Opinion for the Court filed by Circuit Judge TATEL, in which Chief Judge EDWARDS and Circuit Judges WALD, SILBERMAN, BUCKLEY, WILLIAMS, and ROGERS concur.
Concurring opinion filed by Circuit Judge SILBERMAN, in which Circuit Judges BUCKLEY and WILLIAMS join.
Concurring opinion filed by Circuit Judge ROGERS.
Dissenting opinion filed by Circuit Judge RANDOLPH, in which Circuit Judges GINSBURG, SENTELLE, and HENDERSON join.
TATEL, Circuit Judge:
In these consolidated cases, we address the recurring question of what effect a defendant’s unequivocal offer to concede elements of a crime has on the admissibility of prior bad acts evidence under Federal Rule of Evidence 404(b). Where a defendant offers unequivocally to concede elements of a crime — intent and knowledge in these prosecutions under 21 U.S.C. § 841(a)(1) — and agrees to a jury instruction that the Government need not prove those elements, we hold that bad acts evidence offered solely to prove those elements is inadmissible because the defendant’s concession of intent and knowledge deprives the evidence of any value other than what Rule 404(b)’s first sentence unambiguously prohibits: “to prove the character of a person in order to show action in conformity therewith.” Because the district court in No. 93-3059 admitted bad acts evidence notwithstanding defendant’s unequivocal concession of both intent and knowledge, we reverse and remand for a new trial. In No. 92-3133, in addition to admitting the bad acts evidence to show intent to distribute, which defendant had offered to concede, the district court also admitted the evidence to show knowledge of drug dealing, a fact that defendant had not conceded. We nevertheless remand for another Rule 403 balancing because, in weighing the probative value of the bad acts evidence against its prejudicial effect, the district court improperly took account of the conceded fact — intent to distribute.
I.
The Government charged Horace Lee Davis with possession with intent to distribute and unlawful distribution of cocaine base (crack) in violation of 21 U.S.C. § 841(a)(1). The Government relied primarily on an undercover officer who testified that an intermediary directed him to a seller whom he had never met. After giving the seller twenty dollars, the undercover officer waited while the seller acquired the drugs from [1408]*1408another individual sitting in a car parked nearby. The seller placed the drugs on a store window ledge, motioning for the officer to retrieve them. After recovering the drugs, the officer left the area and broadcast a “look-out” to the arrest team, describing both the seller and the man in the car. After the arrest team radioed the undercover officer that they had stopped two individuals fitting the descriptions in the broadcast, the officer returned to the scene and identified the two, including Davis as the seller.
Davis’s defense was mistaken identity. According to his theory of the case, he had been in a nearby liquor store at the time of the drug sale and had simply walked out of the store immediately before his arrest. At trial, the liquor store clerk confirmed that Davis had been in the store until just before the arrest. Further supporting the defense theory, one of the arresting officers testified that Davis had a beer in his hand when arrested. The defense attacked the undercover officer’s identification of Davis, noting the officer’s inexperience (he had been undercover only three weeks at the time of the incident); his inability to identify the intermediary; his broadcast of the look-out on an unrecorded police channel; and his failure to prepare the “buy report,” the only other evidence of his description of the seller, until after Davis’s arrest.
Prior to trial, the Government notified Davis that it would introduce evidence of three prior cocaine sales to prove knowledge and intent. In a written opposition and again orally at trial, Davis offered to stipulate to both knowledge and intent — in other words, he offered to concede that the person who possessed the drugs both knew that they were drugs and intended to sell them — arguing that in view of this concession, the Government wanted to use the prior incidents only to show his character in violation of Rule 404(b). Davis’s attorney reiterated both to the judge and to the jury that the only disputed issue in the case was whether the Government had proven possession beyond a reasonable doubt. After balancing the bad acts evidence’s probative value against its prejudicial effect pursuant to Rule 403, the district court admitted the evidence to show intent and knowledge. The jury convicted Davis on both counts.
