[686]*686OPINION OF THE COURT
BECKER, Circuit Judge.
This appeal by defendants Luis F. Hidalgo and Jesus Cepeda, Jr. raises an important issue concerning the appropriate remedy for juror misconduct discovered mid-trial. On the seventh day of a nine-day trial, the district court learned that jurors, in disregard of the district court’s admonition, had been discussing the case among themselves. The court responded by distributing a two-part questionnaire asking the jurors whether they had discussed the case with other jurors and, if so, whether those discussions had led them to form an opinion as to the guilt or innocence of the defendants. The questionnaire asked only for “yes” or “no” answers. In response, all the jurors admitted that they had prematurely discussed the ease, but also responded that they had not arrived at a conclusion as to guilt or innocence.
The district court denied the defendants’ requests for individualized voir dire designed to ascertain more precisely what had occurred and the extent of any prejudice flowing from the premature deliberations. Concluding that the defendants had suffered no prejudice as a result of the jury’s misconduct, the district court denied their motions for a mistrial. The trial resumed, and the defendants were convicted of conspiracy to distribute cocaine and heroin, 21 U.S.C. §§ 841(a)(1) and 846; conspiracy to possess with intent to distribute cocaine and heroin, 21 U.S.C. §§ 841(a)(2) and 846; and the use of a firearm in relation to the drug trafficking offenses. 18 U.S.C. § 924(c)(1).
We conclude that the district court erred by refusing to conduct a more searching inquiry into the potential prejudice to the defendants from the jury’s misconduct. Ordinarily, a defendant must show that the error was prejudicial in order to obtain a new trial. However, because the two-part questionnaire did not provide any significant information about the nature or extent of the jurors’ discussions, we fail to see how the district court could have made a reasoned determination that the defendants would suffer no prejudice due to the jurors’ premature discussions. Under these circumstances, i.e., where the jury misconduct was discovered mid-trial but there is no way for us to determine whether the defendants were or were not prejudiced, we will vacate the convictions and remand for a new trial, even though the defendants have not established prejudice.
Hidalgo and Cepeda further contend that the government presented insufficient evidence to support the conviction on the firearms charge. In particular, the defendants submit that the individual who brandished the gun, Juan Hernandez, was involved in a different drug distribution conspiracy and that they should not be responsible for his use of the firearm. We address this issue, which was vigorously pressed, to obviate any double jeopardy challenges in connection with the retrial. Concluding that the government provided ample evidence that there was one conspiracy to distribute drugs and that Hidalgo and Cepeda should have reasonably foreseen that the gun would be used to further the conspiracy, we reject the defendants’ contentions.
I. FACTS AND PROCEDURAL HISTORY
The evidence adduced at trial concerned the efforts of two New York City drug suppliers to infiltrate the Pittsburgh market for illegal drugs. The government’s case, developed through two key witnesses, Nancy Ramos and Juan Hernandez, disclosed the following evidence. Niriz Filix, a/k/a “Papo” and Rosa Queroll were competing drug suppliers in New York. Before Hidalgo and Cepeda left for Pittsburgh in an attempt to sell drugs in that market on behalf of Que-roll, they had worked only for Queroll and not for Papo. Ramos, in contrast, worked for both Papo and Queroll while Hernandez had worked exclusively for Papo. Joseph Resko had been a drug distributor and customer of Papos’ in Pittsburgh. Queroll was interested in selling drugs in the Pittsburgh area and thus hired Resko to also serve as her middleman in the Pittsburgh drug market.
Consequently, at a meeting of Ramos, Res-ko, Hernandez, Queroll, Hidalgo and Cepeda, it was agreed that Queroll — with Hidalgo, Cepeda, and Ramos working for her — would [687]*687supply Resko with cocaine and heroin which would be transported by Hidalgo, Cepeda, Ramos and Resko to Pittsburgh by car at the same time that Ramos, Resko and Hernandez would also transport cocaine supplied by Papo. The above-named individuals agreed that the Papo supply and the Queroll supply were to be transported to the Pittsburgh area on the same trip. According to Ramos, Hidalgo and Cepeda were to participate in this trip in order to oversee the drug delivery and ensure that “nobody got ripped off.”
