dissenting in part and concurring in part:
Nearly a half century ago, Justice Jackson expressed concern that the history of the law of conspiracy “exemplifies the ‘tendency of a principle to expand itself to the limit of its logic.’” Krulewitch v. United States, 336 U.S. 440, 445, 69 S.Ct. 716, 719, 93 L.Ed. 790 (1949) (Jackson, J., concurring) (quoting B. *878Cardozo, The Nature of the Judicial Process 51). Justice Jackson further warned that:
The unavailing protest of courts against the growing habit to indict for conspiracy in lieu of prosecuting for the substantive offense itself, or in addition thereto, suggests that loose practice as to this offense constitutes a threat to fairness in our administration of justice.
Id. at 445-46, 69 S.Ct. at 719-20. Unfortunately, I believe that today’s majority opinion confirms the fears expressed by Justice Jackson. The majority has turned the law of conspiracy — at least in the context of alleged drug conspiracies — into the law of “mere association.” Burgos’s conviction stands simply because he hung around with the wrong people on a long train ride. Indeed, as of today there seems to be little that does not allow inference of conspiracy. I, therefore, respectfully dissent insofar as the majority opinion affirms Burgos’s conviction. ■
I do believe that based on traditional principles of conspiracy law, there was sufficient evidence presented at Gobem’s trial to show' that he was involved in a conspiracy with Gonzales — though not with Burgos. I, therefore, concur in the judgment, but not the reasoning, of the majority insofar as it sustains Gobern’s conspiracy conviction. I also concur in the majority’s treatment and disposition of the sentencing issues raised by Go-bern.
I.
A.
At Burgos’s trial the following facts were presented to the jury. On January 25, 1993, three men, Alexio B. Gobern, Anthony Gonzales, and Frank K. Burgos, traveled on the same train from New York City to Greensboro, North Carolina. Gobern, Gonzales, and Burgos got off the train at the AmTrak station in Greensboro. Burgos was first off, followed by Gonzales and then Gobern. They did not walk together on the platform. Gobern carried a Christmas-wrapped package and a blue knapsack. Although the package was not elaborately wrapped, it had no loose ends.
As part of routine drug interdiction work, Police Officer Daniel Kaplan stopped and questioned Burgos before he reached the terminal building. Burgos produced his train ticket, which was in his name. Police Officer Berkley Blanks stopped and questioned Gonzales, and Gobern paused to watch the questioning of Gonzales. Officer Blanks and Gonzales then walked to the front of the terminal building, and Gobern followed and watched. At Officer Blanks’ request, Gonzales produced his train ticket, which was in the name of “Anthony Flores.” By this time, Officer Kaplan had finished questioning Bur-gos and had joined Officer Blanks and Gonzales in front of the terminal building.
Gobern then entered the terminal building and walked to the men’s restroom with the wrapped package and the knapsack. Approximately one to two minutes later, Gobern came out of the restroom with the knapsack but without the wrapped package. No one else entered or exited the restroom during this time. After Gobern exited the restroom, Officer Kaplan stopped and questioned him. Officer Kaplan asked for Gobern’s train ticket, which was also in the name of “Anthony Flores.” Both Gobern and Gonzales denied that they knew each other or that they were traveling together. However, both Gobem’s and Gonzales’s tickets were in sequential order, bore the same name, were issued on January 25, and had identical return dates of January 27.
Officers Kaplan and Blanks then searched the restroom after telling Officer Cameron Piner to watch Gobern, Gonzales, and Bur-gos. Inside the restroom the officers saw the Christmas package and a cereal box torn open on the sink, and they found a balled-up New York City newspaper, The Daily News, behind the commode. Inside the newspaper, the officers found aluminum foil; inside the aluminum foil was a ziploe plastic bag containing “crack” cocaine. After finding the cocaine, Officers Kaplan and Blanks went outside and saw Gobern, Burgos, and Gonzales standing near a pay phone. There was a cab with an open door, and it appeared that Gobern, Burgos, and Gonzales were preparing to leave. Gobern was arrested, and Bur-gos and Gonzales were asked to go to the *879police station. At the police station all three men were questioned and fingerprinted. Special Agent Wayne Kowalski of the DEA conducted the questioning and fingerprinting. At the end of the evening Gobern was detained, and Burgos and Gonzales were released.
The Greensboro Police sent the wrapping paper, cereal box, newspaper, and aluminum foil to a North Carolina lab for fingerprint analysis. Agent Kowalski sent the cocaine and plastic bag to a DEA lab in Miami for analysis. Three of Gobern’s prints were found on the wrapping paper, and one of Burgos’s prints was found on the plastic bag. It could not be determined when the cocaine was placed in the bag or when Burgos touched the bag. The aluminum foil was not tested for cocaine residue.
At Burgos’s trial Agent Kowalski testified that when he questioned Burgos on January 25, Burgos said (1) that he (Burgos) knew Gonzales but did not know Gobern, (2) that Burgos “drove down on the train with Gonzales and Gobern from New York,” (3) that “Gobern had the package throughout the trip down,” (4) that Gobern and Gonzales “had dope although Burgos did not see it,” and (5) that Gobern and Gonzales “were going to sell the dope at the A & T University.”
In addition, the Government and the majority claim that Agent Kowalski testified that Burgos said (1) that he knew that crack cocaine was in the wrapped package and (2) that Gobern had the crack cocaine inside the wrapped package from the time he left New York. Govt. Br. at 12; ante at 854-55 & 865-67. While it is evident from the trial transcript that the Assistant United States Attorney repeatedly attempted to elicit such testimony from Agent Kowalski, a reading of the transcript shows that Agent Kowalski never in fact so testified:
Q. Did [Burgos] say whether he talked to Gonzales and Gobern while they were on the train?
A. Yes, he did. He rode with them. He told me that Mr. Gobern was the one who carried the package wrapped as a Christmas package.
Q: All right. Did he make any statements as to how long Mr. Gobern had this package?
A: He said Mr. Gobern had the package throughout the trip down.
Q: All right. And did he tell you anything about what Mr. Burgos or Mr. Gonzales— what Mr. Gonzales or Mr. Gobern told him about what was in the package during the train trip?
A: What Mr. Burgos told me was that he knew that they had the dope although he didn’t see it. And it was his understanding — this is what he told me — it was his understanding they were going to sell the dope at the A & T University.
******
Q: Did you ask him whether he knew that there was crack cocaine in the package? A: Yes. He said he knew they had it, but he actually didn’t see it.
That last question and answer are the closest that Agent Kowalski came to testifying that Burgos said that the crack was inside the package throughout the trip. Yet, Agent Kowalski’s answer of “yes” is only to the question of whether he (Kowalski) asked Burgos whether he (Burgos) knew that there was crack in the package. Agent Kowalski never testified, as the majority mistakenly contends, that Burgos admitted that he (Bur-gos) “knew cocaine base was in the Christmas package that Gobern possessed from the commencement of the trip in New York....” Ante at 867. Nor can it be said, based on Agent Kowalski’s testimony or any other evidence in Burgos’s trial, that the plastic bag was inside the wrapped package “since the inception of the trip from New York[.]” Id. at 866.
In his defense Burgos testified that he was currently attending college in New York City and that he was traveling to North Carolina to visit friends. He planned to spend a day in Greensboro, but his ultimate destination was Laurinburg Institute, a prep school he had attended.
Burgos boarded the train in New York City. While near the front of the ticket line at Penn Station, he was approached by Anthony Gonzales. Burgos knew Gonzales *880from his neighborhood as “Tone,” though he did not know Gonzales’s last name. Because he was running late, Gonzales asked Burgos to get his (Gonzales’s) tickets, and he gave Burgos the ticket money and a piece of paper with the reservation numbers. The tickets were in the name of “Anthony Flores,” and not knowing Gonzales’s last name, Burgos thought nothing of it. He gave Gonzales the tickets, boarded the train, and sat down by himself. After he was seated, Gonzales, Go-bern, and two young women came by. Gonzales introduced Burgos to Gobern. Gonzales, Gobern, and the two women sat behind Burgos on the train.
The trip from New York to Greensboro lasted approximately twelve hours, and Bur-gos had brought sandwiches and snacks to eat. He had packed the food in ziploc bags. He offered some of his food to Gonzales, Gobern, and the women. Eventually, Gonzales and Gobern ate some of the sandwiches. Burgos did not see what they did with the ziploc bags after they ate the food. According to Burgos, the ziploc bags were of the same type as the one found containing the cocaine.
Burgos further testified that the conversation on the train consisted of “general stuff: music, girls.” Burgos also said that he exchanged telephone numbers with Gonzales, though he did not exchange numbers with Gobern.
In contrast to the testimony provided by Agent Kowalski, Burgos testified that there was no conversation about drugs on the train and that he did not see the wrapped package until Gonzales and Gobern were being questioned by the police at the train station. Burgos explained that he called a cab after Officer Blanks told him in a hostile manner to leave the train station. He called the cab for himself, not for Gobern and Gonzales. When the cab arrived, Burgos moved to open the door, but Officer Piner stopped him and told him not to leave. After Burgos stepped back to the platform, Officers Kaplan and Blanks came out of the train station and arrested Gobern.