In the second case before us, the Government charged Rochelle Ardall Crowder with possession with intent to distribute both crack and heroin. According to the Government’s witnesses, three police officers in a marked car observed Crowder exchange a small object for cash with another man. Seeing one of the officers motion to him, Crowder fled. According to the pursuing officer, Crowder removed a brown object from his pocket, dropping it as he scaled a fence. The officer testified that the object was a brown paper bag containing ninety-three “ziplock” bags of crack and thirty-eight glassine packets of heroin. After placing Crowder under arrest and searching him, the arresting officer recovered a pager and $988 in currency.
Acknowledging that the chase and arrest had occurred as described by the Government’s witnesses, Crowder contended at trial that he had possessed neither the paper bag nor the drugs. According to his theory of the case, the police officers came to his neighborhood looking for information about an unrelated, unsolved murder. He claimed that when he refused to talk, the police beat him and falsely accused him of possessing drugs. Defense witnesses confirmed that a detective investigating a murder had previously questioned Crowder. They also testified that they saw the same detective at the scene of Crowder’s arrest, that the police beat him during the arrest, and that the officer who first motioned to Crowder from the marked car called him by name. To refute the Government’s claim that he was selling drugs at the time, Crowder’s witnesses testified that the object he passed to the other man was a cigarette, that the large amount of cash was for home repair materials, and that, having no telephone, he had borrowed the beeper from the mother of his eight-year-old daughter to communicate when their daughter was with him.
Crowder’s first trial ended in a mistrial. Prior to the second trial, the Government advised Crowder that it would introduce evidence that he sold drugs to an undercover officer five months after the first trial. Objecting in writing and orally, Crowder offered [1409]*1409to concede every element of the crime except whether he possessed the drugs on the day of arrest. At a conference prior to trial, the district court reserved decision on the admission of the evidence. After the defense rested, the Government filed a second motion to introduce the evidence. Over Crowder’s continued objection and willingness to concede all the elements of the offense except possession, the district court admitted the evidence to show intent to distribute and knowledge of drug dealing. The jury convicted Crowder on both counts.
Both defendants appealed. After separate panels heard oral argument, the court on its own initiative consolidated the cases, hearing them en banc to resolve the question of the admissibility of prior bad acts evidence when criminal defendants offer to concede the element of a crime for which the Government seeks to introduce the evidence.
II.
The first sentence of Federal Rule of Evidence 404(b) is unambiguous: “[e]vidence of other crimes, wrongs, or acts is not admissible to prove the character of a person in order to show action in conformity therewith.” Such evidence, however, may be admissible under the second sentence of Rule 404(b) if introduced “for other purposes,” such as showing “knowledge” or “intent.” Intent and knowledge are two of the three elements — the third is possession — of a “possession with intent to distribute” charge under 21 U.S.C. § 841(a)(1), the crime charged in these cases. Thus, under the second sentence of Rule 404(b), the Government may ordinarily introduce bad acts evidence to prove intent and knowledge in a section 841(a)(1) prosecution if, under Rule 403, “its probative value is [not] substantially outweighed by the danger of unfair prejudice.” If the district court admits the evidence after Rule 403 balancing, the defendant may request an instruction limiting the jury’s use of the evidence to proof of intent and knowledge. See, e.g., 1 Edward J. Devitt, Charles B. Blackmar & Kevin F. O’Malley, Federal Jury Practice and Instructions § 17.08 (4th ed.1992).
These cases require that we determine the effect of a defendant’s offer to stipulate to— to concede — one or more of the issues for which the Government seeks to introduce the bad acts evidence. In the Davis case, when the Government introduced the evidence to prove the knowledge and intent elements of the crime, Davis offered to concede both. Following Rule 403 balancing, the district court admitted the evidence. In Crowder’s case, the district court admitted the evidence, also after Rule 403 balancing, to show intent to distribute and knowledge of drug dealing; Crowder conceded only the former.
Appellants argue that a defense offer to concede an element of the offense when combined with a jury instruction that the Government need not prove that element renders bad acts evidence inadmissible because the evidence is irrelevant to a disputed issue and substantially more prejudicial than probative. According to the Government, an offer to concede is only one factor the district court should consider when balancing probative value against prejudicial effect.