Ramos testified that in April of 1991 in New York, she, Hernandez and Resko obtained one kilogram of cocaine from Papo, which they hid in the door panel of a rented stationwagon. They then picked up Cepeda and Hidalgo who, according to Ramos, provided eight ounces of cocaine and bundles of heroin (supplied by Queroll), which they and Resko stored in the other door of the station wagon. Ramos, Resko, Hernandez, Hidalgo and Cepeda then drove to Pittsburgh to distribute heroin and cocaine in the Pittsburgh area. According to Ramos, the group arrived in Pittsburgh and checked into a single motel room. Ramos testified that Cepeda took all the drugs — both those supplied by Papo and those supplied by Queroll through Hidalgo and Cepeda — and deposited them in a white plastic bag which Cepeda placed in the ceiling of the motel room. Ramos testified that later that night, with Hidalgo present, she telephoned someone she knew as “Drew,” who was actually Andrew Toth, an undercover agent from the Pennsylvania Bureau of Narcotics. Toth expressed interest in buying all of the cocaine and the heroin. Ramos testified that both Hidalgo and Cepe-da assented to this plan to sell all the drugs together to Toth.
According to Ramos, the next morning, while the entire group was present, Resko removed a pistol from his pants and placed it on the bed. Ramos testified that the pistol had been supplied by Papo. Hernandez testified that Resko later placed the gun in the ceiling, where the drags were stored. Hernandez further testified that Resko subsequently placed the gun on a table in the motel room, and that Hernandez picked it up and placed it in his pocket.
That day, Toth met with Resko, Hernandez, and Cepeda, ostensibly to complete the drug transaction. After Resko showed Toth the white plastic bag with the drugs, Toth handed Resko a briefcase meant to contain the purchase money. At that point, Toth and a fellow officer announced that they were police officers and attempted to arrest Res-ko, Ramos and Hernandez. According to Toth, Resko fled while Hernandez brandished the pistol that Hernandez testified belonged to Resko. Having subdued Hernandez, Toth testified that he and his partners then located Cepeda, Resko, Hidalgo and Ramos, all of whom were arrested.
In January, 1992 a grand jury indicted Cepeda, Hidalgo and Resko. The grand jury charged Cepeda and Hidalgo on counts two only. Count one charged that they, along with Resko, had entered into a conspiracy to distribute and to possess with intent to distribute more than five kilograms of cocaine and more than one kilogram of heroin in violation of 21 U.S.C. §§ 841(a)(1), 841(a)(2) and 846. Count nine charged that Hidalgo and Cepeda, along with Resko, had unlawfully carried and used a firearm in conjunction with count one, in violation of 18 U.S.C. § 924(c)(1).
Hidalgo and Cepeda went to trial together.1 Commencing with its preliminary instructions, the district court repeatedly admonished the jury that it was to refrain from discussing the case until the close of the trial, that is until all the evidence was in and the jury was formally given its instructions. Nonetheless, approximately seven days into a nine-day trial, during a break in the presentation of Cepeda’s case, a juror approached a court officer and told him that the members of the jury had been discussing the case during their recesses and while waiting in the jury room. The court officer informed the trial court of this fact, and the court informed counsel. Hidalgo and Cepeda then moved the court to conduct individualized voir dire of the jurors to ascertain what had occurred and the extent of any prejudice caused by the alleged premature delibera[688]*688tions. In the alternative, they moved for a mistrial.
The district court denied both motions. Rather than conduct an individualized voir dire of the jurors, the court summoned the jurors en masse, informed them of the problem,2 and then gave each a written questionnaire which asked:
1. Have you participated in discussing the facts of this case with one or more other jurors during the trial? Yes _ No _
2. If your answer to Question No. 1 is “Yes,” have you formed an opinion about the guilt or non-guilt of either defendant as a result of your discussions with other jurors? Yes_No__
I verify that the answers to these questions are true and correct. I understand that false statements are subject to the penalties of 18 Pa.C.S. § 4904, relating to unsworn falsification to authorities.
Name: _
Juror No.: _
The jury was left alone in the courtroom while answering the questionnaires.
All twelve jurors answered “yes” to the first question and “no” to the second question. Defendants again moved the court for further inquiry or the grant of a mistrial, but their motion was denied. The trial then resumed, and the jury convicted both Hidal-go and Cepeda on counts one and nine.
The district court had jurisdiction under 18 U.S.C. § 3231. We have jurisdiction over the defendants’ timely appeal from the judgment under 28 U.S.C. § 1291. We review the district court’s investigation of the juror misconduct as well as its denial of a mistrial for abuse of discretion. See United States v. Clapps, 732 F.2d 1148 (3d Cir.1984). With respect to the defendants’ contention that there is insufficient evidence to support their convictions on count nine, the firearms charge, we view the evidence in the light most favorable to the government to determine whether there is sufficient evidence to support the jury’s verdict. See United States v. McNeill, 887 F.2d 448, 454 (3d Cir.1989), cert. denied, 493 U.S. 1087, 110 S.Ct. 1152, 107 L.Ed.2d 1055 (1990).