Robert Lewis, a friend of Burgos’s and a student at North Carolina A & T University, testified that Burgos called him a few days before the train trip. Burgos told Lewis that although he was coming down to visit him in Greensboro, his main purpose was to visit Laurinburg Institute. In addition, Vernon Johnson, a dormitory director at Laurinburg Institute testified that Burgos called him in January 1993 and said that he (Burgos) was coming down to Laurinburg Institute to play basketball, which he had done in the past. Burgos’s final witness was a character witness, Frank McDuffy, the headmaster of Laurinburg Institute. McDuffy testified that he and Burgos had developed a relationship like that between father and son, that Burgos was in his opinion truthful, and that Burgos would visit Laurinburg when he was in North Carolina.
Based on this evidence, the jury convicted Burgos of conspiracy to possess cocaine base with intent to distribute. The jury also convicted Burgos of possession of cocaine base with intent to distribute. He was sentenced to 131 months imprisonment.
B.
At Gobern’s trial many of the same facts were also presented to the jury. For example, the jury was told that Gobern entered the AmTrak terminal building and walked to the men’s restroom with the wrapped package and the knapsack; that Gobern came out of the restroom with the knapsack but without the wrapped package; and that after Gobern left the restroom, Officer Kaplan stopped and questioned him. In addition, there was testimony that when questioned both Gobern and Gonzales denied that they knew each other or that they were traveling together. Also, the two train tickets introduced at Gobern’s trial showed that both his and Gonzales’s tickets were in sequential order, bore the same name, were issued on January 25, and had identical return dates of January 27. And, of course, Gobern’s jury was presented with evidence showing that the crack cocaine was found in the restroom.1
*881The evidence presented at Gobern’s trial was not, however, identical to that presented at Burgos’s trial. At Gobem’s trial Agent Kowalski did not testify concerning statements made by Burgos at the police station. Thus, Gobern’s jury was not told that Burgos said (1) that he (Burgos) knew Gonzales but did not know Gobern, (2) that Burgos rode down on the train with Gonzales and Gobern from New York, (3) that Gobern had the package throughout the trip down, (4) that Gobern and Gonzales had dope although Bur-gos did not see it, and (5) that Gobern and Gonzales were going to sell the dope at the A & T University.
In addition, Gobern did not testify at his trial and neither did Burgos or Gonzales.
The jury convicted Gobern of conspiracy to possess cocaine base with intent to distribute and possession with intent to distribute cocaine base. Gobern was sentenced to 121 months imprisonment.
II.
At the outset, I emphasize that I agree with the majority’s conclusion that once a conspiracy has been established, the Government need only show a “slight connection” between the defendant and the established conspiracy. See ante at 860-61 In addition, while the majority does not expressly state what it believes to constitute a “slight connection,” I assume the majority agrees that to establish a “slight connection” the Government must present substantial evidence showing that the defendant knowingly and wilfully participated in the conspiracy. Although we have not always been completely faithful to this standard, see United States v. Truglio, 731 F.2d 1123 (4th Cir.), cert. denied, 469 U.S. 862, 105 S.Ct. 197, 83 L.Ed.2d 130 (1984), it is the law of this circuit and others. United States v. Chindawongse, 771 F.2d 840, 844 (4th Cir.1985) (“The evidence of the ‘slight connection’ ... must be of a quality which will reasonably support a conclusion that the defendant wilfully participated in the unlawful plan with intent to further some object or purpose of the conspiracy.”) (quoting United States v. Miranda-Uriarte, 649 F.2d 1345, 1349-50 (9th Cir.1981)), cert. denied, 474 U.S. 1085, 106 S.Ct. 859, 88 L.Ed.2d 898 (1986); United States v. Laughman, 618 F.2d 1067, 1075 (4th Cir.) (“Simply proving the existence of a conspiracy ... cannot sustain a verdict against an individual defendant. There must also be a showing of that defendant’s knowledge of the conspiracy’s purpose and some action indicating his participation.”) (citations omitted), cert. denied, 447 U.S. 925, 100 S.Ct. 3018, 65 L.Ed.2d 1117 (1980); see United States v. Brooks, 957 F.2d 1138, 1147 (4th Cir.) (upholding drug conspiracy conviction when there was “ample evidence establishing the active participation of [the defendants] in the conspiracy”) (emphasis added), cert. denied, 505 U.S. 1228, 112 S.Ct. 3051, 120 L.Ed.2d 917 (1992). See also United States v. Campbell, 985 F.2d 341, 345 (7th Cir.1993) (“To establish that participatory link, the Government must offer sufficient evidence to demonstrate that the defendant knew of the conspiracy and that he intended to join and associate himself with its criminal design and purpose.”); United States v. Cloughessy, 572 F.2d 190, 191 (9th Cir.1977) (To establish a slight connection, “[e]vidence has to be produced to show that [the defendant] had knowledge of the conspiracy and acted in furtherance of it. Mere casual association with conspiring people is not enough.”) (citations omitted).
With that said, however, I believe that the majority has confused the concept of what elements make a conspiracy with the principles by which we determine whether sufficient evidence has been presented to establish those elements. Thus, I cannot agree with the majority’s sweeping statement that “our conspiracy jurisprudence” is a collection of “elastic, ad hoc principles.” Ante at 861.
No doubt because of the covert nature of most conspiracies, the particular evidence upon which the Government may rely to establish a conspiracy and a defendant’s connection thereto often differs from case to case. And, of course, the Government may *882rely upon circumstantial evidence to prove a conspiratorial agreement, a defendant’s knowledge of the conspiracy, and a defendant’s participation in it. Id. at 857. Co-conspirators do not often advertise the fact that they are involved in an illegal endeavor, and it is uncommon for them to memorialize their agreement in a written document. See Iannelli v. United States, 420 U.S. 770, 777 n. 10, 95 S.Ct. 1284, 1289 n. 10, 43 L.Ed.2d 616 (1975) (“The agreement need not be shown to be explicit. It can instead be inferred from the facts and circumstances of the ease.”) (citation omitted).2
Likewise, because conspiracies have “an elusive quality,” ante at 858-59, and are often “loosely-knit, haphazard, or ill-conceived,” id., the conspiracy concept has an elastic quality to it. See Krulewitch, 336 U.S. at 445, 69 S.Ct. at 719 (Jackson, J., concurring) (“This case illustrates a present drift in the federal law of conspiracy which warrants some further comment because it is characteristic of the long evolution of that elastic, sprawling and pervasive offense.”).
Yet, the elasticity of the concept does not-in turn soften the principles by which we determine whether substantial evidence has been presented showing that a crime has been committed. In addition, the elasticity of the concept does not reheve the Government of its burden to prove each element of the crime beyond a reasonable doubt. Accordingly, though the law of conspiracy is flexible in its application, it is not boundless. United States v. Bell, 954 F.2d 232, 236 (4th Cir.1992), cert. denied, 510 U.S. 835, 114 S.Ct. 112, 126 L.Ed.2d 77 (1993). As we have rightly recognized in the past, under the black letter law of conspiracy:
[ M]ore than being a bad person with intent to commit a crime is required [before a conspiracy conviction can be sustained]. A conspiracy is not shown until the government has presented evidence of a specific agreement to commit a specific crime, for the same criminal purpose, on the part of all indicted conspirators.
Id. at 237-38 (emphasis in original); see United States v. Giunta, 925 F.2d 758, 764 (4th Cir.1991) (“[C]ircumstantial evidence that proves nothing more than association between two persons, even if one has a fixed intent known to the other to commit an unlawful act, is not sufficient to permit the inference of the requisite agreement between the two to act in concert to commit the act.”) (citations omitted); United States v. Pupo, 841 F.2d 1235, 1238 (4th Cir.) (en banc) (agreeing that “mere knowledge, acquiescence, or approval of a crime is not enough to establish that an individual is part of a conspiracy to distribute drugs. Nor is mere presence at the scene of a distribution of drugs sufficient to prove participation in a conspiracy.”) (citations omitted), cert. denied, 488 U.S. 842, 109 S.Ct. 113, 102 L.Ed.2d 87 (1988). See also Laughman, 618 F.2d at 1074 (“[T]he gist or gravamen of the crime of conspiracy is an agreement to effectuate a criminal act.”) (emphasis in original; citation omitted).
To hold, as the majority does, that “elastic, ad hoc principles” govern the application of the law of conspiracy to the facts of a case is to condone decisionmaking without clear principles — decisionmaking that is particularly ill-suited for determining whether any crime has been committed. Indeed, one need look no further than the cases presented here to be certain that under the majority’s approach any fact that shows that individuals associated together (or were simply acquainted) may be turned into evidence showing the existence of a conspiracy and a defendant’s participation therein. I will mention just a few examples at this point (the rest are discussed in detail infra): as the majority sees it, (1) that Burgos was *883acquainted with Gonzales, (2) that Burgos, Gonzales, and Gobern were from the same neighborhood, and (3) that Burgos and Gonzales exchanged phone numbers are all facts showing that Burgos participated in a conspiracy. See ante at 869-70. While I believe one would be hard-pressed to say that these facts show anything more than mere association, and a very weak association at that, under the majority’s “elastic, ad hoc principles,” these are all facts that establish Bur-gos’s guilt.