Nine of our sister circuits have considered the admissibility of bad acts evidence following a defense offer to concede intent, reaching essentially three different positions. The Third and Fifth Circuits permit district courts to balance the evidence’s prejudice against its probative value, although both circuits have held that an unequivocal offer to stipulate to intent generally renders extrinsic bad acts evidence inadmissible to show intent. See, e.g., United States v. Jemal, 26 F.3d 1267, 1274 (3d Cir.1994); United States v. Yeagin, 927 F.2d 798, 801-02 (5th Cir.1991). The Fourth, Sixth, Seventh, and Ninth Circuits similarly permit district courts to weigh probative value against prejudice, though they leave the district courts with broader discretion than the Third and Fifth Circuits. See, e.g., United States v. Hernandez, 975 F.2d 1035, 1039-41 (4th Cir.1992); United States v. Johnson, 27 F.3d 1186, 1193-94 (6th Cir.1994), cert. denied, — U.S. -, 115 S.Ct. 910, 130 L.Ed.2d 792 (1995); United States v. Brown, 34 F.3d 569, 572-74 (7th Cir.1994); United States v. Hadley, 918 F.2d 848, 852 (9th Cir.1990), cert. granted, 503 U.S. 905, 112 S.Ct. 1261, 117 [1410]*1410L.Ed.2d 491, and cert. dismissed, 506 U.S. 19, 113 S.Ct. 486, 121 L.Ed.2d 324 (1992). But see United States v. Kramer, 955 F.2d 479, 492-93 (7th Cir.) (Cudahy, J., concurring) (expressing view that law of circuit is mistaken), cert. denied, 506 U.S. 998, 113 S.Ct. 595, 121 L.Ed.2d 533 (1992).
In contrast, the Second Circuit has squarely held that under Rule 404(b) bad acts evidence must be relevant to an “actual” issue and that an offer to stipulate to an issue removes it from the case. United States v. Mohel, 604 F.2d 748, 751 (2d Cir.1979). Going further, the Second Circuit has held that even in the absence of a formal offer to stipulate, defendants may remove an issue from consideration as long as the decision not to dispute the issue is made “with sufficient clarity that the trial court will be justified (a) in sustaining objection to any subsequent cross-examination or jury argument that seeks to raise the issue and (b) in charging the jury that if they find all other elements established beyond a reasonable doubt, they can resolve the issue against the defendant because it is not disputed.” United States v. Figueroa, 618 F.2d 934, 942 (2d Cir.1980) (citing Mohel, 604 F.2d at 754). The First, Eighth, and Eleventh Circuits align themselves with the Second, see United States v. Garcia, 983 F.2d 1160, 1173-76 (1st Cir.1993) (citing United States v. Ferrer-Cruz, 899 F.2d 135, 139 (1st Cir.1990) (Breyer, J.)); United States v. Thomas, 58 F.3d 1318, 1321-23 (8th Cir.1995) (citing United States v. Jenkins, 7 F.3d 803, 806 (8th Cir.1993)); United States v. Taylor, 17 F.3d 333, 338 (11th Cir.) (citing United States v. Costa, 947 F.2d 919, 925 (11th Cir.1991), cert. denied, 504 U.S. 946, 112 S.Ct. 2289, 119 L.Ed.2d 213 (1992), and cert. denied, 506 U.S. 929, 113 S.Ct. 360, 121 L.Ed.2d 272 (1992)), cert. denied, — U.S. -, 115 S.Ct. 364, 130 L.Ed.2d 317 (1994), although neither the First nor the Eleventh Circuit has ever found a sufficiently explicit stipulation to require reversal.
Reviewing these opinions, we think the First, Second, Eighth, and Eleventh Circuits’ treatment of an offer to concede is most convincing. As we see it, a defendant’s offer to concede knowledge and intent combined with an explicit jury instruction that the Government no longer needs to prove either element gives the Government everything the evidence could show with respect to those two elements, doing so without risk that the jury will use the evidence for impermissible propensity purposes. In the absence of any other non-propensity purpose for the bad acts evidence, the evidence is therefore inadmissible because its only purpose could be “to prove the character of a person in order to show action in conformity therewith,” precisely what Rule 404(b)’s first sentence prohibits.
