honest conviction as to the weight or effect of evidence solely because of the opinion of your fellow jurors or for the mere purpose of returning a verdict.
Id.
. See supra note 3.
. 383 A.2d 336, 342 n. 11 (D.C.1978).
. See supra note 4.
. 622 A.2d 697 (D.C.1993).
. The juror in seat number eight was not the previously dissenting juror but sat in the same row as that dissenter and the previously un-polled juror.
. Later that day, the court issued an anti-deadlock charge. The jury eventually acquitted one of the remaining defendants on all charges and remained deadlocked as to the other. Of the five defendants, therefore, Brown was the only one convicted at this trial.
. See supra note 4.
. Green v. United States, 740 A.2d 21, 26 (D.C.1999).
. See id. at 28 ("The problem of a deadlocked or ‘hung’ jury, however, is not necessarily the same as a jury that, in polling, simply reveals a split.’’). As explained more fully below, the jury instructions, respectively, for a poll breakdown and a deadlock are different overall but, in part, contain identical language for use, when appropriate, in the trial court's discretion. Compare Criminal Jury Instruction No. 2.603, supra notes 3 & 4, with Criminal Jury Instructions for the District of Columbia, No. 2.601(III)(A) (5th ed.2012) ("AntiDeadlock Instructions”).
. See supra note 6. The trial court gave the jury an instruction that mirrored the first portion of Instruction 2.603, see supra note 3, but also informed the jurors that "[ajfter you return to the jury room, any member is free to change his or her vote on any issue submitted to you. Each juror is free to change his or her vote until the jury is discharged. So you may return to the jury room.” 383 A.2d at 341.
. Id.
. Id. at 342.
. See supra notes 3 & 14.
. See Crowder, supra note 6, 383 A.2d at 342-343.
. Id. at 342 n. 11. This charge is also referred to as the Thomas charge. See United States v. Thomas, 146 U.S.App.D.C. 101, 108 n. 45, 449 F.2d 1177, 1185 n. 45 (1971). The language referred to is identical to the bracketed portion of Criminal Jury Instruction 2.603, supra note 4, requested by Brown’s counsel here and is also contained as an alternative in anti-deadlock Criminal Jury Instruction 2.601(III)(A). See supra note 13.
. Supra note 8, 622 A.2d 697.
. Id. at 699.
. Id.
. Id. at 699-700; see supra notes 4 & 19.
. Id. at 705. We wrote that these factors include:
the degree of isolation of a dissenting juror (or jurors), whether the identity of a dissenting juror (or jurors) is revealed in open court as opposed to in a note, whether the exact numerical division of the jury is revealed, whether the judge knows the identity of a dissenting juror (or jurors) and whether the juror is aware of the judge’s knowledge, whether other jurors may feel 'bound' by a vote they have announced, and whether an ‘anti-deadlock’ instruction has been given and, if so, whether this has occurred under circumstances where the potential for coercion is high.
. Id. at 706.
. Id.
. Id. at 707 n. 20.
. 633 A.2d 27, 36 (D.C.1993).
. Id.
. Id. at 31 n. 4. The instruction was otherwise substantially identical to the first paragraph of Criminal Jury Instruction 2.603, than numbered 2.93.
. Id. at 36.
. Id.
. Supra note 12, 740 A.2d at 24.
. Id.
. Id. at 25. Again, at the time, the instruction was numbered 2.93.
. id.
. Id.; see also id. at 23 n. 2 (setting out pre-deliberation instruction).
. Id. at 28.
. Id. at 29.
. Id. at 29, 31. The court also noted that "no anti-deadlock instruction” had been given to the jury. Id. at 30.
. Id. at 30; see supra note 13.
. Id. at 30.
. 669 A.2d 680 (D.C.1995).
. 692 A.2d 1359 (D.C.1997).
. Green, supra note 12, 740 A.2d at 30. The Winters instruction, derived from Winters v. United States, 317 A.2d 530, 534 (D.C.1974)(era banc), is the most coercive of the anti-deadlock instructions approved in this jurisdiction. Green at 27-28. See Criminal Jury Instruction 2.601(111), supra note 13.
. Green, supra note 12, 740 A.2d at 31.
. See Green, supra note 12.
. See Crowder, supra note 6.
. See Harris, supra note 8. This remains true even though, as we noted in Green, a deadlock and a split revealed in polling are not the same thing.
. See Davis, supra note 43, and Benlamine, supra note 44.
. Harris, supra note 8, 622 A.2d at 705.
. Id.
. Id.
. Id. at 701.
. Crowder, supra note 6, 383 A.2d at 342.
. Id. at 342 n. 11.
. Harris, supra note 8, 622 A.2d at 705.
. Id. Our holding in Davis, supra note 43, 669 A.2d at 685, is particularly instructive on the numerical division issue. There, this court discounted the importance of knowing the "exact numerical division among the jurors,” even though a great number of combinations were possible after the third juror revealed disagreement with the verdict. We noted that because "the jury thought it had a unanimous verdict," the “in-court identification of one juror who disagreed with the initial verdict reveals a high probability that a minority of jurors, with a strong possibility of a minority of one, was not in favor of the guilty verdict.” We find this logic especially compelling on these facts, where only a small minority of dissenters was even possible and a minority of one was highly likely.
. Harris, supra note 8, 622 A.2d at 705.
. Id.
. See Elliott, supra, note 28, 633 A.2d at 36 (noting, even when full Crowder instruction was not given, that judge nonetheless instructed jurors that they could change their votes).
. Harris, supra note 8, 622 A.2d at 705. We also note that after the announcement of the jury’s verdict, there were exclamations from the gallery. Although we have not addressed this before, it is reasonable to conclude that some jurors would feel more bound by a vote announced before a courtroom of spectators than one announced before just the courtroom personnel, some attorneys, and a defendant.
. Id. at 706.
. Davis, supra note 43, 669 A.2d at 685.
. Green, supra note 12, 740 A.2d at 29.
. In the past, we have been inclined to find less coercion when the jury took more time than the jury took in this case to return with a verdict after polling breakdown. See id. at 25 (jury took 40 minutes to return with verdict after polling breakdown); Harris, supra note 8, 622 A.2d at 700 (jury deliberated for two more days after receiving Crowder instruction). In any event, coercion can occur either quickly or slowly. See In re Pearson, 262 A.2d 337, 338 (D.C.1970)(finding coercion warranting reversal where, roughly 15 minutes after a polling breakdown, dissenting juror changed her vote from not guilty to guilty in open court, upon "neutral" questioning from the judge).
. Further, we have no record like that in Williams v. United States, 136 U.S.App.D.C. 158, 162-163, 419 F.2d 740, 744-745 (1969), where the juror clearly did not understand the questions posed during the poll.
. See supra note 6, 383 A.2d at 341 ("I said not guilty on Count 2 because of the lack of evidence.”).
. See supra note 8, 622 A.2d at 699 (when asked about agreement with verdict, juror replied, "Part of it and not all of it”).
. Id.
. Id. at 705.
.Id. at 707 n. 20.
. Id..; see Crowder, supra note 6, 383 A.2d at 342 n. 11; supra note 4.
. Crowder, supra note 6, 383 A.2d at 341; see supra note 14.
. Harris, supra note 8, 622 A.2d at 706-707.
. See supra note 13.