Needless to say, analysis such as that represents the kind of “loose practice” that Justice Jackson specifically warned against in Krulewitch, a practice that “constitutes a threat to fairness in our administration of justice.” 336 U.S. at 446, 69 S.Ct. at 720 (Jackson, J., concurring). See also Bell, 954 F.2d at 237 (recognizing that courts should have “a significant reluctance to apply a conspiracy conviction to individuals whose most heinous crime is choosing the wrong friends”); United States v. Falcone, 109 F.2d 579, 581 (2d Cir.) (Hand, J.) (“so many prosecutors seek to sweep within the drag-net of conspiracy all those who have been associated in any degree whatever with the main offenders. That there are opportunities of great oppression in such a doctrine is very plain, and it is only by circumscribing the scope of such all comprehensive indictments that they can be avoided.”), aff'd, 311 U.S. 205, 61 S.Ct. 204, 85 L.Ed. 128 (1940). In short, reliance on “elastic, ad hoc principles” allows for conspiracy convictions to be based on no more than a defendant’s “mere association” with alleged co-conspirators, rather than substantial evidence establishing the existence of a conspiracy and a defendant’s knowing and wilful participation therein.
Moreover, I cannot agree with the majority that our decisions in either United States v. Giunta or United States v. Bell are incorrectly decided. These decisions properly stated and applied basic principles of conspiracy law. Yet, the majority has now chosen to overrule them because, according to the majority, (1) the decision in Giunta “appli[es] ... ‘heightened vigilance’ to reverse a [conspiracy] conviction,” ante at 859, suggesting “a heightened standard for reviewing conspiracy convictions,” id. at 859; (2) the decision in Bell (along with Giunta) “cannot be squared with” the law of conspiracy, id. at 860; and (3) “Bell can be read as increasing the quantitative connection required to tie a defendant to a conspiracy,” id. at 861.
The majority is, however, flatly wrong to perceive the decisions in Giunta and Bell as creating any conflict with the established law of conspiracy. And, in fact, any perceived conflict is simply a manifestation of the majority’s own misunderstanding about the law of conspiracy and our role in reviewing conspiracy convictions.3
*884In Giunta, a case involving an alleged drug conspiracy, we reviewed the entire trial record and found that it was a “shadowy one” and that there was little evidence of conduct, independent of conversations between investigative agents and the alleged conspirators, pointing to the requisite criminal agreement that the law of conspiracy demands. 925 F.2d at 765. We expressed concern that “[i]n this circumstance, the danger of guilt being found on the basis of speculation from mere association between criminally disposed people, related criminal behavior, and like considerations is acute.” Id. We then cited Justice Jackson’s opinion in Krulewitch and Judge Goldberg’s opinion in United States v. Caro, 569 F.2d 411, 418 (5th Cir.1978), saying that:
In the decade since Judge Goldberg wrote in Caro, conspiracy has of course continued a “potent and oft-used weapon in the prosecutorial arsenal,” particularly in connection with the drug trafficking prosecutions that increasingly dominate federal criminal dockets. And Justice Jackson’s concerns about the special risks of unfairness in conspiracy prosecutions in general have simply been magnified in drug conspiracy prosecutions as understandable outrage and frustration have charged the “war on drugs” presently underway as avowed national policy.
Giunta, 925 F.2d at 766.
Contrary to what the majority believes, see ante at 858-59, these “pronouncements” only recognized the obvious. That is, to guard against the danger of convicting a defendant based on speculation rather than fact — a danger especially acute when conspiracy is charged — we must take care to ensure that sufficient evidence is presented to the jury. The majority is now bothered that in Giunta we chose the words “heightened vigilance” to make this clear. But far from suggesting that it is appropriate to apply “a heightened standard for reviewing conspiracy convictions,” id. at 859, we did just the opposite. After stating the facts in the light most favorable to the Government, we went on to state and apply the appropriate standard of review that governs sufficiency of the evidence challenges on appeal. And, of course, that standard demands that we not only draw all reasonable inferences in favor of the Government, but that we also look to the entire record, including both the evidentiary strengths and weaknesses of the Government’s case, to determine whether the convictions could be sustained. As we stated quite clearly and comprehensively:
[ W]e do not of course suggest that the judiciary may under any circumstances properly view conspiracy prosecutions with general disfavor, nor skew the judicial function at the trial or appellate level to reflect that disfavor. Conspiracy is a crime no less than are “substantive” offenses; the Jackson v. Virginia standard applies equally to assessing proof of both. Our assessment of the sufficiency of the evidence to convict of conspiracy here must therefore of course proceed as it would in respect of any substantive offense. There is no special, more stringent standard in conspiracy. What is appropriate, however, is the recognition, carrying forward Justice Jackson’s in Krulewitch, that the very “elasticity” of the conspiracy concept may and frequently does tend to produce more “slippery” facts in proof than do prosecutions for less “elastic” substantive offenses, with a corresponding greater danger that speculation rather than reason may be required to make the necessary leaps of inferences to find guilt. Heightened vigilance to guard against the increased risks of speculation, though not a heightened standard, is warranted in conspiracy prosecutions.
Our assessment of the evidence here reveals exactly the degree of “slipperiness” of factual proof that Justice Jackson feared as a general phenomenon and that Judge Goldberg found present in Caro. Our conclusion — a raw judgment call at odds with that of the district court — is that because *885of its cumulative weakness in various respects, the evidence here could only lead to a finding of guilt by an unacceptable process of raw speculation rather than by a reasoned process of inferring guilt beyond a reasonable doubt.
Because our ultimate assessment is one of cumulative weakness of the evidence in various respects, we analyze it in terms of those specific weaknesses. We precede the analysis proper with the observation that our earlier account of the facts has been set out in considerable detail and in light of the appropriate evidence assessment standard, precisely to serve as a framework for analysis. That is to say, the facts as recited are essentially those developed by the evidence considered in the light most favorable to the government.
Giunta, 925 F.2d at 766 (emphasis added; footnote omitted).4
As for Bell, there is nothing remarkable about requiring the Government to show the existence of “a specific agreement to commit a specific crime” before a conspiracy conviction can be obtained. See 954 F.2d at 237-38 (emphasis in original). Suppose, for instance, law enforcement agents discover that drugs are being stored and sold at a certain location, go to that location, and find that one individual is selling drugs and another individual is using drugs. There is, of course, no conspiracy shown from these facts. Both individuals are guilty of some crime, the seller for possession with intent to distribute and distribution, and the user for simple possession. Neither, however, is guilty of conspiracy because there is no evidence showing that they specifically agreed to commit any crime. Nor, for that matter, is there any evidence of knowing and wilful participation in a conspiracy.5
In addition, our decision in Bell does not “increas[e] the quantitative connection required to tie a defendant to a conspiracy.” See ante at 861. In Bell we held that the defendants’ conspiracy convictions had to be reversed because the Government failed to present sufficient evidence showing that there existed a specific agreement to commit a specific crime. 954 F.2d at 238 (“The evidence is far too general to support a conviction for any concerted action incident to a specific conspiracy.”). Therefore, other than saying that “mere association” is not enough, id. at 237, we had no occasion to discuss what “quantitative connection” the Government must show between an existing conspiracy and a defendant before the defendant’s conspiracy conviction can be obtained.
In sum, a defendant’s connection with an existing conspiracy is not established unless there is substantial evidence showing that the defendant voluntarily took part in some action that affirmatively links him with the *886conspiracy and showing that he knowingly and wilfully joined the conspiracy. E.g. United States v. Chindawongse, 771 F.2d 840, 844 (4th Cir.1985), cert. denied, 474 U.S. 1085, 106 S.Ct. 859, 88 L.Ed.2d 898 (1986). To hold otherwise allows the Government to convict and punish a person based upon that person’s mere association with conspirators. That has never been the law.
I will now apply these basic and long-established conspiracy principles to the facts presented at the separate trials of Burgos and Gobern.
A.
At Burgos’s trial the Government presented substantial evidence to the jury showing that Burgos associated with Gobern and Gonzales. I think it clear, however, that the Government failed to present substantial evidence that Burgos knowingly and wilfully participated in a Gobern/Gonzales drug conspiracy to possess crack cocaine with intent to distribute.