This approach is faithful to the plain language of Rule 404(b) as well as to its purpose: protecting defendants from the prejudice inherent in bad acts evidence. Underlying Rule 404(b) is the concern that a jury will use prior bad acts evidence to draw general conclusions about a defendant’s character and to make unfounded judgments about the defendant’s conduct on a particular occasion. The problem with character evidence in general and pri- or bad acts evidence in particular is not that character is irrelevant — indeed, it is relevant, see Concurring Opinion of Silberman, J., at 1416 — but that using bad acts evidence to show character can “weigh too much with the jury and ... so overpersuade them as to prejudge one with a bad general record and deny him a fair opportunity to defend against a particular charge.” Michelson v. United States, 335 U.S. 469, 475-76, 69 S.Ct. 213, 218, 93 L.Ed. 168 (1948). Evidence of extrinsic bad acts thus carries the risk of prejudice — a risk long recognized in the common law and explicitly embraced by the drafters of the Federal Rules — for in our criminal justice system, “ ‘a defendant must be tried for what he did, not for who he is.’ ” United States v. Daniels, 770 F.2d 1111, 1116 (D.C.Cir.1985) (quoting United States v. Myers, 550 F.2d 1036, 1044 (5th Cir.1977)). As we have held, “ ‘[t]he exclusion of other crimes evidence is not simply a “technicality”[;] ... it reflects and gives meaning to the central precept of our system of criminal justice, the presumption of innocence.’ ” United States v. Dockery, [1411]*1411955 F.2d 50, 53 (D.C.Cir.1992) (quoting Daniels, 770 F.2d at 1118).
In order to protect defendants from the prejudicial effects of bad acts evidence without also weakening the Government’s ability to prove its case, the conceded elements must be completely removed from the trial. A defendant’s offer to concede, therefore, must be unequivocal: the concession’s language must be unambiguous and the defendant may not undermine the offer by later challenging the conceded element. See United States v. Manner, 887 F.2d 317, 322 (D.C.Cir.1989) (bad acts evidence may be admissible when concession equivocal). To ensure that the jury clearly understands that the concession releases the Government from its burden of proof on the conceded elements, the defendant must accept, as counsel for Davis acknowledged at oral argument, an instruction stating clearly that in order to convict, the jury need find only the contested element(s). In Davis’s case, the instruction describing the elements of the crime, see, e.g., Devitt at § 54.07, could be followed by something along the following lines: “By Davis’s agreement, the Government need not prove either knowledge or intent. Your job is thus limited to the possession element of the crime. Therefore, in order to meet its burden of proof, the Government must prove beyond a reasonable doubt only one element of the crime, that Horace Davis was in possession of the cocaine base charged in the indictment.” By offering this suggestion, we do not mean to dictate the precise language district courts must use, merely to indicate one way to explain Davis’s concession to the jury. Whatever language the district court chooses would, in order to avoid confusing the jury, replace the standard instruction regarding the Government’s obligation to prove every element of the crime beyond a reasonable doubt. See, e.g., Criminal Jury Instructions for the District of Columbia, Instruction 2.08. This will not, as our dissenting colleagues argue, see Dissent at 1427, create a problem in multi-defendant cases. Because trial judges may instruct the jury differently for different defendants, see, e.g., Criminal Jury Instructions for the District of Columbia, Instruction 2.51-2.52 (in apparent recognition of the permissibility of different instructions for different defendants, recommending the following instruction: “Unless I have stated otherwise to you, you should consider each instruction that the Court has given to apply separately and individually to each defendant on trial” (emphasis added)), defendants who do not concede any elements of the crime would continue to be entitled to an instruction requiring the Government to prove every element beyond a reasonable doubt. And because an unequivocal concession of an element of a crime waives a defendant’s constitutional right to have the Government prove that element beyond a reasonable doubt, the district court will have to ensure that the concession is knowing and voluntary. See Garcia, 983 F.2d at 1175-76; cf. Brady v. United States, 397 U.S. 742, 748, 90 S.Ct. 1463, 1468-69, 25 L.Ed.2d 747 (1970) (“Waivers of constitutional rights not only must be voluntary but must be knowing, intelligent acts done with sufficient awareness of the relevant circumstances and likely consequences.”).
