concurring in part, dissenting in part, and dissenting from the judgment:
On no fewer than eight occasions during its closing and rebuttal arguments, the *206government misstated the legal standard governing the key issue disputed by the parties at trial — whether the genuine and allegedly counterfeit marks were “substantially indistinguishable.” Six of these misstatements occurred after the district court sustained the first of four defense objections to the remarks. Nevertheless, the district court denied Lam and Chan’s Rule 33 motion for a new trial, concluding that, although the statements were prejudicial, the court’s curative instruction negated the prejudice. Because I believe that these statements deprived Appellants of a fair trial, and because I think the district court abused its discretion in concluding they did not, I respectfully dissent in part from the majority opinion.
I.
To establish that a defendant has trafficked in counterfeit goods, in violation of 18 U.S.C. § 2320, the government must prove that, in connection with his goods, the defendant used a “spurious mark ... that is identical with, or substantially indistinguishable from” a registered trademark and that the use of this mark was “likely to cause confusion, to cause mistake, or to deceive.” Id. § 2320(f)(1)(A). At trial, Lam and Chan did not contest that they engaged in the conduct alleged by the government — that is, that they imported handbags bearing the “Marco mark.” Rather, the central issue disputed by the parties was whether Appellants’ Marco mark was “substantially indistinguishable” from the registered Burberry Check mark.
Yet during closing and rebuttal arguments, government counsel repeatedly misstated the standard governing this key issue. He instructed the jurors to assume the perspective of the average consumer who failed to conduct a side-by-side examination of the marks, specifically directing them to adopt the view of two government witnesses in place of their own conclusions regarding the marks. Despite numerous defense objections sustained by the district court, the prosecutor continued to assert this erroneous legal standard.
A.
Government counsel first addressed the issue of the standard for evaluating whether the marks were substantially indistinguishable in his closing argument. That determination, the prosecutor informed the jury, should be made “from the standpoint of the average purchaser.” J.A. at 1426. He went on to tell the jurors that they no longer occupied such a position, stating the following:
Now, like you all, after spending more than a week in this courtroom, we all probably know a whole lot more about these marks than the average purchaser does. All of us do. I mean, after all, how long do you think an average purchaser spends looking at a mark on a handbag or wallet to determine whether that mark is substantially indistinguishable to a registered mark that they know and remember before deciding to make the purchase or move on to something else? How long do you think they spend? 30 seconds? A minute? It is not going to be a lot of time, certainly not as long as the defendants have asked everyone in the Court to stare and look at them.
Id. at 1427.
The prosecutor then urged that, instead of relying on their own examinations of the marks, jurors should adopt the perspective of two government witnesses, Beatrice Bartko and Francis Golden Abel, consumers who had purchased Lam and Chan’s handbags believing they were authentic Burberry products. When moving to call Bartko and Abel as witnesses, the govern*207ment asserted that their testimony would be “relevant to [the] material issue” of “likelihood of confusion,” id. at 892, and the district court admitted their testimony solely for this purpose, id. at 640-41. See also United States v. Lam, Action No. 3:07-CR-374, 2010 WL 5178839, at *5 (E.D.Va. Dec. 14, 2010) (“The Government called Bartko and Abel as witnesses for the likelihood of confusion element of the counterfeiting claim only.... The Government was not allowed, however, to use these witnesses as evidence that the marks were substantially indistinguishable because they were not called for that purpose.”).
As already noted, however, in his closing argument, after cautioning the jurors that they no longer possessed the perspective of an average consumer, the prosecutor told them to “put[] [them]selves in the shoes of average purchasers on the street, purchasers like Bea Bartko, purchasers like Fran Abel.” J.A. at 1427-28. Bartko and Abel, he recounted, compared the “bag they [were] about to buy with what they remember[ed] ... of what the genuine mark look[ed] like,” without conducting a side-by-side comparison. Id. at 1428-29.