Viewed in the light most favorable to the Government, the evidence at Burgos’s trial established the existence of a conspiracy between Gobern and Gonzales. Most importantly, Agent Kowalski’s testimony about Burgos’s alleged admissions shows that Go-bern and Gonzales traveled together from New York City with the intent to distribute crack cocaine in Greensboro, North .Carolina.6
Agent Kowalski’s testimony further shows (as does Burgos’s) that Burgos associated with Gobern and Gonzales during the trip from New York City. Indeed, Burgos does not deny that he bought two train tickets on Gonzales’ behalf, that he sat near Gobern and Gonzales throughout the trip, or that he talked with Gobern and Gonzales during the trip. What, of course, Burgos does deny is that he participated in a Gobern/Gonzales conspiracy, and Burgos has denied any such participation from the moment he was questioned by Agent Kowalski. Thus, the Government had to prove beyond a reasonable doubt that Burgos wilfully joined and participated in the Gobern/Gonzales conspiracy with the intent to further, promote, and accomplish its criminal purpose.
At trial, however, the Government offered no testimony, admission, or co-conspirator declaration showing that Burgos wilfully joined and participated in the Gobern/Gon-zales conspiracy. The Government offered no evidence that Burgos possessed tools of the drug trade (e.g., firearms or unexplained cash). The Government offered no evidence that drugs were found in Burgos’s possession. The Government offered no evidence that Burgos used an alias. The Government offered no evidence that Burgos was scheduled to travel back to New York on the same train as Gobern and Gonzales. The Government offered no evidence that Burgos made reservations to travel with Gobern and Gonzales. And the Government offered no evidence that Burgos knew Gobern prior to the time of the train trip.
There are thus only two facts offered by the Government to show that Burgos participated in a Gobern/Gonzales conspiracy. First, Burgos’s fingerprint was discovered on the plastic bag that contained the crack cocaine, and second, Burgos, Gobern, and Gon*887zales were traveling “together” on January 25, 1993. Despite the majority’s attempt to weave these two fibers into a cloak of guilt, see ante at 863, these facts (separately or together) do not provide substantial evidence that Burgos knowingly and wilfully participated in a Gobera/Gonzáles conspiracy.
1.
The Government says the fingerprint is probative evidence that Burgos participated in the conspiracy because the jury could rationally infer that the fingerprint was impressed on the baggie when it contained crack cocaine. Although the Government’s fingerprint expert acknowledged he did not know whether Burgos touched the bag when it contained crack,7 the Government argues that the inference nevertheless can be drawn.
The majority agrees with the Government and points out that Gobern did not have sufficient time to transfer the crack cocaine to the plastic bag during his one or two minutes in the train station restroom. Ante at 865-66. The majority also points to the testimony of Agent Kowalski and contends that “a reasonable jury could infer from the facts that Burgos did assist in packaging the cocaine base before boarding the train in New York.” Id. at 865. And, in particular, the majority believes that because Agent Kowalski testified that Burgos had knowledge of the conspiracy and did not see wrapping materials on board the train, a rational juror could infer that Burgos touched the plastic bag while it contained crack and that he assisted in the packaging before the men boarded the train in New York. Id. at 865-66. The majority then rhetorically asks, “How else could Burgos’s fingerprint be found on an item inside a wrapped package that was wrapped since the inception of the trip from New York?” Id. at 866 (emphasis in original).
Obviously, I agree with the majority to the extent that it is implausible to think that Gobern put the cocaine base in the plastic bag during the one or two minutes he spent in the restroom. Moreover, even if I were to disagree with the majority on this point, the jury could certainly infer that one or two minutes is too little time. Where, however, I do disagree with the majority is on its contention that the evidence and reasonable inferences show that Burgos must have assisted in the packaging of the crack cocaine prior to boarding the train in New York. A rational juror would disagree on this too.
It is axiomatic that when a defendant challenges his conviction on sufficiency of the evidence grounds, we must take the facts in the light most favorable to the Government, and we must further draw all reasonable inferences in favor of the Government. That standard of review is not, however, a license to state an inference as a fact and to pile one inference on top of another until the conclusion reached is far removed from the actual facts presented. But that is what the majority has done here.
In particular, I do not dispute that the jury could infer that Burgos touched the bag when it contained cocaine if there was evidence that the cocaine was in the wrapped package throughout the entire trip from New York. There is, however, no such evidence. While Agent Kowalski testified that Burgos said that Gobern possessed the package throughout the trip down from New York, Agent Kowalski never testified that Burgos said that the plastic bag containing the cocaine was inside the wrapped package throughout the entire trip. Gobern may have carried a "wrapped package all the way from New York. He may also have carried the crack all the way from New York. Still there is no evidence that Gobern carried the plastic bag inside the wrapped package “since the inception of the trip from New York[.]” Accordingly, there is no evidence to support the majority’s assertion that Burgos assisted in the packaging of the cocaine. Go-bern was on the train for twelve hours, so he did not have just one or two minutes to put the cocaine in the plastic bag, he had half a day.
*888Indeed, the fact that Burgos testified that he did not see any materials with which Gobern could wrap or rewrap the package aboard the train does not mean that Gobern did not wrap or open and close the package during the half day he spent on the train. Any child that has ever looked at presents under a Christmas tree before Christmas morning knows that packages can be opened and closed without the aid of scissors, wrapping paper, or tape. Here, there are several different reasons why Gobern may have handled and packaged or repackaged the crack cocaine during the train trip from New York.
The majority is also wrong when it says that:
To accept Burgos’s rendition of the testimony, the jury would have had to find that Gobern saved the plastic bag, entered the lavatory, tore open the Christmas package and the cereal box, unwrapped the newspaper and aluminum foil from the cocaine base, placed the cocaine base in the plastic bag without leaving his fingerprint, wrapped the plastic bag in aluminum foil, swaddled the aluminum foil in the newspaper, and secreted the cocaine base behind the commode.
Ante at 867. This is simply not true. Once again, Agent Kowalski never testified that Burgos said that he (Burgos) knew cocaine base was in the Christmas package that Go-bern possessed from the commencement of the trip in New York. Therefore, although a juror would be free to believe Agent Kowal-ski’s testimony and disbelieve Burgos’s testimony in its entirety (which I assume that Burgos’s jurors did), a rational juror would still have no basis to conclude that the plastic baggie with Burgos’s fingerprint on it was placed in the wrapped package prior to the time that the men left New York.
Nonetheless, could a reasonable juror still infer (and believe it true beyond a reasonable doubt) that Burgos must have assisted in the packaging of the cocaine prior to boarding the train in New York? I think not. Because there was no testimony from Agent Kowalski (or anyone else) that the cocaine was in the wrapped package throughout the entire trip from New York, there is no evidence that Gobern did not put the cocaine into an empty plastic bag bearing Burgos’s fingerprint during the twelve hour train ride from New York. Indeed, all a rational juror could know and infer based on the evidence presented here is that the plastic baggie was placed inside the wrapped package at some unknown time before Gobern got off the train.
Furthermore, suppose that I am wrong and that a rational juror could reasonably infer that Burgos did in fact touch or hold Gobern’s crack-filled bag at some point before the men boarded the train (or during the train trip itself).8 How does a juror jump from the fingerprint to a conclusion that Burgos wilfully joined and participated in the Gobern/Gonzales conspiracy with the intent to further and accomplish its criminal purpose? Although a juror could speculate that one who simply touched another’s container of drugs had the intent to join in and associate himself with an endeavor to possess and ultimately distribute those drugs, I do not think this would qualify as a rational inference. “Unquestionably the print raises a suspicion. But a suspicion, even a strong one, is not enough.” Hiet v. United States, 365 F.2d 504, 505 (D.C.Cir.1966). I therefore think it clear that Burgos’s conviction cannot be sustained based on the fingerprint evidence.9
*8892.
The question, of course, remains whether the Government’s other evidence somehow bridges the gap between the fingerprint and Burgos’s alleged intent to participate in the Gobem/Gonzales conspiracy. Viewed in the light most favorable to the Government, the other evidence showed the following: Bur-gos, Gonzales, and Gobern were all from New York; Burgos was acquainted with Gonzales in New York prior to the train ride; Burgos sat near Gobern and Gonzales on the train ride; Burgos and Gonzales exchanged telephone numbers; Burgos was told they had drugs; the three men got off the train one after the other; and the three men were seen getting into a cab together when the police arrested Gobern.10
That being said, not a single witness testified that Burgos had ever met Gobern before the train ride. Also, while Burgos said he was acquainted with Gonzales in New York, no evidence was presented on the scope of their relationship, save for the fact that Bur-gos and Gonzales exchanged telephone numbers while on the train — a fact that shows that Burgos had hardly any relationship with Gonzales, much less a conspiratorial one.
Likewise, the Government presented no witnesses from New York or the train ride. The Government presented no evidence to show that Burgos had planned to travel to Greensboro with Gobern and Gonzales. There was no evidence that Burgos had previously engaged in any drug activity with Gobern and Gonzales or anyone else. And, as stated above, Burgos was not found to be in possession of any drugs, any weapon, any large amount of cash, or any other items that might indicate an involvement with drugs or the other two men.
Indeed, evidence that Burgos traveled on the same train with Gobern and Gonzales, while going far to show that Burgos associated with Gobern and Gonzales, does little to show that Burgos knowingly and wilfully participated in a Gobern/Gonzales drug conspiracy. See Bell, 954 F.2d at 237 (“More than mere association with bad people who are committing crimes is required for a conspiracy conviction.”). See also Giunta, 925 F.2d at 764 (“[Cjircumstantial evidence that proves nothing more than association between two persons, even if one has a fixed intent known to the other to commit an unlawful act, is not sufficient to permit the inference of the requisite agreement between the two to act in concert to commit the act.”) (citations omitted).