II. THE JURY MISCONDUCT ISSUE
A. Contentions of the Parties
Invoking them Sixth Amendment right to a fair trial before an impartial jury, Hidalgo and Cepeda submit that the district court abused its discretion by neither engaging in further investigation into whether the jurors were prejudiced by their- premature deliberations nor declaring a mistrial. The government argues in response that the district court has broad discretion to determine how to address a situation in which there is evidence of juror misconduct involving improper ■mira-jury influences (in contrast to external influences on the jury), and that the questionnaire was adequate when viewed under this standard. Because the questionnaire produced no evidence that any jurors were prejudiced by their premature discussions and because there is no other evidence in the record of juror prejudice, the government submits that the district court properly denied the defendants’. motions for individualized voir dire or a mistrial.
B. Premature Jury Deliberations as Juror Misconduct — General Principles
It is a generally accepted principle of trial administration that jurors must not engage in discussions of a case before they have heard both the evidence and the court’s legal instructions and have begun formally deliberating as a collective body. Although we have been unable to locate the origins of [689]*689this rule, it is now firmly established in practice in federal court, as well as in most, if not all, state courts. Accordingly, the practice has developed that trial judges admonish juries at the outset of trial not to discuss the case with anyone before the conclusion of the trial. See, e.g., United States v. Wiesner, 789 F.2d 1264, 1269 n. 3 (7th Cir.1986) (“Admonishing the jury [regarding premature deliberations] is a critical and important duty and cannot be over-emphasized.”); United States v. Lemus, 542 F.2d 222, 224 (4th Cir.1976), cert. denied, 430 U.S. 947, 97 S.Ct. 1584, 51 L.Ed.2d 794 (1977); State v. Joyner, 289 S.C. 436, 346 S.E.2d 711 (1986); Commonwealth v. Kerpan, 508 Pa. 418, 498 A.2d 829 (1985); People v. Feldman, 87 Mich.App. 157, 274 N.W.2d 1 (1978); Lillian B. Hardwick & B. Lee Ware, Juror Misconduct § 7.04, at 7-27 (1988); 2 Charles A. Wright, Federal Practice & Procedure § 486, at 718 (2d ed. 1982); 1 Edward J. Devitt & Charles B. Blackmar, Federal Jury Practice & Instructions § 5.07, at 127 (1977);3 89 C.J.S. Trial, § 460 (1955). But see United States v. Viale, 312 F.2d 595, 602 (2d Cir.) (holding that while it is the practice of most judges to admonish the jury regarding premature deliberations, this admonition is not required), cert. denied, 373 U.S. 903, 83 S.Ct. 1291, 10 L.Ed.2d 199 (1963); United States v. Carter, 430 F.2d 1278, 1279-80 (10th Cir.1970) (holding that failure to give the admonition in a two-day trial was not plain error).
There are a number of reasons for this prohibition on premature deliberations in a criminal case. See generally Lillian B. Hardwick & B. Lee Ware, Juror Misconduct § 7.04, at 7-27 (1988). First, since the prosecution presents its evidence first, any premature discussions are likely to occur before the defendant has a chance to present all of his or her evidence, and it is likely that any initial opinions formed by the jurors, which will likely influence other jurors, will be unfavorable to the defendant for this reason. See Commonwealth v. Kerpan, 508 Pa. 418, 498 A.2d 829 (1985). Second, once a juror expresses his or her views in the presence of other jurors, he or she is likely to continue to adhere to that opinion and to pay greater attention to evidence presented that comports with that opinion. Consequently, the mere act of openly expressing his or her views may tend to cause the juror to approach the case with less than a fully open mind and to adhere to the publicly expressed viewpoint. See Winebrenner v. United States, 147 F.2d 322, 328 (8th Cir.1945); State v. Joyner, 289 S.C. 436, 346 S.E.2d 711, 712 (1986).
Third, the jury system is meant to involve decisionmaking as a collective, deliberative process and premature discussions among individual jurors may thwart that goal. See Winebrenner, 147 F.2d at 329; Kerpan, 498 A.2d at 831. Fourth, because the court provides the jury with legal instructions only after all the evidence has been presented, jurors who engage in premature deliberations do so without the benefit of the court’s instructions on the reasonable doubt standard. See Winebrenner, 147 F.2d at 327. Fifth, if premature deliberations occur before the defendant has had an opportunity to present all of his or her evidence (as occurred here) and jurors form premature conclusions about the case, the burden of proof will have been, in effect, shifted from the government to the defendant, who has “the burden of changing by evidence the opinion thus formed.” Id. at 328.