Davis had no opportunity to consider such an instruction because the district court admitted the bad acts evidence under Rule 403 to show intent. But Davis offered to concede both the intent and knowledge elements of the crime, and we think his concession was unequivocal. His written motion offered to concede “that the person who sold the undercover officer the drugs in this case had the intent to distribute any and all drugs recovered by the police.” Davis’s attorney also told the district court that “[o]nce the government has proved possession in this case, it’s our position they’ve proved knowledge and intent as well.” In both opening statement and closing argument, Davis’s attorney reiterated that knowledge and intent were not at issue. See J.A. 33 (“[T]he real person who sold drugs to [the undercover officer] had the intent ... to sell the drugs ... [and] had knowledge of the drugs”); J.A. 51 (“We don’t contest that the real seller had the intent to distribute the drugs.”). Because this offer to concede was unequivocal, the evidence’s only remaining purpose was to [1412]*1412show Davis’s propensity. Assuming Davis’s willingness to accept the type of instruction we suggest above, the district court should have excluded the evidence under Rule 404(b)’s first sentence.
That Davis conceded that the person with the drugs knew what they were and intended to distribute them, rather than conceding more particularly that he had such knowledge and intent, does not undermine the concession. See Dissent at 1424. Because Davis’s position was that he did not possess the drugs at all, he could not have conceded that he knew what the substance was or that he intended to distribute the drugs. By conceding that the person with the drugs had knowledge of what they were and had the intent to distribute them and thus conceding that he had the knowledge and intent to distribute if the jury were to find that he possessed the drugs, Davis clearly conveyed that the only remaining element was whether he possessed the drugs. If, on remand, Davis reiterates his unequivocal concession and remains willing, as he was at oral argument, to accept an instruction limiting the jury’s consideration to the possession element of the crime, Davis’s concession would completely relieve the Government of having to prove knowledge and intent.
Our dissenting colleagues suggest that a defendant’s intent to distribute drugs could also be his “motive” to possess the drugs, another permissible purpose under Rule 404(b)’s second sentence. See Dissent at 1424. In other words, a defendant’s motive for having the drugs might be that he seeks to distribute them, thus making the evidence, according to the dissent, directly relevant to possession, the sole unconceded element of this crime. If the dissent is correct, unlike a concession of intent or knowledge, defendant’s concession of motive to possess would not categorically bar the evidence under Rule 404(b)’s first sentence. In the case of knowledge and intent, both of which are elements of a section 841(a)(1) crime, the concession, combined with an instruction limiting the jury’s consideration to possession, completely relieves the Government of having to prove anything with respect to either element. A defendant’s concession of a motive to possess is different. Because the Government seeks to introduce the bad acts evidence to show motive in order to have the jury infer beyond motive to an element of the crime that it must prove — possession—a concession may not completely substitute for live evidence. Of course, the Government did not offer the evidence for this purpose in Davis’s case. We thus have no need to resolve the question of the admissibility of other drug offenses to show motive to possess. We note only that with respect to such evidence, as long as a defendant remains willing to concede knowledge and intent to distribute and to accept an instruction limiting the jury’s consideration to possession, the district court could consider neither knowledge nor intent in its Rule 403 balancing.
In Crowder’s ease, he too offered a defense based solely on the theory that he did not possess the drugs. Like Davis, he contested neither intent nor knowledge, offering instead to concede both that the “substances seized were cocaine base and heroin,” and that “anybody who possessed those drugs possessed them with the intent to distribute.” His attorney stated that “[t]he issue in this case is: Did he or did he not possess those drugs? That’s the threshold thing that the Government has to be able to prove in this case. The rest of it in terms of what the facts — -what the evidence will show, we concede.” Read together, these statements amount to an unequivocal offer to concede both the knowledge and intent to distribute elements of the crime.