Nevertheless, to buttress its conclusion that Burgos participated in a Gobern/Gon-zales conspiracy, the majority attempts to distill a number of discrete facts and inferences from the single fact that Burgos, Go-bern, and Gonzales traveled on the same train. For example, the majority evidently believes that persons traveling from “source cities,” such as New York City, are more likely to be drug conspirators; and, according to the majority, traveling from such a city is affirmative evidence of guilt. Ante at 868-69. While I might agree that traveling from a “source city” is a fact that has bearing on whether law enforcement officials have a “reasonable suspicion” that drugs are being smuggled, I fail to see how our Fourth Amendment jurisprudence bears on the question of whether sufficient evidence was presented showing that Burgos (or anyone else traveling from New York City) knowingly and wilfully participated in a drug conspiracy.
Similarly, that Burgos picked up Gobern’s and Gonzales’s train tickets, that Burgos *890knew Gonzales, that Burgos, Gonzales, and Gobem were from the same neighborhood in New York, that Burgos and Gobern exchanged telephone numbers, and that the three appeared to be leaving the train station in a cab together are all “facts” that the majority believes help prove Burgos’s guilt. Ante at 869-70. Yet, these are facts that when taken together (or separately) are innocent on their face. They provide little, if any, evidence that Burgos participated in a drug conspiracy.
In addition, while Burgos, Gobern, and Gonzales were on the same train from New York City, there is no evidence showing that Burgos scheduled the same return trip with Gobem and Gonzales, that Burgos lied to Agent Kowalski when he said he was not traveling with Gobern and Gonzales, or that Burgos remained seated near Gobem and Gonzales because they were his co-conspirators. See ante at 868-69 & n. 8, 870. Also, while Gobem and Gonzales possessed sequentially numbered train tickets, both in the name of “Anthony Flores,” Burgos’s ticket was in his own name and evidently not in sequential order with the tickets of Gobern and Gonzales. Indeed, even when taken in the light most favorable to the Government, these facts undercut, rather than support, the majority’s conclusion that Burgos participated in a drug conspiracy. See Giunta, 925 F.2d at 768 (considering evidence, presented by the Government, that had the effect of undercutting rather than supporting the Government’s ease when deciding that insufficient evidence was presented on drag conspiracy charge).
Could then the association evidence and the fingerprint evidence, taken together, lead a juror rationally to conclude beyond a reasonable doubt that Burgos wilfully participated in, and intended to further and accomplish the purpose of, the Gobem/Gonzales conspiracy? Again, I think not. If the fact that a defendant merely associates with a drag dealer is insufficient to prove a conspiracy, I doubt the evidence would suddenly be propelled into the realm of sufficiency by the additional fact that the defendant at one time touched the container that stored the dealer’s drags (especially when the Government has not shown that the defendant touched the container while it contained the drags).
The Ninth Circuit’s decision in United States v. Vasquez-Chan, 978 F.2d 546 (9th Cir.1992), is on point. In that case, DEA Agents infiltrated a drag ring and identified several persons who were suspected of participating in drag trafficking. Eventually, the DEA seized over six hundred kilograms of cocaine found in a house in which the defendants were temporarily residing. One defendant, Vasquez, was the housekeeper, and the other defendant, Gaxiola, was a houseguest. Neither defendant had been mentioned by the co-conspirators as having participated in the drag ring. However, when the DEA agents arrived at the house both defendants were present. Id. at 549.
Both Vasquez and Gaxiola, like Burgos, agreed to speak with the law enforcement officials. Vasquez admitted that she had resided in the house for three months. She said that she worked as a caretaker for a man named Peralta and was paid between $300 and $800 every fifteen days; the money arrived by messenger. According to Vasquez, she did not lease or rent the residence, although agents found a utility bill in her name and the name of Peralta. Vasquez also told the officers that the cocaine had been delivered by messengers about three days earlier, although she could not describe them to the police. In addition, Vasquez possessed a false passport. Id.
As for Gaxiola, she told the agents that she had been Vasquez’s roommate in Mexico and that she and her infant child had been visiting Vasquez and staying at the house for a few weeks. The vast majority of the cocaine was found in the bedroom where Gaxiola and her child slept. Twelve of Gaxiola’s fingerprints were found on the containers in which the cocaine was stored. While no fingerprints were found on the plastic bags inside of the containers, one of Gaxiola’s fingerprints was found on the inside surface of the cover for a container housing the cocaine. Id.
Both Vasquez and Gaxiola were indicted and convicted by a jury of conspiracy to possess cocaine with intent to distribute, possession, and aiding and abetting. On appeal *891they claimed that the evidence was insufficient on all counts. The Ninth Circuit reversed the convictions. On the conspiracy charge, the court held:
While the government submitted more than enough evidence that a narcotics conspiracy existed among several defendants other that Gaxiola and Vasquez, the evidence does not establish that the defendants here agreed to or knowingly assisted that conspiracy. Gaxiola’s and Vasquez’s actions are consistent with those of an innocent housekeeper and houseguest who have no involvement in the ongoing narcotics transaction....
Id. at 553 (emphasis in original).
As for the fingerprint evidence against Gaxiola, the court said:
The canisters — some opened, some closed, some empty, some filled with cocaine— were located in her bedroom; it is reasonable to assume that she touched them at some time, including on one occasion the inside lid of a canister, as she passed in and out of the room or made space in the small bedroom so that she and her infant child could have a comfortable place to sleep. The evidence presented at Gaxiola’s trial did not establish any reason to believe that an innocent explanation of that evidence was any less likely than the incriminating explanation advanced by the government .... Even when the fingerprint evidence is combined with the other evidence against Gaxiola, it is legally insufficient to establish in the mind of a reasonable juror, beyond a reasonable doubt, that she possessed the cocaine located in the house in which she was staying.
Id, at 551-52 (emphasis in original).
Needless to say, the evidence presented against Burgos is no more damning (and actually far less so) than the evidence presented against Vasquez and Gaxiola.
3.
There remains the question whether in reviewing the sufficiency of the evidence on appellate review, can we affirm by assuming that the jury converted its apparent disbelief of Burgos’s testimony into positive evidence of guilt. In other words, can we, as the majority has done, see ante at 865-66, 867-69, fill in the factual gaps in the Government’s proof by negative inference grounded on the assumption that the jury disbelieved Burgos and believed Agent Kowalski’s account of what Burgos said to him about Gobern and Gonzales? Specifically, if the majority is correct, Burgos’s participation in the conspiracy can be established by demeanor evidence arising from Burgos’s denial that he told Agent Kowalski that Gobern and Gonzales were involved in drug trafficking.
Although this circuit has not squarely addressed the issue of what inferences are permissible from disbelief, we have (before today) indicated that the Government cannot prove its case by negative inferences based on demeanor evidence. See United States v. Fountain, 993 F.2d 1136, 1139 (4th Cir.1993) (“while [the defendant’s] evidence may be disbelieved, it contained nothing which, through disbelief, could be converted to positive proof of distribution of marijuana by him”). I would take this opportunity to adopt the longstanding view that demeanor evidence (ie., evidence inferred from the assumption that the jury disbelieved the defendant’s testimony) does not allow us on appellate review to fill the evidentiary gaps in the Government’s case. See Dyer v. MacDougall, 201 F.2d 265 (2d Cir.1952) (Hand, J.). See generally Olin Guy Wellborn, III, Demeanor, 76 Cornell L.Rev. 1075, 1101-02 (1991) (“Hundreds of eases say ... [m]ere disbelief of testimony is not proof of facts of an opposite nature or tendency.”).
When, as here, there is a conviction and the jury made credibility determinations based on the defendant’s testimony and conflicting testimony from the Government, the only affirmative evidence of guilt we may assume on appellate review is attributable to the testimony offered by the Government. There is, of course, nothing remarkable about this standard. It is black letter law that we must assume that the jury credited the Government’s evidence, and we must accept such evidence as true for purposes of deciding whether sufficient evidence of guilt was presented at trial.
*892I think it clear, however, that we may not assume more than this. If we did, then appellate review would be rendered meaningless because there is no basis (except speculation) to decide what additional evidence the jury inferred in light of its credibility determinations. Indeed, to hold that the Government can be credited with additional affirmative evidence of guilt based on negative credibility determinations made against the defendant would relieve the Government of the burden of proving its case. It would raise the problem discussed by Judge Learned Hand in Dyer v. MacDougall, 201 F.2d 265 (2d Cir.1952), and allow the Government to prove its case by having us assume that which is at issue: whether “there is substantial evidence, taking the view most favorable to the Government, to support [the jury’s verdict].” Glasser v. United States, 315 U.S. 60, 80, 62 S.Ct. 457, 469, 86 L.Ed. 680 (1942).