Finally, requiring the jury to refrain from prematurely discussing the case with fellow jurors in a criminal case helps protect a defendant’s Sixth Amendment right to a fair trial as well as his or her due process [690]*690right to place the burden on the government to prove its case beyond a reasonable doubt. See In re Winship, 397 U.S. 358, 364, 90 S.Ct. 1068, 1072, 25 L.Ed.2d 368 (1970).
Despite the importance of the prohibition against all premature discussions, there is a clear doctrinal distinction between evidence of improper minx-jury communications and ertra-jury influences. It is well-established that the latter pose a far more serious threat to the defendant’s right to be tried by an impartial jury. See, e.g., Tanner v. United States, 483 U.S. 107, 117-21, 107 S.Ct. 2739, 2746-48, 97 L.Ed.2d 90 (1987); Smith v. Phillips, 455 U.S. 209, 102 S.Ct. 940, 71 L.Ed.2d 78 (1982); Remmer v. United States, 347 U.S. 227, 228-30, 74 S.Ct. 450, 450-52, 98 L.Ed. 654 (1954); Mattox v. United States, 146 U.S. 140, 149, 13 S.Ct. 50, 53, 36 L.Ed. 917 (1892); Government of V.I. v. Gereau, 523 F.2d 140 (3d Cir.1975), cert. denied, 424 U.S. 917, 96 S.Ct. 1119, 47 L.Ed.2d 323 (1976); Fed.R.Evid. 606(b) (juror may only testify post-verdict on the question whether extraneous influence was brought to bear on jurors). It has long been recognized that when jurors are influenced by the media and other publicity, or when they engage in communications with third parties, these extra-record influences pose a substantial threat to the fairness of the criminal proceeding because the extraneous information completely evades the safeguards of the judicial process. In contrast, when there are premature deliberations among jurors with no allegations of external influence on the jury, the proper process for jury deci-sionmaking has been violated, but there is no reason to doubt that the jury based its ultimate decision only on evidence formally presented at trial.
Moreover, the trial judge has discretion, both in cases involving intra- and extra-jury misconduct, to decide how to deal with a situation in which there is an allegation of jury misconduct, including premature jury deliberations. See Government of V.I. v. Dowling, 814 F.2d 134, 137-38 (3d Cir.1987) (recognizing discretion of district court regarding method for addressing allegations of extraneous influence on the jury). This discretion extends to the determination of whether prejudice has been demonstrated. See United States v. Clapps, 732 F.2d 1148, 1152 (3d Cir.1984); United States v. Pantone, 609 F.2d 675, 679 (3d Cir.1979). Accord United States v. Ortiz-Arrigoitia, 996 F.2d 436 (1st Cir.1993); United States v. Chiantese, 582 F.2d 974, 978, 980 (5th Cir.1978), cert. denied, 441 U.S. 922, 99 S.Ct. 2030, 60 L.Ed.2d 395 (1979). As we have explained, “[t]he trial court is obviously in a better position (than the appellate court) to observe the impact of premature jury discussions of guilt, and to make a considered judgement as to the effectiveness of a cautionary instruction.” Pantone, 609 F.2d at 679. Moreover, the trial court is in a superi- or position to observe the “mood at trial and the predilections of the jury.” Chiantese, 582 F.2d at 980.
C. Analysis
This is a difficult case. On one hand, the case involves only improper mim-jury influence over which the trial court exercises broad discretion. On the other hand, every juror admitted to partaking in premature discussions. In addition, because the district court failed to engage in any investigation beyond the cursory questionnaire, there is no evidence in the record one way or the other regarding prejudice to the defendants. We simply have no way to know the nature of the jurors’ discussions and whether these discussions in fact resulted in prejudice to the defendants. We appreciate the discretion that is generally afforded the district court in these situations and its superior ability to assess the jury’s demeanor. However, the absence of information and the consequent inability of the district court meaningfully to assess the nature and extent of the jurors’ premature discussions in order to ascertain whether there has been any prejudice to the defendants creates a highly problematic situation.