The district court apparently believed that Crowder undermined his concession of intent by presenting evidence that he had legitimate reasons for the beeper and $988, and a benign explanation — selling a cigarette — for the observed transaction. See J.A. 243. We disagree. The beeper, the money, and the transaction were each relevant to show not only Crowder’s intent, but also his possession. That Crowder contested the significance the Government attributed to its evidence is therefore completely consistent with his defense that he did not possess the drugs at all. Although Crowder’s evidence might also counter the probative value of the Government’s evidence to show intent, this does [1413]*1413not make intent an issue because Crowder’s concession of intent, combined with an instruction limiting the jury’s consideration to possession, would completely eliminate the possibility that the jury would find against the Government on intent. Indeed, preventing a defendant wishing to concede intent from countering evidence showing possession simply because that evidence is also relevant to intent would make a defense based on lack of actual possession virtually impossible because almost all of the Government’s evidence could be relevant to intent.
Unlike the Davis case, the Government also sought to introduce the evidence for another reason: to show Crowder’s modus operandi. Although not listed in Rule 404(b)’s second sentence as a permissible purpose for prior bad acts evidence, modus operandi evidence may be admissible when it reveals a defendant’s personal criminal “identity.” See, e.g., United States v. Connelly, 874 F.2d 412, 416 n. 7 (7th Cir.1989); see also United States v. Levi, 45 F.3d 453, 455 (D.C.Cir.1995). While we have some concern that using bad acts evidence to show identity in a drug possession case is tantamount to showing propensity, we need not decide that question here. The district court did not admit the evidence for that purpose, and for good reason: The Government failed to establish any basis for admitting the evidence to show criminal identity. In order to use bad acts evidence to show criminal identity through modus operandi, the Government must establish not only that the extrinsic act “bears some peculiar or striking similarity” to the charged crime, United States v. Garbett, 867 F.2d 1132, 1135 (8th Cir.1989), but also that it is the defendant’s trademark, “so unusual and distinctive as to be like a signature,” United States v. Garcia-Rosa, 876 F.2d 209, 224 (1st Cir.1989) (citations and internal quotation marks omitted). See generally Edward J. Imwinkelried, Uncharged Misconduct Evidence §§ 3:10-3:14 (1995). The facts in Crowder’s case do not satisfy either requirement. The subsequent sale involved clear and black ziplock bags of cocaine and no heroin, while the brown paper bag in the charged offense contained clear and blue ziplock bags of cocaine, as well as glassine packets of heroin. These two events are hardly “striking[ly] similar[].” Even if the subsequent sale resembled the charged offense, the Government presented no evidence, expert or otherwise, that the mode of packaging was distinctive to Crowder or even unusual in the drug trade. Crowder’s subsequent drug sale was thus inadmissible to establish modus operandi.
In addition to admitting the evidence to show intent, the district court accepted it to show “knowledge of drug dealing.” In this respect, Crowder’s case is fundamentally different from Davis’s. Knowledge of drug dealing, of course, is not the same as the knowledge element of the crime. The former shows a general understanding about the drug trade; the latter is a defendant’s awareness that the particular substance he possessed was cocaine base. The question, then, is this: What relevance does Crowder’s knowledge of drug dealing have for proving the only unconceded element of the crime, whether Crowder possessed drugs during the alleged crime? Although using bad acts evidence to show a defendant’s “knowledge of drug dealing” could be a proxy for showing his character, a jury could infer possession from knowledge of drug dealing without relying on an impermissible propensity inference: A familiarity with the drug trade may make it more likely that a defendant knew where or how to acquire drugs, thus making it more likely that he possessed them at the time of arrest. Knowledge of drug dealing can therefore be a permissible purpose for the evidence under the second sentence of Rule 404(b). Cf. 22 Charles A. Wright & Kenneth W. Graham, Jr., Federal Practice and Procedure: Evidence § 5245, at 506 (1978) (“Because of the varying ways in which other crimes, wrongs, or acts can prove knowledge, it is not possible to specify the requisite similarity between the charged crime and the other incident.”). Although the district court admitted the evidence for that purpose, its Rule 403 balancing included consideration of the evidence’s probative value to show intent to distribute. Because Crowder offered to concede intent to distribute, we remand for another “on the record determination as to whether the preju[1414]*1414dicial effect of the prior bad acts evidence substantially outweighs its probative value.” See Manner, 887 F.2d at 323. In weighing the evidence, the district court may consider its probative value to show “knowledge of drug dealing” only to the extent that “knowledge of drug dealing” shows that Crowder knew where or how to acquire drugs. In view of the attenuated nature of this inference, as well as the Government’s strong evidence of possession — the observed transaction, beeper, money, and the police officer’s testimony about the chase — and the likelihood of the jury misusing the evidence, we think the Government’s burden, demonstrating that the evidence’s probative value is not substantially outweighed by the danger of unfair prejudice, will be heavy.