At that time, defense counsel did not object to the government’s comments. But during his closing argument, Chan’s counsel stated that the government was “confusing] two issues”: whether “the bags confuse somebody if they saw them” and “whether they are substantially indistinguishable or not.” Id. at 1482. By mingling these issues, defense counsel noted, the government was “getfting the jury] away from ... comparing the marks and making [its] own determination” regarding distinguishability. Id.
On rebuttal, government counsel once again asserted the jurors should judge whether the marks “are substantially indistinguishable from the perspective of an average person, an average consumer.” Id. at 1536. Appellants objected to this statement, and the district court sustained the objection. Nevertheless, the government immediately continued, “But what the government suggests, what the government submits, is this: That the evidence has proven that these bags are substantially indistinguishable from authentic marks owned by Burberry and Gucci to the average consumer.” Id. at 1537. Again, Appellants objected. The district court sustained the objection on the ground that the government was “conflating two issues,” but the court instructed the government, “Go ahead and make your argument. I will clear it up.” Id.
Shortly thereafter, the government again argued that the determination of whether the marks were substantially indistinguishable should be made “from the perspective of the average, ordinary consumer.” Id. at 1540; see also id. (“[T]he test is an average consumer.”); id. at 1541 (“This case is about assessing the evidence from the perspective of an average person, an average consumer.”). And the prosecutor further admonished the jury that they — unlike Bartko and Abel — no longer enjoyed that perspective. He stated the following:
And ladies and gentlemen of the jury, you have a difficult task. You have a difficult task, because you are supposed to assess these facts and look at these Burberry knock-offs ... from the standpoint of an average consumer. You have been in this case listening to it for a week. You have already developed a level of expertise and a perception of detail that I would argue is way above the average consumer. The average consumer includes people like Beatrice Bartko, someone who was familiar with the Burberry bags from shopping, from seeing it in catalogs, who wanted to add *208to her collection. She saw a photograph of a Marco bag on the Internet and purchased it and owned it for a year before determining, before learning that the bag was counterfeit. And she said that she thought it was a Burberry based on the pattern, based on the mark.... She thought it was substantially indistinguishable....
Another average consumer, an ordinary consumer, someone who didn’t have the benefit of a week-long trial as you have, was Frances Golden Abel.... She bought this bag and for some time she thought it was a real Burberry. She did it based on her familiarity with the pattern.... She thought this was a real Burberry. That is your evidence of an ordinary consumer.... And I submit, the government submits, that this is evidence both of substantially indistinguishable and the likelihood of confusion.
Now, [Chan’s defense counsel] in his closing arguments basically said you have to look at the differences. Sure, you do look at the differences. But you start with the similarities. That’s what the ordinary consumer does. The ordinary consumer sees things and sees the similarities and is drawn to it based on the similarities.... There are so many similarities, they are substantially indistinguishable to the ordinary purchaser.
Id. at 1541-43. The district court overruled defense counsels’ objection to this statement without explanation. The district court went on to sustain one additional objection to the government’s instruction that the jury should “[k]eep [its] eye on the perspective of the average person when [it is] looking at substantially indistinguishable.” Id. at 1558-59.
In total, the government told the jury to adopt the perspective of the average consumer on no fewer than eight distinct occasions during closing arguments. By my count, six of these comments occurred after the first sustained objection. The defense entered four objections, three of which the district court sustained. Yet the government carried on in making the erroneous statements and never sought clarification from the district court regarding the nature of its error.
B.
As the majority notes, the district court provided several instructions that, it concluded, cured the prejudice caused by the government’s statements. First, the district court instructed the jurors generally that they should listen to the court, not the lawyers, regarding “what the law is.” Id. at 1559-60. The district court then directed the jury to determine whether the marks were substantially indistinguishable as follows:
[Y]ou have to determine whether or not the mark that is alleged to be counterfeit is identical to or substantially indistinguishable from the [registered] mark.... Now, in order to carry out your responsibility as it relates to that part of these instructions, you have to compare the marks, the mark ... alleged to be counterfeit and the mark that is the genuine mark.