Simply put, the questionnaire raised more questions than it answered. Based on the results of the questionnaire, we know that every juror engaged in premature discussions. Yet there is no way to know the nature of those discussions — whether they involved merely brief and inconsequential [691]*691conversations about minor matters or whether they involved full-blown discussions of the defendants’ guilt or innocence. The government asserts that the jurors were only discussing the facts of the ease and not the guilt or innocence of the defendants. Yet the very crux of the problem here is that neither we nor the district court know anything about the nature of the jurors’ discussions.
Similarly, there is no way to know the extent of the discussions — whether they had been occurring throughout the trial, or whether they were only a few, brief conversations. This lack of information concerns us for two reasons. First, as we have noted, the district court’s ability to evaluate the situation was necessarily diminished by the dearth of information about the jury’s misconduct. Even taking into consideration the discretion enjoyed by the district court as well as the court’s general observance of the jury, we do not see how the district court could have had enough information to make a reasoned determination that the defendants would suffer no prejudice due to the jury misconduct or to fashion an appropriate cautionary instruction. Second, our own ability to review the district court’s determination that there was no prejudice to the defendants is hampered by this absence of information in the record.
In addition, the manner in which the questionnaire was administered raises further questions about its reliability. The jurors were together when filling out the questionnaires and no court personnel was present at that time. Consequently, there was a potential for collaboration among the jurors in filling out the questionnaires, which the district judge did not instruct against. Thus, while individual jurors may have felt that their views as to the guilt/non-guilt of the defendants were influenced by the premature discussions, they may have been swayed against noting this fact on the questionnaires due to a collective agreement to answer “no” to the questionnaire’s second question. We have no assurance that such collaboration did not occur.
We recognize that the district court did not abuse its discretion by its initial decision to resort to the questionnaire. It was clearly within its broad discretion to elicit information through this method, thereby learning that each juror had engaged in premature discussions. However, given the discovery that the jurors had all engaged in premature discussions of the case, we conclude that this method was inadequate to enable the court to fulfill its responsibility of providing an appropriate cautionary instruction and of determining whether prejudice resulted from the jury misconduct.
We conclude that the district court erred by declining to engage in further inquiry— such as individualized voir dire — upon which it could have determined whether the jurors had maintained open minds. By failing to do so, the district court allowed the jurors, in effect, to be the judge of whether or not they had been influenced by their premature discussions. In this way, the district court effectively ceded to the jury its responsibility for determining whether or not the defendants would be prejudiced by the jurors’ misconduct. Our holding comports with the approach of the First Circuit, which has specified that when jury misconduct (including improper intra-jury influences) has been alleged, the district court should: ascertain whether the misconduct actually occurred; if it did, determine whether it was prejudicial; and if there are no grounds for a new trial, specify the reasons it decided that misconduct did not occur, or occurred but was nonprejudicial. See United States v. Richman, 600 F.2d 286, 295 (1st Cir.1979).
We also find support for our conclusion that the district court erred in our decision in United States v. Dowling, 814 F.2d 134 (3d Cir.1987). Although in that case the district court faced allegations of improper extra-jury influence, a far more serious charge, our reasoning in Dowling is relevant here. In Dowling, a juror in a trial for bank robbery informed the district court that the jury had been exposed to extra-record information about the facts of the case and about the defendant’s past criminal record, which included a prior conviction for a bank robbery. See 814 F.2d at 135. The court proceeded to question the jury in banc, asking each juror to raise his or her juror’s card if he or she had been exposed to “any information that [692]*692had rendered [him or her] incapable of giving a fair trial.” Id. at 136 (internal quotations omitted). None of the jurors raised his or her hand to signify his or her inability to render a fair verdict. The trial then resumed without further investigation by the district judge. See id. at 137.
We held in Dowling that while the district court had not abused its discretion by its initial resort to in banc questioning given the latitude a district judge enjoys regarding such decisions, it had erred by failing to engage in further inquiry to ascertain whether the jurors remained impartial. See id. at 137-41. More specifically, we decided that the trial court’s exclusive reliance on the in banc questioning procedure resulted in the creation of an inadequate record for evaluation of the potential prejudice to the defendant. See id. at 141. We reasoned that the in banc questioning did not provide the district court with information about what extra-record information the jurors had been exposed to, nor did it allow the court to evaluate the jurors’ demeanor and credibility in making a finding regarding a historical fact: whether or not the jurors retained open minds. See id. at 138-41. Moreover, we found that the in banc procedure “effectively transferred to the jury the responsibility of determining its own impartiality.” Id. at 141.
