dissenting:
In the vast majority of criminal cases tried in this circuit, transcripts of witness testimony are not available at the time of closing arguments. This means that pros-*704eeutors and defense counsel must rely on their recollections in making those arguments, and that judges must rely on theirs in ruling on objections. Innocent mistakes of recollection are inevitable and hardly uncommon. For protection from prejudice, our adversary system relies on the opportunity each side has to challenge the other’s misstatements before the jury, and upon the court’s standard admonition that it is the jury’s recollection that controls. In the end, the jury’s memory of what a witness actually said provides the corrective for errors made by the parties.
In light of this reality, it is not surprising that although “it is error for a prosecutor to mischaracterize evidence in a summation[,] [i]t is also clear ... that an error in a prosecutor’s summation will only rarely warrant reversal of a conviction.” United States v. Donato, 99 F.3d 426, 432 (D.C.Cir.1996); see also United States v. Edelin, 996 F.2d 1238, 1243 (D.C.Cir.1993) (“[W]e have generally been ‘chary of reversing convictions solely on the grounds of a misstatement in a closing argument.’ ”) (citation omitted). Indeed, it is so rare that my colleagues are unable to cite a single case in which we have reversed a conviction solely for a prosecutor’s misquotation of testimony that the jury itself heard.
It is “the law of this circuit that, even where challenges to a prosecutor’s closing argument have been preserved through timely objection, we will reverse a conviction and require a new trial only if we determine that the defendant has suffered ‘substantial prejudice.’ ” United States v. Childress, 58 F.3d 693, 715 (D.C.Cir.1995) (quoting United States v. North, 910 F.2d 843, 897-98 (D.C.Cir.1990)). We have “framed the test for prejudice in terms of the severity of the prosecutor’s misconduct, the measures adopted to cure the misconduct, and the certainty of conviction absent the improper remarks.” Id. at 715; see also United States v. Gartmon, 146 F.3d 1015, 1026 (D.C.Cir.1998) (noting variety of similar formulations). It is only in the most egregious of cases that we will consider reversal, see North, 910 F.2d at 897 n. 33, and an examination of the applicable factors makes clear that this is not such a case.
A
As the court recognizes, the first step in determining the severity of a prosecutor’s misstatement is to compare it with the witness’ actual testimony. A misstatement is error, but only “to the extent that [it] overstate^]” the testimony. Gartmon, 146 F.3d at 1025 (quoting United States v. Perholtz, 842 F.2d 343, 360 (D.C.Cir.1988)). In this case there clearly was an overstatement, but the difference between the witness’ testimony and the prosecutor’s characterization is not as substantial as the court’s opinion suggests.
The problem in this case began with a classic error in trial technique. See Thomas A. Mauet, Fundamentals of Trial Techniques 385 (1980). The prosecutor asked what was in essence a compound question: “Mr. Thomas, you believe that you know Watson’s girlfriend, Tyra Jackson, right?” In so doing, he effectively asked both whether the witness knew Ms. Jackson, and whether the witness knew her to be the defendant’s girlfriend. At that point, the equally-elassic “objection as to form” would have been in order. Defense counsel, however, did not make it. Instead, the cross-examination unfolded as follows:
Prosecutor: Mr. Thomas, you believe that you know Watson’s girlfriend, Tyra Jackson, right?
Thomas: I never testified I knew her or not.
Prosecutor: You believe that you may have met her once or twice, right?
Thomas: Maybe.
The court may be correct in stating that the defense had no obligation to object to the prosecutor’s question. But my colleagues are wrong in suggesting that the *705defense could “rest satisfied” with the witness’ response because it “did not produce damaging testimony.” Op. at 701. In fact it did. As the court notes, the compound question yielded an ambiguous response— the classic consequence of asking such a question. But ambiguity is not the same as the absence of evidence. A reasonable jury could have concluded that Thomas would have disputed the implication that Jackson was Watson’s girlfriend if it were untrue or if he did not know it to be true— particularly since he had already exhibited a willingness to resist the prosecutor’s assumptions. See 5/1/96 Tr. at 50. (“I never testified I knew her or not.”). Accordingly, a reasonable jury could well have interpreted Thomas’ answers as assent to the implied question — do you know Tyra Jackson to be Watson’s girlfriend? Although the defendant did not have to risk “per-fectfing] the government’s case by clarifying the witness’ response,” Op. at 701, by not doing so he accepted the risk that the jury would reasonably read the ambiguity against him.1
It is true that when the prosecutor recounted the exchange in closing argument, he erred by “eliminating the ambiguity” in Thomas’ testimony. Op. at 701. The prosecutor told the jury that Thomas had said: “I think I’ve met Watson’s girlfriend, Tyra Jackson once or twice.” This was a stronger version of the witness’ testimony and hence was error. But since it was an inference that a reasonable jury could have derived on its own, the measure of the difference is one of degree. The prosecutor did no more than make express what a juror could reasonably have found implicit in the witness’ answers. This was error, but not egregious error.