Now, you do that and you make a decision. This is based on your side-by-side comparison, use of your own eyes, and any other evidence that came into the record that might help you in that task. But it is for you to decide----If you decide that it is substantially indistinguishable, then you have another task to deal with, and that task has to do with now that you have decided that it is counterfeit under the definition that I have given you, you have to decide whether or not this counterfeit mark is a mark the use of which is likely to cause confusion or cause mistake or to deceive.
*209Everybody understanding that? It is two separate things. Now, a lot of that got clouded up in some of these arguments and some of the discussion. What must be counterfeit is the mark itself, not the goods.
Id. at 1587-88.
The jury found Lam and Chan guilty on all counts relating to the Burberry mark.
C.
Following their convictions, Lam and Chan filed a motion under Federal Rule of Criminal Procedure 33 seeking a new trial on the ground that the government’s improper statements constituted reversible prosecutorial misconduct. The district court concluded that the government’s remarks were both improper and prejudicial. Lam, 2010 WL 5178839, at *4-9. Nevertheless, it denied Appellants’ motion because it found that its instructions to the jury cured this prejudice. Id. at *9.
II.
On appeal, we review the district court’s denial of Lam and Chan’s motion for a new trial for abuse of discretion. See United States v. Ibisevic, 675 F.3d 342, 348-49 (4th Cir.2012). And our review of the district court’s factual findings regarding a claim of prosecutorial misconduct is for clear error. United States v. Ellis, 121 F.3d 908, 927 (4th Cir.1997).
“A prosecutor’s improper closing argument may ‘so infect[ ] the trial with unfairness as to make the resulting conviction a denial of due process.’ ” United States v. Lighty, 616 F.3d 321, 359 (4th Cir.2010) (quoting United States v. Wilson, 135 F.3d 291, 297 (4th Cir.1998)). To determine whether a prosecutor’s remarks require reversal, we consider “(1) whether the remarks were, in fact, improper, and, (2) if so, whether the improper remarks so prejudiced the defendant’s substantial rights that the defendant was denied a fair trial.”1 Id.
*210A.
The government first urges that, contrary to the district court’s conclusion, its statements were not improper and, as a result, this court need not review the district court’s finding with respect to prejudice. I disagree.
The district court found that the prosecutor’s remarks were improper in two ways. First, “[t]he [gjovernment’s comments regarding the standard for evaluating whether two marks are substantially indistinguishable were erroneous” because, in instructing the jurors that this determination should be made from the perspective of the average consumer, the prosecutor told the jury that they should not rely on their own visual examination to make this factual finding. Lam, 2010 WL 5178839, at *5. Second, the government improperly cited Bartko’s and Abel’s testimony to establish whether an average consumer would find the marks substantially indistinguishable. This testimony, however, could be used to prove only the likelihood-of-confusion element, as that was the sole purpose for which these witnesses were called. “Thus, even if an ‘average consumer’ analysis was relevant to the substantially indistinguishable element, the [gjovernment presented no evidence that the average consumer would have thought the marks were substantially indistinguishable.” Id. I agree with the district court on both points.
It was clearly within the province of the jury, as fact finder, to decide whether the Marco and Burberry marks were substantially indistinguishable. The government nevertheless argues that its statements were correct in light of Montres Rolex, S.A. v. Snyder, 718 F.2d 524 (2d Cir.1983), and Pepe (U.K.) Ltd. v. Ocean View Factory Outlet Corp., 770 F.Supp. 754 (D.P.R.1991). See Montres Rolex, 718 F.2d at 533 (holding that the determination of whether two marks were substantially indistinguishable should be made from the standpoint of an average purchaser, rather than an expert); Pepe, 770 F.Supp. at 758 (discussing Montres Rolex). The government failed to recognize, however, that both of these cases arose in the context of civil forfeitures or seizures, without a jury present to make the factual determination. *211See Montres Rolex, 718 F.2d at 527; Pepe, 770 F.Supp. at 755-56. And it is clear that the Montres Rolex court was not holding that a side-by-side examination of the contested marks was inappropriate, but rather was rejecting a rule that the court must adopt the view of an expert who, aided by specialized knowledge and equipment, is able to identify differences between marks that are not discernible to an average purchaser. See Montres Rolex, 718 F.2d at 531 (rejecting a test for the substantially indistinguishable element that depends upon an expert’s ability to discern differences “aided by a loupe or microscope” that “would be unnoticed or even undetectable by the average purchaser”).