Had the district court here not effectively delegated to the jurors the duty of deciding their own impartiality, but instead conducted some additional inquiry to ascertain whether the jurors were influenced by their premature deliberations, this most likely would have sufficed to support any determination it made about the absence of prejudice against the defendants. In both United States v. Clapps, 732 F.2d 1148 (3d Cir.1984) and United States v. Pantone, 609 F.2d 675 (3d Cir.1979), the district court, when faced with allegations of premature jury deliberations, responded with individualized voir dire of the jurors. As a result of this voir dire, the district court concluded that the jury remained impartial. In both Clapps, 732 F.2d at 1152 and Pantone, 609 F.2d at 679, we upheld the determination of non-prejudice by the district court out of deference to the district court’s superior ability to assess the situation and to evaluate the jurors’ impartiality because it had questioned them individually. Accord United States v. Armstrong, 909 F.2d 1238 (9th Cir.) (no abuse of discretion where trial judge responded to allegations of premature juror communications by conducting two evidentiary hearings involving extensive questioning of jurors), cert. denied, 498 U.S. 870, 111 S.Ct. 191, 112 L.Ed.2d 153 (1990). Here, because of the lack of individualized questioning of the jurors, we have less reason to trust the district court’s judgment that the defendants were not prejudiced.
This case is distinguishable from our recent decision in United States v. Thornton, 1 F.3d 149 (3d Cir.1993), an appeal from the convictions of several members of the “Junior Black Mafia.” In Thornton, we upheld the district court’s rejection of the defendants’ motion for voir dire of the jurors in the face of allegations that some jurors experienced “ ‘a general feeling of apprehensiveness’ ” for their safety despite the heightened security measures taken for their benefit. At 689. The district court explained that it rejected the defendants’ motion because, in light of the testimony presented, it believed that such apprehension was normal. See id. Moreover, the district court reasoned that by holding a colloquy, it would only exacerbate the problem by giving added attention to the security considerations in the trial. See id. We concluded that this potential exacerbation of the jurors’ apprehensions was a legitimate consideration and that the district court therefore had not abused its discretion by relying on it as the basis for its decision. See id. at 690.
In contrast, the district court in this case had no countervailing consideration — such as exacerbation of the problem' — -upon which to base its decision not to engage in further inquiry into the jury’s misconduct. Although the district court expressed concern that if it engaged in any colloquy with the jurors it might invade their deliberative process, the court could have easily tailored a colloquy to elicit information about the jurors’ impartiality without so intruding. In short, the district court simply assumed that a determina[693]*693tion of the impact of the premature deliberations was entirely incompatible with respect for the privacy of the jury’s decisionmaking when, in fact, it was not.
D. Prejudice
There are a number of reported cases in which other federal courts of appeals have upheld criminal convictions when there was evidence of premature jury communications but no individualized voir dire of the jurors.4 Most of those cases, however, are distinguishable from the case at bar because there was some basis for determining whether the defendants had been prejudiced.5 For example, in Stockton v. Virginia, 852 F.2d 740, 747 (4th Cir.1988), cert. denied, 489 U.S. 1071, 109 S.Ct. 1354, 103 L.Ed.2d 822 (1989), United States v. Wiesner, 789 F.2d 1264 (7th Cir.1986), and United States v. Chiantese, 582 F.2d 974 (5th Cir.1978), cert. denied, 441 U.S. 922, 99 S.Ct. 2030, 60 L.Ed.2d 395 (1979), the substance of the premature communications were known to both the trial courts and the courts of appeals. Accordingly, the courts of appeals had some means to evaluate the district courts’ determinations that the premature discussions resulted in no prejudice to the defendants. See Stockton, 852 F.2d at 747 (jurors had engaged in “small talk about the manner in which the trial was being conducted and the appearance of certain witnesses”); Wiesner, 789 F.2d at 1268-69 (jurors had discussed the meaning of several words used during the trial); Chiantese, 582 F.2d at 978-79 (three jurors made remarks about an attorney’s conduct during a cross-examination). Here, in contrast, neither we nor the trial court have any means to evaluate the possible prejudicial effect of the premature deliberations since we are unaware of the substance or the frequency of the discussions.
There is, however, at least one ease which is inconsistent with our reasoning: United States v. Klee, 494 F.2d 394, 395-96 (9th Cir.), cert. denied, 419 U.S. 835, 95 S.Ct. 62, 42 L.Ed.2d 61 (1974).6 In Klee, the district court repeatedly admonished the jurors not to discuss the case among themselves at recesses. However, the defendant, in support of a motion for a new trial, presented an affidavit of one of the jurors which stated that eleven of the fourteen jurors (including alternates) had discussed the case during recesses and that nine of the jurors expressed premature opinions about Klee’s guilt. The district court denied the motion for a new trial — without any investigation into the allegations of jury misconduct — and the Ninth Circuit affirmed.