Nor was the prosecutor’s misstatement an intentional one. Recalling the precise contours of a witness’ testimony is a difficult task in the best of circumstances, made all the more difficult here by the witness’ ambiguous answer to the prosecutor’s compound question. When defense counsel made his motion in limine concerning the statement, he did so orally, without notice, and without obtaining a transcript to support his motion. In the absence of that transcript, all of the participants were forced to rely on their recollections — and all of those recollections were erroneous to some degree. See Revised Appendix (“App.”) 161. Although the prosecutor’s memory was worse than that of defense counsel, it did not vary significantly from that of the judge.2 Thus, it can hardly be said that the defense’s uncorroborated allegation put the prosecutor on notice that he had not obtained the admission he thought he had.
Finally, in measuring the severity of the prosecutor’s error, it is also important to note that it involved just one sentence in each of the prosecutor’s two closing arguments.3 Those arguments spanned more than twenty pages of transcript. As we have said many times before, such isolated misstatements rarely amount to severe misconduct. See, e.g., Gartmon, 146 F.3d at 1026; North, 910 F.2d at 897; Perholtz, 842 F.2d at 361.
B
The next factor to consider in measuring the substantiality of prejudice is whether *706measures were available to mitigate its impact. The error at issue here was the inaccurate recitation of testimony that the jury itself heard. Hence, if the jury relied on its own recollection, rather than on that of the prosecutor, the error would be without effect. The judge gave two separate instructions designed to ensure precisely that result:
If any reference by the court or the attorneys to evidence does not coincide with your own recollection of the evidence, it is your recollection which should control during your deliberations.
The statements and arguments of the lawyers are not evidence. They are only intended to assist you in understanding the evidence.
The court also gave an additional instruction aimed directly at the kind of problem engendered by a compound question:
Sometimes a lawyer’s question suggests that something is a fact. Whether or not something is a fact depends on the witness’s answer, not the lawyer’s question. A lawyer’s question is not evidence.
Both the Supreme Court and this court have repeatedly held such instructions sufficient to mitigate prejudice caused by prosecutors’ misstatements in closing arguments.4 Moreover, it bears emphasizing that this is not a case in which the prosecutor asserted knowledge of evidence neither seen nor heard by the jury, nor subject to cross-examination by the defense.5 In such a case, it might be argued that an instruction that the jury’s recollection controls is of questionable value since the jury has no recollection on which to rely. Here, by contrast, the dispute was solely about evidence the jury did hear, and as long as the jury followed the court’s instructions the prosecutor’s error would be mitigated. See Richardson v. Marsh, 481 U.S. 200, 211, 107 S.Ct. 1702, 95 L.Ed.2d 176 (1987) (“[J]uries are presumed to follow their instructions.”).
We also cannot ignore defense counsel’s failure to use his closing argument to tell the jury that the prosecutor had misstated the evidence. Pointing out such a misstatement can have a powerful, even devastating effect on an opponent’s case. Had defense counsel used his closing argument in that fashion, we doubtless would have found it sufficient to mitigate the impact of the misstatement. See, e.g., United States v. Williams-Davis, 90 F.3d 490, 507-08 (D.C.Cir.1996) (holding that although government’s opening statement blamed defendants for two murders as to which it never introduced evidence, defense counsel was able “to use the variance between the government’s opening and its proof to sow doubt of the prosecution among the jurors”); North, 910 F.2d at 895; Cross v. United States, 353 F.2d 454, *707455 (D.C.Cir.1965). The defense’s failure to take advantage of this curative opportunity cannot put it in a better position than if it had — at least not without creating a powerful incentive to let misstatements pass without comment in the hope of obtaining a second bite at the apple if the jury’s verdict should be unfavorable.
C
Finally, we must consider the weight of the government’s evidence. As my colleagues correctly note, the government’s case against Watson cannot be characterized as overwhelming. But the evidence against the defendant was certainly “weighty,” and that is sufficient to uphold his conviction in light of the other factors discussed above. See Brecht v. Abrahamson, 507 U.S. 619, 639, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993) (holding Kotteakos harmless error standard satisfied where “evidence of guilt was, if not overwhelming, certainly weighty”); Childress, 58 F.3d at 716 (indicating that the various factors must be weighed against each other in determining whether prosecutorial remarks caused substantial prejudice). The evidence tying Watson to $14,000 worth of crack cocaine was as follows.