Both the Second Circuit, in Montres Rolex, and the District of Puerto Rico, in Pepe, conducted visual comparisons of the goods at issue, rather than relying on the views of individuals who merely recalled the general appearance of the registered trademarks. See Montres Rolex, 718 F.2d at 533 (“We examined the actual bracelets at oral argument....”); Pepe, 770 F.Supp. at 759 (“Using the ‘ocular test’ of direct comparison of the defendants’ goods with the genuine[,] the Blantor defendants’ use of the mark PEPE is use of a counterfeit mark.”) (citing Fimab-Finanziaria Maglificio Biellese Fratelli Fila S.p.A. v. Helio Import/Export, Inc., 601 F.Supp. 1 (S.D.Fla.1983)). Use of the “ocular test” is particularly apt where a jury is available to undertake the comparison. Thus, the government erred in instructing the jurors that they should not conduct a visual comparison and that, after hearing evidence at trial, they were no longer positioned to make the requisite factual determination.
In addition, the district court correctly found that the government erred in directing the jury to adopt the views of Bartko and Abel on this issue. The district court allowed these witnesses’ testimony because it was relevant to the likelihood-of-confusion issue.2 Yet the government repeatedly and extensively invoked their testimony for the purpose of establishing that the marks were substantially indistinguishable. This was, as the district court found, improper.
I must next consider the district court’s determination regarding whether the improper remarks caused prejudice depriving Lam and Chan of a fair trial.
B.
We have identified the following factors as relevant to the prejudice inquiry:
“(1) the degree to which the prosecutor’s remarks have a tendency to mislead the jury and to prejudice the accused; (2) whether the remarks were isolated or extensive; (3) absent the remarks, the strength of competent proof introduced to establish the guilt of the accused; and (4) whether the comments were deliberately placed before the jury to divert attention to extraneous matters.” We also consider (5) whether the prosecutor’s remarks were invited by improper conduct of defense counsel, and (6) whether curative instructions were given to the jury.
Lighty, 616 F.3d at 361 (citations omitted) (quoting United States v. Adam, 70 F.3d 776, 780 (4th Cir.1995)). No single factor is dispositive. Id. Here, the district court found that factors one, two, and five favored granting a new trial; factor three was mixed; and factors four and six fa*212vored the government. See Lam, 2010 WL 5178839, at *6-11.
I agree with the district court that factor one, the tendency of the prosecutor’s remarks to mislead the jury and prejudice the defendant; factor two, the extensiveness of the remarks; and factor five, whether Appellants invited the remarks, favor Lam and Chan. The remarks were highly prejudicial. Whether the marks were substantially indistinguishable was the central issue disputed by the parties at trial. The government not only misstated the relevant legal standard but also suggested that the jurors should adopt the view of government witnesses in place of their own. Accordingly, the government’s statements were highly likely to mislead the jury.
Moreover, the remarks were extensive, particularly during the government’s rebuttal argument, in which the prosecutor cited the average-consumer standard no fewer than seven times. The prosecutor also spoke at length on two occasions about why he believed the jurors had become too sophisticated to judge the evidence correctly and should, instead, adopt the perspective of Bartko and Abel. The improper remarks pervaded the government’s argument, and the district court correctly found they were extensive. I also agree with the district court that Lam’s and Chan’s counsel did nothing to invite these remarks.