In Dyer, an action for libel and slander, the plaintiff proposed to call the defendants to the stand, have them deny uttering the slander, and then rely upon the jury to disbelieve the witnesses, thus proving the plaintiffs case. Judge Hand rejected the argument that such “demeanor” evidence would be sufficient to prove the plaintiffs ease. Judge Hand explained that while a witness’s demeanor “is part of the evidence” and that the jury “may and indeed they should” take it into consideration, to allow one party to prove its case based on negative inferences would render appellate review meaningless:
He, who has seen and heard the “demean- or” evidence, may have been right or wrong in thinking that it gave rational support to a verdict; yet, since that evidence has disappeared, it will be impossible for an appellate court to say which he was. Thus, he would become the final arbiter in all cases where the evidence of witnesses presented in court might be determinative.
201 F.2d at 269. See United States v. Zeigler, 994 F.2d 845, 850 (D.C.Cir.1993) (“There is no principled way of deciding when the government’s proof, less than enough to sustain the conviction, is nevertheless enough to allow adding negative inferences from the defendant’s testimony to fill the gaps.”); United States v. Sliker, 751 F.2d 477, 495 n. 11 (2d Cir.1984) (“Although [demeanor] is a legitimate factor for the jury to consider, this could not remedy a deficiency in the Government’s proof if one existed.”), cert. denied, 470 U.S. 1058, 105 S.Ct. 1772, 84 L.Ed.2d 832 (1985).
Accordingly, Burgos’s jury was free to believe Agent Kowalski’s testimony as to Bur-gos’s alleged admissions, and we must accept that the jury believed to be true everything that Agent Kowalski said. Likewise, to the extent that Burgos denied making the statements to Agent Kowalski, we must accept that the jury disbelieved Burgos. Yet, where the testimony of Agent Kowalski is not in conflict with the testimony of Burgos, such as on the critical question of whether Burgos participated in a Gobern/Gonzales conspiracy, we are still not free to fill in the gaps in the Government’s proof by considering Burgos’s denial that he was a part of that alleged conspiracy as affirmative evidence of guilt.11
*893Finally, in making the sufficiency determination, I have not given any credit to Burgos’s innocent explanations. Nor have I required the Government to disprove all innocent hypotheses. The law is clear that the Government need not exclude every reasonable hypothesis consistent with innocence. Holland v. United States, 348 U.S. 121, 139-40, 75 S.Ct. 127, 137-38, 99 L.Ed. 150 (1954) (rejecting the view that where the Government relies on circumstantial evidence, it must “exclude every reasonable hypothesis other than that of guilt”); see Jackson v. Virginia, 443 U.S. 307, 326, 99 S.Ct. 2781, 2792, 61 L.Ed.2d 560 (1979) (ha-beas case confirming that the Government has no “affirmative duty to rule out every hypothesis except that of guilt beyond a reasonable doubt”). “The question is not whether the evidence forecloses all possibility of doubt in the mind of the court, but merely whether the evidence, construed most favorably for the prosecution, is such that a jury might find the defendant guilty beyond a reasonable doubt.” Crawley v. United States, 268 F.2d 808, 811-12 (4th Cir.1959).
What we have here is a case where the Government urges inference upon inference, not all of which are rational or supported by the record. The Supreme Court advised long ago that “charges of conspiracy are not to be made out by piling inference upon inference,” Direct Sales Co. v. United States, 319 U.S. 703, 711, 63 S.Ct. 1265, 1269, 87 L.Ed. 1674 (1943), and we too have warned that, although a jury may infer a conspiratorial relationship, “[s]uch inferential analysis is not boundless,” Bell, 954 F.2d at 236. Viewed in the light most favorable to the Government, the evidence here was insufficient for a rational juror to conclude to a “near certitude,” Jackson, 443 U.S. at 315, 99 S.Ct. at 2786-87, that Burgos wilfully participated in the Gobern/Gonzales conspiracy. Consequently, the evidence was insufficient as a matter of law to sustain Burgos’s conspiracy conviction.
B.
From what I have said thus far, it would appear evident that Gobern’s conspiracy conviction must be sustained. Indeed, at Bur-gos’s trial, the testimony of Agent Kowalski as to Burgos’s alleged admissions provided sufficient evidence to establish the existence of a Gobern/Gonzales conspiracy. At Go-bern’s trial, however, Agent Kowalski did not testify as to Burgos’s alleged admissions. Therefore, unlike at Burgos’s trial, the Government not only had to present evidence that Gobem knowingly and wilfully participated in a conspiracy to possess crack cocaine with intent to distribute (i.e., evidence of a “slight connection” to an established conspiracy), but the Government also had to present substantial evidence to the jury showing the existence of a conspiratorial agreement to possess crack cocaine with intent to distribute. See, e.g., Bell, 954 F.2d at 237-38; Giunta, 925 F.2d at 764; Laughman, 618 F.2d at 1074.
In its analysis the majority never directly acknowledges the fact that Agent Kowalski did not testify as to Burgos’s alleged admissions at Gobem’s trial. Nor, for that matter, *894does the majority state what evidence presented at Gobern’s trial shows that Gobern had a specific agreement with either Gonzales, Burgos, or both to possess crack cocaine with intent to distribute. Because the evidence presented about Burgos at Gobern’s trial was even less than that presented at Burgos’s own trial, there is, of course, insufficient evidence in the Gobern record upon which to find a conspiratorial agreement between Gobern and Burgos.
I believe, however, that even without Agent Kowalski’s testimony, there is still sufficient evidence upon which to find a conspiratorial agreement between Gobern and Gonzales, and I would, therefore, sustain Go-bern’s conspiracy conviction on that ground. Gobern’s jury was presented with evidence in the form of the sequentially numbered train tickets that showed that Gobern and Gonzales were traveling together under an alias and planned to return to New York City together within two days. The jury was also presented with evidence that Gobern and Gonzales denied knowing each other when questioned by the police, an obvious attempt to cover up their relationship. And, without doubt, the evidence presented to the jury showed that Gobern possessed the crack cocaine. From this evidence I believe that a rational juror could find that there existed a conspiratorial agreement between Gobern and Gonzales to possess and distribute the drugs that Gobern carried.
Of course, none of these facts implicate Burgos in any way. Rather, they undercut any showing of his guilt. That is, there was no evidence presented showing that Burgos was traveling under an alias (the evidence was actually to the contrary), had a return trip scheduled with Gobern and Gonzales, or lied to the police about knowing Gobern and Gonzales. Nonetheless, based on the fingerprint evidence the majority again erroneously concludes that “[t]he fact that Burgos’s fingerprint was on an interior article wrapped in a package held by Gobern from the inception of the train trip could demonstrate to a rational finder of fact that Burgos and Gobern conspired to distribute the cocaine base.” Ante at 872-73 (emphasis in original). However, the majority does not explain what evidence shows that Gobern possessed cocaine in an “interior article” (ie the baggie) of the wrapped package from the inception of the train trip. The reason for this is quite simple. There is no such evidence.12
III.
The evidence fares no better on Count Two of the indictment returned against Burgos, which charged both possession of cocaine base with intent to distribute, in violation of 21 U.S.C. § 841, and aiding and abetting the possession with intent to distribute, in violation of 18 U.S.C. § 2.
As for the possession charge, while the Government need not establish that the defendant actually possessed drugs, the Government must show “constructive possession, which occurs when the defendant ‘exercises, or has power to exercise, dominion and control over the item.’ ” United States v. Sa-*895mad, 754 F.2d 1091, 1096 (4th Cir.1984) (quoting Laughman, 618 F.2d at 1077). Also, our circuit has said that “[a]n individual may possess drugs jointly with another, but mere presence on the premises where drugs are found, or association with one who possesses drugs, is insufficient to establish the possession needed for a conviction under 21 U.S.C. § 841(a).” Samad, 754 F.2d at 1096; accord United States v. Rusher, 966 F.2d 868, 878 (4th Cir.), cert. denied, 506 U.S. 926, 113 S.Ct. 351, 121 L.Ed.2d 266 (1992). Thus, a rational juror could not find Burgos guilty of possession beyond a reasonable doubt merely because he associated with Gonzales and Gobem and knew that they had drugs to distribute. And despite the majority’s conclusion to the contrary, ante at 875, the fingerprint does not rescue the Government’s case.
The Government’s argument is that the jury could infer that Burgos touched the bag when it contained cocaine. But how does the jury get to ownership, dominion, or control on January 25, 1993, from this evidence of prior touching? The Government’s argument suggests a juror could rationally infer that if a defendant at one time touched an object serving as a container for drugs, the defendant has the power to exercise ownership, dominion, or control over those drugs. The Government argues that the inference is rational here even though (1) the container was not found on Burgos or at his premises, (2) there was evidence the drugs belonged to someone else, (3) there was no evidence of any prior drug activity by Burgos, and (4) Burgos was not found in possession of any drugs or weapons or an unusually large amount of cash. I believe, however, that the Government has failed to prove its case. See United States v. Vasquez-Chan, 978 F.2d 546, 550-53 (9th Cir.1992).