First, a search of Watson’s person produced a key to a car that was parked fifteen feet from the spot at which Watson was arrested. Although the court describes the searching officer’s testimony as “disputed,” Op. at 698, that description is overstated. Only one defense witness, Anthony Shank, testified about the search of Watson, and he merely said that “the only thing I saw them remove from him was his shoe strings and belt.” App. 152. Shank did not affirmatively testify that there was no key; he was not even asked whether he saw a key. Nor is Shank’s testimony inconsistent with the officer finding the key when Shank was not looking; there was no testimony that anyone saw the officer find the key elsewhere. Indeed, although the court may regard Shank’s testimony as a powerful attack on the officer’s credibility, apparently defense counsel did not appraise it the same way: he did not even mention Shank’s testimony in his closing argument.6
Second, inside the car’s glove compartment the police found large rocks of crack cocaine wrapped in a Shaw’s Jewelry bag. The officers testified that when Watson saw they had found the bag, he began struggling violently to escape. Inside the Shaw’s bag, along with the crack, were five black ziplock bags matching five other bags that fell from Watson’s hand when he was arrested. See 4/25/96 Tr. at 19-20, 224.7 And inside Watson’s house was a receipt for a purchase at Shaw’s Jewelry just seven weeks before — a purchase made using the same alias Watson gave police *708when he was arrested on the instant charge. It would be surprising if the jury regarded that purchase as nothing more than an unfortunate coincidence.
Third, the district court properly admitted, under Federal Rules of Criminal Procedure 403 and 404(b), evidence that Watson previously had been convicted of committing the same crime' — possession with intent to distribute cocaine — on the same city block. See 4/26/96 Tr. at 13-14. Although this cannot alone prove that Watson possessed the drugs on the instant occasion, it can be used to prove he intended to distribute the cocaine in the Shaw’s bag, and “may be a ‘brick in the wall of evidence’ proving possession.” United States v. Crowder, 141 F.3d 1202, 1208 n. 5 (D.C.Cir.1998) (en banc).
My colleagues suggest that upon retrial, the district court may wish to reconsider the admissibility of Watson’s prior conviction. Although the district court is certainly free to reconsider anything it likes, there is no reason to reconsider its decision to admit this prior crimes evidence. We have repeatedly upheld the admission of such evidence in similar circumstances,8 and the reason the court gives for regarding admissibility as a close question in this case is unpersuasive.
The' court suggests that the evidence of Watson’s prior crime can go only to prove “non-contested issues.” Op. at 702. The court apparently adopts defendant’s argument that the element of intent was uncontested in this case, because his defense was mistaken identification rather than the absence of an intent to distribute cocaine.9 But that is precisely the argument we rejected, en banc, in Crowder, where we held prior crimes evidence relevant notwithstanding a defense of mistaken identification and notwithstanding defendant’s offer to stipulate that whoever did possess the drugs in question had the necessary intent. See Crowder, 141 F.3d at 1206; see also Estelle v. McGuire, 502 U.S. 62, 69-70, 112 S.Ct. 475, 116 L.Ed.2d 385 (1991) (“[T]he prosecution’s burden to prove every element of the crime is not relieved by a defendant’s tactical decision not to contest an essential element of the offense.... [The prosecution is not required] to refrain from introducing relevant evidence simply .because the defense chooses not to contest the point.”).
In short, the district court properly admitted the evidence of Watson’s prior drug crime to prove Watson’s intent with respect to the cocaine at issue in this case. Moreover, as we noted in Crowder, “[p]roof of an individual’s intent to commit an act may itself serve as proof that the individual committed the act,” and hence “other-offense evidence of intent would have probative value not just on the intent element, but also on the possession element of the offense.” 141 F.3d at 1208. When this is taken together with the other evidence connecting Watson to the bag of crack cocaine, the government’s evidence is sufficiently weighty to bar a conclusion that Watson was substantially prejudiced by the limited (and mitigated) error the prosecutor committed in closing argument.
D
It may well be that in the not-too-distant future even routine criminal trials will *709have the benefit of real-time transcripts of witness testimony. See Toni Locy, Law Meets Technology in Courtroom No. 9, Wash. Post, Aug. 21, 1997, at J1. When that day comes, disputes over testimony will be resolved by reference to transcripts rather than memories. In the meantime, however, it is inevitable that trial lawyers will suffer from innocent misrecollections. We have always relied on the self-corrective nature of the adversary system, combined with instructions from the court, to police all but the most egregious of these kinds of errors. Because I am unable to conclude that the defendant suffered substantial prejudice as a consequence of the error that occurred in this case, I respectfully dissent from the reversal of his conviction.