III.
Contending that our holding will interfere with the Government’s ability to prove its case, our dissenting colleagues would permit the Government to introduce bad acts evidence even in the face of an unequivocal concession. Underlying the dissent’s concerns is its failure to distinguish between bad acts evidence, with its own specialized rule of admissibility, and other evidence susceptible only to the general provisions of Rules 401 and 403. As stated in the Advisory Committee Notes to Rule 401,
[a]n enormous number of [relevancy questions] fall in no set pattern, and [Rule 401] is designed as a guide for handling them. On the other hand, some situations recur with sufficient frequency to create patterns susceptible of treatment by specific rules. Rule 404 ... [is] of that variety; [it] also serve[s] as [an] illustration! ] of the application of [Rule 401] as limited by the exclusionary principles of Rule 403.
The plain language of Rule 404(b) requires that district courts evaluate the admissibility of bad acts evidence by deciding whether it is inadmissible under the rule’s first sentence or possibly admissible under its second. Grounded in Rule 404(b), our holding excluding bad acts evidence thus does not, as the dissent fears, encompass non-bad acts evidence intrinsic to a crime such as the details of a victim’s wounds in an assault case or the actual film in a prosecution for distributing obscene material. See Dissent at 1421 n.2. For the same reason, we would not exclude Crowder’s beeper, see Dissent at 1421; it is not bad acts evidence, and in any event, it is relevant to possession. Because neither Estelle v. McGuire, 502 U.S. 62, 69, 112 S.Ct. 475, 480-81, 116 L.Ed.2d 385 (1991), nor Mathews v. United States, 485 U.S. 58, 65, 108 S.Ct. 883, 887-88, 99 L.Ed.2d 54 (1988), on which the dissent relies so heavily, see Dissent at 1421-1426, involved an explicit concession of an element of a crime, they are irrelevant here.
The dissent accuses us of relying on a “psychological insight” that prosecutors seek to use bad acts evidence in lieu of a concession in order to convince juries to rely on the evidence for impermissible purposes. Dissent at 1422. Our concern under Rule 404(b), however, is not with the prosecutor’s motive, but rather with the jury’s use of the evidence to show a defendant’s propensity. Right or wrong, Rule 404(b) establishes a separate standard for bad acts evidence precisely because in the absence of such a standard, juries might use the evidence to make propensity inferences. With all respect, we think our colleagues’ disagreement is with Rule 404(b), not with us.
Along the same lines, the dissent argues that our holding will undermine prosecutors’ “legitimate trial strategy” of presenting “live testimony and documentary evidence” rather than accepting a concession and a “cold” jury instruction. Dissent at 5. The dissent does not articulate exactly how the Government’s inability to introduce extrinsic bad acts evidence to prove knowledge or intent could affect its case, and with good reason — it will not. If the jury fails to find possession beyond a reasonable doubt, then the Government’s inability to introduce bad acts evidence for the other two elements will not affect its ease at all unless the jury uses the evidence to show possession through propensity, a purpose explicitly prohibited by the first sentence of Rule 404(b). If the jury does find possession beyond a reasonable doubt, the only way the Government’s inability to introduce bad acts evidence could affect [1415]*1415the outcome is if the jury ignores its instruction that the Government need only prove possession and acquits the defendant because the Government failed to prove intent and knowledge beyond a reasonable doubt. Put another way, not only would the jury have to disregard a clear, simple instruction, but it also would have to acquit on a theory of the facts not advanced by either party and unsupported by any record evidence: that the defendant possessed the drugs, but either had no knowledge of what they were or had no intent to distribute them. The dissent apparently believes that juries would be less likely to develop their own theories if they hear “live” bad acts evidence on intent and knowledge than if courts clearly instruct them that they need only consider possession. Such an assumption strains credulity and ignores the principle that “[a] jury is presumed to follow a trial court’s instructions ... particularly ... when ... the trial judge’s instructions [are] clear, concise, [and] careful.” United States v. Jackson, 627 F.2d 1198, 1213 (D.C.Cir.1980) (citing Shotwell Mfg. Co. v. United States, 371 U.S. 341, 367, 83 S.Ct. 448, 463-64, 9 L.Ed.2d 357 (1963)).