A conclusion that Burgos had ownership, dominion, or control over the contents of the container “requires bridging an evidentiary gap with rank speculation,” Goldsmith v. Witkowski, 981 F.2d 697, 703 (4th Cir.1992), cert. denied, 509 U.S. 913, 113 S.Ct. 3020, 125 L.Ed.2d 709 (1993). Consequently, the evidence was not sufficient to show that Burgos had constructive possession of the crack ultimately seized from the train station bathroom. Thus, it was insufficient to sustain Burgos’s conviction for possession with intent to distribute as charged in Count Two of the indictment.13
As for the aiding and abetting charge, the Government’s brief on appeal makes no argument on how or why the evidence was sufficient to support that charge. The only mention of it in the Government’s brief is this unenlightening sentence: “Aiding and abetting was charged in connection with the possession with intent to distribute.” Govt. Br. at 10. The majority does not do much better and summarily concludes that because the evidence was sufficient to sustain Burgos’s conspiracy conviction, it is sufficient to sustain the aiding and abetting conviction. Ante at 875. I disagree.
To establish aiding and abetting, the Government had to prove beyond a reasonable doubt that Burgos became associated with and participated in Goberris possession of “crack” cocaine and that Burgos acted with the specific intent to facilitate that crime and *896the desire to make it succeed. United States v. Poston, 902 F.2d 90, 93 (D.C.Cir.1990); see Flowers v. Tandy Corp., 773 F.2d 585, 590 (4th Cir.1985) (noting that, in criminal context, a defendant may be found guilty of aiding and abetting if he “shared in the principal’s criminal intent”); United States v. Winstead, 708 F.2d 925, 927 (4th Cir.1983).
Here, viewed in the light most favorable to the Government, the evidence was insufficient for a rational juror to conclude beyond a reasonable doubt that Burgos assisted Go-bern’s criminal activity with the intent to facilitate it and the desire to make it succeed. The fingerprint and the associational evidence is simply not that strong.14 Accordingly, the aiding and abetting charge does not salvage Burgos’s conviction on Count Two.
* * * * * *
Again, I respectfully dissent from the majority’s affirmance of Burgos’s conviction.
Judge HALL, Judge MURNAGHAN, Judge ERVIN, and Judge MOTZ join in this opinion.. Although the fingerprint evidence was also introduced at Gobem’s trial, and the Government said that Burgos was involved in the alleged conspiracy, the Government’s main pitch to Go-*881bem's jury was that Gobern and Gonzales conspired together to possess and distribute the crack cocaine.
. Of course, the fact that the agreement need not be explicit (i.e., the agreement may be tacit) does not mean that the agreement need not be specific, as the majority incorrectly suggests. See ante at 860 (mistakenly citing Iannelli as support for the proposition that the conspiratorial agreement need not be specific). And, I am not "misper-ceiv[ing] the difference between conspiracy to commit the crime and the activity that can constitute participation in the conspiracy.” See id. at 860-61 n. 2. Again, the Government may prove its case based on circumstantial evidence, but that evidence (to be sufficient) must be such that a rational juror could believe beyond a reasonable doubt that there existed a specific agreement to commit a specific crime and that the defendant knowingly and wilfully participated in the conspiracy.
. Since Giunta was decided in 1991, we have cited the case in sixteen published decisions, and we have never once before today either criticized or questioned its analysis or statements of law. Of those sixteen published decisions, thirteen involved conspiracy convictions. We upheld convictions in ten of those thirteen decisions, see United States v. Heater, 63 F.3d 311 (4th Cir.1995); United States v. Johnson, 54 F.3d 1150 (4th Cir.), cert. denied, - U.S. -, 116 S.Ct. 266, 133 L.Ed.2d 188 (1995); United States v. Harris, 39 F.3d 1262 (4th Cir.1994); United States v. Kennedy, 32 F.3d 876 (4th Cir.1994), cert. denied, - U.S. -, 115 S.Ct. 939, 130 L.Ed.2d 883 (1995); United States v. Moore, 11, F.3d 475 (4th Cir.1993), cert. denied, - U.S. -, 114 S.Ct. 1864, 128 L.Ed.2d 486 (1994); United States v. Banks, 10 F.3d 1044 (4th Cir.1993), cert. denied, - U.S. -, 114 S.Ct. 1850, 128 L.Ed.2d 475 (1994); United States v. Mills, 995 F.2d 480 (4th Cir.), cert. denied, 510 U.S. 904, 114 S.Ct. 283, 126 L.Ed.2d 233 (1993); United States v. Chambers, 985 F.2d 1263 (4th Cir.), cert. denied, 510 U.S. 834, 114 S.Ct. 107, 126 L.Ed.2d 73 (1993); United States v. Baker, 985 F.2d 1248 (4th Cir.1993), cert. denied, 510 U.S. 1040, 114 S.Ct. 682, 126 L.Ed.2d 650 (1994); United States v. Mabry, 953 F.2d 127 (4th Cir.1991), cert. denied, 504 U.S. 914, 112 S.Ct. 1951, 118 L.Ed.2d 555 (1992); and we reversed convictions in three of those decisions, see United States v. Dozie, 27 F.3d 95 (4th Cir.1994); United States v. Winfield, 997 F.2d 1076 (4th Cir.1993); United States v. Bell, supra.
Since Bell was decided in 1992, we have cited the case in ten published decisions, and like Giunta, we have never once before today either criticized or questioned its analysis or statements of law. Of those ten published decisions, eight involved conspiracy convictions. We upheld convictions in seven of those eight decisions, see United States v. Lamarr, 75 F.3d 964 (4th Cir.1996); United States v. Morsley, 64 F.3d 907 (4th Cir.1995), cert. denied, - U.S. -, 116 S.Ct. 749, 133 L.Ed.2d 697 (1996); United States v. Heater, supra; United States v. Capers, 61 F.3d 1100 (4th Cir.1995); United States v. Johnson, *884supra; United States v. Mills, supra; United States v. Chambers, supra; and we reversed in one, see United States v. Winfield, supra.
If (as the majority suggests) I believe that "conspiracy is a Frankenstein's monster that has grown out of control,” ante at 859, neither Giun-ta or Bell have been used to curb its growth.
. As in Giunta, the cases presented here ultimately come down to "a raw judgment call" about the sufficiency of the Government's evidence against Burgos and Gobem. As I explain infra, I believe that "call" requires that Burgos’s conviction be reversed and that Gobem's be affirmed. Surprisingly, the majority evidently believes that once the facts are presented and the law is properly stated and applied, the ultimate decision of whether sufficient evidence has been presented to the jury is not a judgment call. See ante at 858-59 & 861-62 n. 4. Yet, if deciding whether sufficient evidence has been presented to the jury is not a judgment call, then I am not sure what is.
Moreover, in malting this judgment call, I have not, as the majority believes, "disingenuously attempted] to ascribe innocuous purposes to surreptitious conduct.” Ante at 870. Rather than being disingenuous, my analysis is based on the existing law in this circuit and the evidence presented at the separate trials of Burgos and Gobem. Burgos’s counsel made a similar point at oral argument, explaining that based on the evidence presented a jury could infer that a conspiracy existed between Gobem and Gonzales, and that Burgos had knowledge of the conspiracy, but that a jury could not reasonably infer Burgos’s participation in the Gobem/Gonzales conspiracy. Compare id. (asserting that Burgos’s “counsel conceded at oral argument that the guilty verdict was not irrational”).
. Our decision in United States v. Morsley, 64 F.3d 907 (4th Cir.1995), cert. denied, - U.S. -, 116 S.Ct. 749, 133 L.Ed.2d 697 (1996), is not to the contrary. In that case, there existed an established conspiracy to buy and sell unregistered firearms for both cocaine and cash. Id. at 919. Because there was no question that a conspiratorial agreement existed among the individuals selling the unlicensed firearms, we analyzed the case to determine whether the defendants knowingly and wilfully participated in that conspiracy. Id. Compare ante at 860-61 (asserting that Morsley holds that "there need not be evidence of a specific agreement in order to sustain a conspiracy conviction”).
. Agent Kowalski’s testimony on this point does, however, raise a question that was not explored at Burgos’s trial. If, when questioned by Agent Kowalski on the night of Gobem’s arrest, Burgos did in fact implicate Gobern and Gonzales in a conspiracy to possess and distribute crack cocaine, why then was Gonzales released by the police? Indeed, to this day Gonzales has never been indicted on the crimes for which Burgos and Gobern were charged. Of course, Burgos’s alleged admissions to Agent Kowalski were, among other things, not in furtherance of the alleged conspiracy and, thus, would not likely be admissible at a trial of Gonzales (which may also explain why Agent Kowalski did not testify as to Burgos’s alleged admissions at Gobern’s trial). Nevertheless, at the time Burgos made his alleged admissions to Agent Kowalski, the police had no way of knowing that Burgos would later deny making the statements or that he would later be indicted and, therefore, be unlikely to testify at a trial of Gobern or Gonzales. I want to make clear that my central question here is not, as the majority believes, why Gonzales was never prosecuted, see ante at 866 n. 6, but rather it is why Gonzales was released by the police right after Burgos made his alleged admissions. Also, I raise this point not to suggest that the police handled these cases improperly in any way, but merely to suggest that Agent Kowalski’s testimony at Burgos’s trial does raise some inconsistencies which have never been explained.