At the outset, the Ninth Circuit questioned the value of an admonition prohibiting premature discussions among jurors. See id. at 395-96. The court acknowledged that ordinarily individuals working together communicate about their common goal throughout the project.7 See id. Nonetheless, the court did [694]*694not decide either the propriety or the necessity of the admonition against premature deliberations. See id. at 396. It went on to reason that when internal jury misconduct comes to light, the essential issue is the existence or non-existence of prejudice to the defendant, explaining, “[t]he test is whether or not the misconduct has prejudiced the defendant to the extent that he has not received a fair trial.” Id. The Klee court then held that premature deliberations, without an indication of prejudice, do not warrant a new trial. According to the court, the overarching goal is that the jury, once it begins its formal deliberations, should view the evidence with an open mind and, as long as there is no proof that the jury no longer possesses this requisite mental state, evidence of jury misconduct alone is not grounds for a retrial.8 See id. Because there was no such proof on the record before it, the Ninth Circuit affirmed the district court’s denial of the defendant’s motion for a new trial.
The problem with Klee is that its analysis stops precisely where it should have begun. We agree that prejudice is generally the touchstone of entitlement to a new trial when improper intra-jury influences are at issue. But the Klee court lost sight of the prejudice inquiry out of deference to the superior ability of the trial judge to assess the prejudicial impact of the jury’s misconduct. In the instant case, the superior position of the trial judge was of little or no consequence because the district judge did not utilize it in order to ascertain whether the defendants had been prejudiced due to the jury misconduct. A trial judge does not possess talismanic powers. In the absence of any effort to evaluate the impact the premature deliberations may have had on the jurors, the judge can only guess as to the existence or non-existence of prejudice. In addition, in the absence of any record, the reviewing court is similarly left to speculate whether the district court properly denied the defendant’s motion for a mistrial.
Although ordinarily a defendant must establish prejudice before a new trial will be ordered, in the circumstances here, in which there is unequivocal proof of jury misconduct discovered mid-trial coupled with a failure by the district court to evaluate the nature of the jury misconduct or the existence of prejudice, we conclude that a new trial is warranted. Given the importance of the Sixth Amendment rights at stake and the relative ease with which the district court here could have properly assessed the impact of the jury misconduct, it would be unfair to penalize the defendants for the lack of evidence of prejudice. We are thus willing, in these limited circumstances, to carve out an exception to the rule that a defendant must demonstrate prejudice before a new trial is warranted.
We note in this regard that, as a practical matter, it is far easier for a district court to address allegations of jury misconduct when they come to light mid-trial rather than after the verdict has been entered and the jury discharged. Before the jury has been discharged, a district judge can more readily and more fruitfully explore whether misconduct has in fact occurred and if so, its effect on the jurors. Thus, the balance of practical [695]*695considerations counselling in favor of further investigation into intra-jury misconduct are much greater when the misconduct is alleged mid-trial rather than post-verdict, when the district court’s inquiry will likely be less productive while consuming more time and resources.9 In addition, at that point, the district court is also in a position to tailor a cautionary instruction to correct the ascertained damage.
We note that the case law has recognized a number of situations in which prejudice to a defendant need not be established in order to grant relief. In these situations, courts will simply presume prejudice to the defendant because “[prejudice in these circumstances is so likely that ease-by-case inquiry into prejudice is not worth the cost.” Strickland v. Washington, 466 U.S. 668, 692, 104 S.Ct. 2052, 2067, 80 L.Ed.2d 674 (1984); see, e.g., id. (actual or constructive denial of the assistance of counsel altogether); Penson v. Ohio, 488 U.S. 75, 86-89, 109 S.Ct. 346, 353-54, 102 L.Ed.2d 300 (1988) (defendant’s appointed appellate counsel failed to follow prescribed procedures before withdrawing from representation); Cuyler v. Sullivan, 446 U.S. 335, 349-50, 100 S.Ct. 1708, 1719, 64 L.Ed.2d 333 (1980) (defendant’s attorney had an actual conflict of interest).
In this case, in contrast, we are unwilling to assume the existence of prejudice because we are far less certain that premature deliberations will lead to prejudice in every, or nearly every, instance. We nonetheless believe that we must vacate and remand for retrial. We do not remand for further investigation into prejudice to the defendants because it is over a year since the trial occurred, and it would be extremely difficult for the district court to now reassemble the jury in order to inquire further into whether the jurors were influenced by their premature discussions. Given the passage of time, a juror may no longer recall what his or her state of mind was before the conclusion of the trial. While we consider the need to order a new trial unfortunate, we do not anticipate that these circumstances, to which our holding is limited,10 will recur. We are confident that district courts will be careful to explore more fully the ramifications of such situations should they arise again.