. On Lhe other hand, as the district court pointed out, if Jackson were not defendant's girlfriend or if Thomas did not know, defense counsel could easily have covered the point during his redirect examination of Thomas. See 5/2/96 Tr. at 24.
. In response to defense counsel’s contention that “there's no evidence that Tyra Jackson was the girlfriend,” the court responded: "I thought the witness answered 'yes.' ” 5/2/96 Tr. at 23.
.Indeed, while the sentence in the initial closing argument was erroneous because the prosecutor presented it as if it were a direct quotation of the witness’ testimony, the sentence employed in the rebuttal appears more as characterization than quotation, and hence may not have been error at all. See Donato, 99 F.3d at 432 (holding that "fair, if disputed, characterization” of testimony does not constitute error).
. See, e.g., Zafiro v. United States, 506 U.S. 534, 541, 113 S.Ct. 933, 122 L.Ed.2d 317 (1993) ("[T]he District Court admonished the jury that opening and closing arguments are not evidence.... These instructions sufficed to cure any possibility of prejudice.”); Gartmon, 146 F.3d at 1026 ("[T]he judge gave the standard limiting instruction that lawyers’ arguments are not evidence and that the jury’s recollection of the evidence controls. We have repeatedly said this kind of instruction can mitigate the impact of erroneous jury argument.”); North, 910 F.2d at 897 ("Our unwillingness to reverse a conviction has been particularly pronounced when the trial judge issues curative instructions.... Here [the judge] ... explicitly reminded the jurors that 'the statements, opinion and arguments of counsel are not evidence’.... [and that] the jurors' 'recollection alone' is controlling as to 'all aspects of the evidence.' The District Judge could not have more directly communicated to the jury the limited evidentiary value of closing arguments.”).
. United States v. Teffera, 985 F.2d 1082 (D.C.Cir.1993), cited by the court, is an example of such a case. There, we reversed a conviction because the evidence was insufficient to convict. See 985 F.2d at 1089. We indicated in dicta, however, that we would also have reversed based on the prosecutor’s repeated references in closing argument to alleged "eye contact” between the two code-fendants — which we characterized as “phantom evidence” that was not "adduced at trial.” Id. at 1089 n. 6.
. A second defense witness, a high school student, testified that the key the government introduced into evidence ”look[ed] like” a key he saw in the possession of a different individual (Everett Hawkins) five hours earlier on the day of Watson's arrest. See App. 131. Like Shank’s, that testimony was not inconsistent with Watson having the key when he was arrested.
The court also suggests that the jury disbelieved the police witnesses, because it acquitted Watson of a firearms charge despite their testimony that they thought they saw Watson pass a gun to a codefendant. By the same logic, we could say that the jury disbelieved Shank, because it found defendant guilty of assaulting a police officer despite Shank's testimony that the assault was actually perpetrated by the police. In fact, the better view is simply that propounded by the Supreme Court in United States v. Watts: "[I]t is impossible to know why a jury found a defendant not guilty on a certain charge. An acquittal is not a finding of any fact.... Without specific jury findings, no one can logically or realistically draw any factual finding inferences.” 519 U.S. 148, 117 S.Ct. 633, 637, 136 L.Ed.2d 554 (1997).
. The court notes that die drugs in the latter five bags were of a different concentration than the drugs recovered from the car. That difference was quite small (39% vs. 42%), and not at all inconsistent with all of the crack cocaine coming from the same batch. See United States v. Robinson, 59 F.3d 1318, 1320 (D.C.Cir.1995) (citing testimony of DEA chemist).
. See, e.g., United States v. Burch, 156 F.3d 1315, 1324 (D.C.Cir.1998); United States v. Mitchell, 49 F.3d 769, 776 (D.C.Cir.1995); United States v. Johnson, 40 F.3d 436, 441 n. 3 (D.C.Cir.1994). The court notes that Watson’s prior conviction was for possession with intent to distribute cocaine rather than cocaine base, and that it occurred seven years before his arrest in this case. Neither circumstance bars admission of Watson’s prior conviction. See, e.g., United States v. Tomberlin, 130 F.3d 1318, 1319-21 (8th Cir.1997); United States v. Hernandez, 84 F.3d 931, 935 (7th Cir.1996); United States v. Broussard, 80 F.3d 1025, 1040 (5th Cir.1996); Mitchell, 49 F.3d at 775-77.
. I assume that the court is not arguing that the prior crimes evidence here is “remotely probative” merely because it is inadmissible to prove the issue of possession directly, since prior crimes evidence is never admissible for that purpose. See Fed. R.Crim. P. 404(b).