. Compare Stoppelli v. United States, 183 F.2d 391, 393, 394 n. 4 (9th Cir.) (defendant's fingerprint was found on envelope containing heroin; expert testified that in his opinion defendant's fingerprint was impressed at a time when the envelope contained a powdery substance), cert. denied, 340 U.S. 864, 71 S.Ct. 88, 95 L.Ed. 631 (1950).
. Again, the Government’s fingerprint expert testified he had no idea how long Burgos's fingerprint was on the bag. See United States v. Townley, 942 F.2d 1324, 1326 (8th Cir.1991) ("There is no evidence when or where the fingerprints [on tape wrapping a bar of cocaine] were made. The police fingerprint expert testified that fingerprints 'can last for a long time’; '[i]t's possible' ‘even as long as a year.’ ").
. My analysis here is consistent with our circuit's fingerprint cases (i.e., sufficiency of the evidence cases where fingerprint evidence was alleged to be determinative) upon which both parties rely in their briefs. I read those cases to say that, if the fingerprint was impressed on a readily movable object, and if the Government failed to show that the print could have been impressed only during the commission of the crime charged, then the Government must point to additional substantive evidence (beyond the fingerprint) to sustain the conviction. See United States v. Anderson, 611 F.2d 504, 508-09 (4th Cir.1979) (reconciling the following cases: United States v. Harris, 530 F.2d 576 (4th Cir.1976); United *889States v. Van Fossen, 460 F.2d 38 (4th Cir.1972); United States v. Bryant, 454 F.2d 248 (4th Cir.1972); United States v. Corso, 439 F.2d 956 (4th Cir.1971)).
. Officer Blanks’s testimony about the cab was inconsistent with Officer Piner's later testimony (Officer Kaplan said nothing about a cab). Blanks said that when he and Kaplan came out of the bathroom with the cocaine, Burgos, Gonzales, and Gobern were getting into a cab together (notwithstanding that Piner was asked to keep an eye on the three men). According to Piner, however, when Kaplan and Blanks came out of the bathroom with the crack, Gonzales and Go-bern were standing on the sidewalk near a pay phone, at which point Gobern was arrested. (Burgos testified that he went to open the door of the cab but stepped back to the curb when Piner told him he could not leave.) Of course, the jury could have believed Officer Blanks and disbelieved Officer Piner and Burgos.
. Agent Kowalski was not on the train with Burgos, Gobem, and Gonzales; thus, Agent Kowalski could not and did not provide any evidence independent of Burgos's alleged admissions as to what transpired. In light of this, the majority's reliance on the Eleventh Circuit's decision in United States v. Hastamorir, 881 F.2d 1551 (11th Cir.1989), is misplaced. See ante at 864-65. In Hastamorir, customs officers observed six suspected drug traffickers conversing at a mall restaurant. Three of the individuals then left the restaurant for the mall parking lot and proceeded to exchange thirty kilograms of cocaine. These individuals were then arrested. Unaware of the arrests, the three other individuals at the restaurant then left, also for the mall parking lot where they too were arrested. Defendant Ledezma (whom the majority likens to Burgos) was in the second group of three. At the time of his arrest and at trial, he flatly denied knowing any of the three individuals who had been previously arrested — a fact that was clearly a lie in light of the customs officer's prior observations. 881 F.2d at 1554. Ledezma's fingerprints were also found on packages wrapping two kilograms of cocaine. Id. at 1555. At trial Ledezma was convicted of conspiracy to possess cocaine with intent to distribute. On appeal the Eleventh Circuit affirmed on the ground that "Ledezma's false testimony that he was not at [the restaurant] entitled the jury to reasonably infer that he was aware of and participated in criminal activity at the mall.” Id. That is, the jury could infer affirmative evidence of guilt based on testimony shown to be perjured by *893direct eyewitness evidence to the contrary, not demeanor evidence and negative credibility determinations made against the defendant.
In the present case, the majority makes great attempts to show that Burgos (like defendant Ledezma in Hastamorir) fabricated parts of his testimony at trial. See ante at 865-66. However, Burgos’s alleged fabrications are based on the inference that the jury credited Agent Kowalski's testimony as to Burgos's alleged admissions, not evidence independent of Agent Kowalski’s testimony that proves that Burgos lied at the time of his questioning or perjured himself on the stand. I agree that there are certainly inconsistencies between Burgos's alleged admissions made to Agent Kowalski and his testimony at trial, but those inconsistencies are based on inferences derived from demeanor evidence and negative credibility determinations made against Burgos. On appellate review, they are insufficient to fill the gaps in the Government's case. Cf. Wright v. West, 505 U.S. 277, 296, 112 S.Ct. 2482, 2492, 120 L.Ed.2d 225 (1992) (cited by the majority, ante at 867-69) (jury may consider perjured testimony as affirmative evidence of guilt); United States v. Bennett, 848 F.2d 1134, 1139 (11th Cir.1988) (cited by the majority, ante at 868-69) (defendants' admissions to the police and testimony on the stand were viewed as affirmative evidence of guilt when such admissions and testimony were shown to be false by direct eyewitness evidence to the contrary — not credibility determinations based on conflicting testimony).
. To cast Burgos as part of the Gobem/Gon-zales conspiracy, the majority also emphasizes that Gobern, Gonzales, and Burgos appeared to separate when disembarking the train, regrouped at the taxi stand, and were traveling from a "source city." Ante at 871-72. As I have explained, however, these acts fall well short of establishing that Burgos knowingly and wilfully participated in a Gobem/Gonzales conspiracy, much less do they establish a specific agreement between Gobern, Gonzales, and Burgos to possess crack cocaine with intent to distribute.
In addition, the majority’s conclusion that the act of watching another person be questioned by the police is "countersurveillance” amounting to evidence of guilt is especially wrongheaded. See ante at 872. Perhaps if one defendant acts as a "lookout” for a drug purchase, we might have a different situation. See United States v. Pazos, 993 F.2d 136, 140 (7th Cir.1993) (cited by the majority, ante at 872) (sufficient evidence established defendant as part of conspiracy based on act of countersurveillance when, among other things, defendant, armed with a loaded weapon, arrived with alleged co-conspirator at location of prearranged drug transaction, made initial contact with Government informant, and remained in parking lot during the deal’s negotiations, visually surveying the area generally and closely scrutinizing automobile in which transaction was negotiated). I believe, however, that standing in plain sight to observe the questioning of a fellow passenger hardly amounts to surveillance, much less countersurveillance. At most this is a case of rubbernecking. And, of course, while Gobern observed the police questioning Gonzales, there was no evidence presented showing that Gobern countersurveilled the questioning of Burgos.
. Cf. United States v. Lucas, 67 F.3d 956 (D.C.Cir.1995) (reversing conviction for possession of narcotics where defendant sublet apartment in which drugs were found, a receipt with the defendant's name was found in the apartment, and the defendant's fingerprints were found on a shoe box containing drugs); United States v. Earl, 27 F.3d 423 (9th Cir.1994) (per curiam) (DEA agents found defendant inside residence that contained drugs, drug paraphernalia and weapons, and an informant testified that defendant stayed there and "conducted crime” there; the court found this evidence to be insufficient to show constructive possession, and the evidence was not made sufficient by the additional fact that defendant's fingerprints were found on a plate (in the residence) that may have contained traces of cocaine).
In United States v. Van Fossen, 460 F.2d 38, 41 (4th Cir.1972), we left undecided whether the defendant “possessed” certain engraving plates and photographic negatives used in counterfeiting despite the fact that his fingerprints were found on the illegal items. In other words, the fact that the defendant at one time touched the items did not necessarily mean he had ownership, dominion, or control over them. Of course, Burgos's fingerprint was not impressed on the illegal item here (the crack) but rather on that item's container, a plastic bag. Cf. United States v. Keeper, 977 F.2d 1238, 1241 (8th Cir.1992) (in dicta court said defendant's fingerprint on a plastic bag containing a large amount of cocaine, by itself, would not be enough to show possession of the cocaine beyond a reasonable doubt).
. I am not saying that because the evidence was insufficient on the conspiracy count, it necessarily was insufficient to sustain the aiding and abetting charge. The elements of the crime of conspiracy differ from those required to show aiding and abetting, and evidence may be insufficient to sustain the former yet sufficient to sustain the latter. See United States v. Arrington, 719 F.2d 701, 705-06 (4th Cir.1983), cert. denied, 465 U.S. 1028, 104 S.Ct. 1289, 79 L.Ed.2d 691 (1984). In this case, however, the evidence was insufficient to show either that Burgos knowingly and wilfully participated in the conspiracy or that he aided and abetted Gobem's possession with intent to distribute.