III. SUFFICIENCY OF EVIDENCE
We now turn to the defendants’ contention that there was insufficient evidence to support a conviction for use of a firearm in connection with a drug trafficking offense, 18 U.S.C.A. § 924(c) (West Supp.1993).11 Hidalgo and Cepeda submit that there was insufficient evidence linking them to the gun brandished by Hernandez during the group’s confrontation with the undercover agents that led to their arrest. More specifically, they argue that the government’s evidence establishes the existence of two separate conspiracies and that they were not members of the conspiracy that utilized the gun. Hidalgo and Cepeda argue that the evidence at most indicated that there was a conspiracy to sell the drugs supplied by Papo which consisted [696]*696of Ramos, Resko and Hernandez; and a separate conspiracy to sell the drugs supplied by Queroll, which consisted of Ramos, Resko and the defendants. Hidalgo and Cepeda thus reason that because the gun in question, which was owned by Resko, was used by Hernandez, but not by any member of the Queroll conspiracy, they cannot be convicted for its use.
The defendants’ arguments are without merit. It is well-established that co-conspirators are liable for all reasonably foreseeable substantive offenses committed in the course of and in furtherance of the conspiracy. See Pinkerton v. United States, 328 U.S. 640, 645-48, 66 S.Ct. 1180, 1183-84, 90 L.Ed. 1489 (1946). We have held that a single conspiracy is proven when there is “ ‘evidence of a large general scheme, and of aid given by some conspirators to others in aid of that scheme.’ ” United States v. Reyes, 930 F.2d 310, 313 (3d Cir.1991) (quoting United States v. Kenny, 462 F.2d 1205, 1216 (3d Cir.), cert. denied, 409 U.S. 914, 93 S.Ct. 233, 34 L.Ed.2d 176 (1972)). In addition, we have held that a single conspiracy may be proven even when each defendant did not know all of the details or goals of the conspiracy. See United States v. Theodoropoulos, 866 F.2d 587, 593 (3d Cir.1989). Under these standards, the government offered sufficient proof that there was a single conspiracy to deliver drugs in Pittsburgh and that Hidalgo and Cepeda should have foreseen Hernandez’s use of the gun.
Ramos testified that she, along with Hernandez, Resko and the defendants, traveled to Pittsburgh with the purpose of selling drugs. It is true that the drugs were supplied by two separate suppliers (who were competitors, at least in the New York City drug market). It is also true that the defendants were primarily the agents of Queroll alone, for whom they ordinarily worked. However, Ramos’ testimony established that as the trip progressed, the goals of the group coalesced, and there was thus evidence from which the jury could have reasonably concluded that, by the time they got to Pittsburgh, there was but one conspiracy to sell drugs to Agent Toth.
Ramos testified that the defendants checked into one hotel room with Hernandez, Resko and Ramos. She further testified that Cepeda took all the drugs — those supplied by Queroll and Papo alike — and hid them in the ceiling of the hotel room. Ramos testified that Hidalgo was present when she called Agent Toth and agreed to sell him all the drugs at once. Most importantly, Ramos testified that both Hidalgo and Cepeda assented to this plan, which undermines their claim that they were not part of a conspiracy to sell the Papo-supplied drugs.
In addition, there was sufficient evidence to support the jury’s conclusion that both Cepeda and Hidalgo were aware of the gun. Ramos’s testimony shows that the defendants were in the room when Resko placed his gun on a bed before storing it in the ceding with the drugs. Ramos also testified that Resko explained to the group that he kept the gun for protection with respect to his drug transactions, among other reasons. Based on this testimony, the jury could have reasonably inferred that the defendants should have foreseen that Resko might use the gun or that Resko might authorize another conspirator to use the gun. In short, given the testimony supporting an inference that Hidalgo and Cepeda knew about the gun and the considerable evidence that there was a single conspiracy to sell all the drugs to Agent Toth, there was ample evidence upon which the jury could find Hidalgo and Cepe-da liable for the use of the gun. We therefore reject the defendants’ arguments with respect to the gun count.
IV. CONCLUSION
For the foregoing reasons, we will vacate the judgment of the district court on counts one and nine of the indictment and remand for a new trial on those counts.12