Although juries sometimes do ignore instructions, we think the risk of their doing so is significantly greater under the dissent’s approach, which would admit the bad acts evidence and instruct the jury to consider it only for the uncontested elements of intent and knowledge, not for the purpose prohibited by Rule 404(b). In the Davis case, for example, the district court instructed the jury as follows:
If you determine that the government has proven beyond a reasonable doubt that on February 27, 1991, it was Mr. Davis who sold drugs to Officer Farmer and possessed the drugs found in the blue Jet-ta with the intent to distribute them, and if you also determine that he committed the similar acts as well, then you may, but you do not have to, draw an inference that in doing the acts charged in the indictment Mr. Davis knew about the drugs in the Jetta and had the intent to distribute them.
You may not consider the evidence of the other arrests for any other purpose. Specifically, you may not use the evidence to conclude that, based upon his other arrest, Mr. Davis is guilty of the offenses with which he is charged in this case.
The ability of such a limiting instruction to overcome the prejudicial effects of bad acts evidence can be, we have said, an “ ‘unmitigated fiction.’ ” Daniels, 770 F.2d at 1118 (D.C.Cir.1985) (quoting Krulewitch v. United States, 336 U.S. 440, 453, 69 S.Ct. 716, 723, 93 L.Ed. 790 (1949) (Jackson, J., concurring)). As we have observed, the “‘mental gymnastic’” such an instruction requires of the jury may be “well beyond mortal capacities.” Daniels, 770 F.2d at 1118 (quoting Nash v. United States, 54 F.2d 1006, 1007 (2d Cir.), cert. denied, 285 U.S. 556, 52 Ed. 945 (1932)). Our approach does not require juries to perform any “mental gymnastics].” The court excludes the bad acts evidence, instructing the jury that to convict it need find only possession beyond a reasonable doubt. Hardly confusing to juries, this approach protects defendants from the prejudicial effects of bad acts evidence without weakening the Government’s ability to prove its case.
The dissent argues that removing knowledge and intent from the jury’s consideration conflicts with Federal Rule of Criminal Procedure 23(a) which requires the Government’s consent before defendants may waive their rights to jury trials, as well as the Supreme Court’s decision in Singer v. United States, 380 U.S. 24, 85 S.Ct. 783, 13 L.Ed.2d 630 (1965), upholding the rule’s constitutionality. Rule 23, entitled “Trial by Jury or by the Court,” only restricts defendants’ ability to waive a jury trial in favor of a bench trial; it does not limit their ability to waive a jury trial altogether. Neither Rule 23 nor Singer has any bearing on defendants’ ability to concede particular elements of a crime. Indeed, under the dissent’s logic, Rule 23(a) would require a defendant to obtain the Government’s consent to a guilty plea because by pleading guilty, the defendant will have, in effect, “waive[d] his right to a jury trial.” Dissent at 1426.
Finally, the dissent raises a host of practical concerns about our disposition of these two eases. Its “long list of difficult ques[1416]*1416tions,” Dissent at 1428, consists primarily of issues not before us. When, if ever, the dissent asks, could a defendant withdraw a stipulation? Under Federal Rule of Evidence 609(a), may the Government impeach a testifying defendant with his prior offenses? When must a defendant make the offer to concede and must it be in writing? While these may be “obvious questions” to our dissenting colleagues, Dissent at 1429, we need address none of them because Crowder and Davis chose not to withdraw their concessions, neither took the stand, and the Government did not challenge the timing and form of either concession. We think it best to leave resolution of these issues in the first instance to the district courts.
IV.
We reverse Davis’s conviction and remand for a new trial. We remand Crowder’s case for the district court to determine, in accordance with this opinion, whether the bad acts evidence’s probative value is substantially outweighed by the danger of unfair prejudice. Because we remand both cases, we do not reach the other issues that defendants raise.
So ordered.