As to the third factor, the strength of the government’s proof absent the remarks, the district court found “that the [government did not present a wealth of other evidence to prove the substantially indistinguishable element of a counterfeiting charge,” but that the government presented some evidence showing that Lam and Chan “knew they were culpable.” Id. at *8. Although the district court identified only a limited body of evidence of guilt, I accept that this finding was not clearly erroneous.3
I do find, however, that the district court clearly erred in concluding that the prosecutor’s statements were not deliberate. In my view, factor four must be found to favor Appellants. The district court found that the prosecutor did not deliberately seek to mislead the jury because he genuinely believed that he had set forth the correct standard on the “substantially indistinguishable” element. Where, however, a prosecutor continues to make an incorrect statement of the law after the district court has sustained repeated objections to that statement, notifying the prosecutor of the error, the court should, in my opinion, infer some degree of deliberateness. To be clear, I do not disturb the district court’s finding that, when the prosecutor began to speak, he believed that his statements accurately reflected the law. But a prosecutor’s refusal to acknowledge his error after multiple sustained objections should be viewed as deliberate, and a court should deduce from *213his decision to soldier on a deliberate attempt to mislead the jury.
Likewise, the district court erred in failing to make a separate determination of deliberateness with respect to the prosecutor’s references to Bartko’s and Abel’s testimony. The government based its motion to admit this testimony on its relevance to likelihood of confusion, and the prosecutor’s plain disregard of the narrow purpose for which this evidence was admitted further evinces the deliberate nature of his comments.
The district court concluded, after considering the first five factors, that the prosecutor’s statements were prejudicial. I agree, but contrary to the district court, I believe that factor four also militates in favor of prejudice.
C.
Despite finding that the government’s remarks were prejudicial, the district court held that its curative instructions were sufficient to cure the prejudice. The majority, too, has treated this factor as determinative. I cannot share their view that the district court’s instructions dispelled the prejudice and ensured Lam and Chan received a fair trial.
The district court noted that it provided general curative instructions that the jury should follow the directions of the court, rather than the lawyers, as to the applicable law. In light of the pervasive and highly misleading nature of the prosecutor’s comments, I have no difficulty concluding that these general instructions were insufficient to cure the prejudice. See Wilson, 135 F.3d at 302 (finding that general curative instruction was insufficient to cause the jury to disregard the prosecutor’s improper statement in that case).
Furthermore, contrary to the district court’s finding, I cannot agree that the court’s specific instruction on the “substantially indistinguishable” element was sufficient to dissipate the prejudicial effect of the prosecutor’s remarks. The district court did tell the jury to compare the allegedly counterfeit mark with the Burberry mark. J.A. at 1587. But it went on to instruct the jury that it should base its determination of this element not only on the jurors’ “side-by-side comparison” of the marks, but also on “any other evidence that came into the record.” Id.
Thus, the district court allowed the jurors to rely on any and all evidence introduced at trial and failed to focus the jurors on the “ocular test.” Most notably, at no time did the district court limit the jurors’ use of Bartko’s and Abel’s testimony or instruct them that Bartko and Abel had not conducted the requisite visual comparison of the marks.4 Nor did the district court disabuse the jurors of the notion that they had gathered too much expertise to make the indistinguishability determination based on their own perceptions.
The district court’s instructions, therefore, failed to counter several of the serious errors that the court had identified in the prosecutor’s statements. And I do not think that the district court was within the bounds of its proper discretion to find that the limited guidance it provided to the jury was sufficient to counter the strong evi*214dence of prejudice established by the remaining factors.
III.
Ultimately, considering all factors relevant to the determination of prejudice, I am compelled to conclude that the government’s highly misleading and erroneous statements regarding a central issue disputed at trial prejudiced Lam and Chan, depriving them of the fair trial to which they are entitled. Moreover, I find that the district court abused its discretion in finding that its instructions to the jury cured this prejudice. Accordingly, with all due respect to the district court and my esteemed colleagues, I dissent.
. In setting out the standard for a Rule 33 motion, the district court stated that a court should grant a new trial "only ... if 'the evidence weighs so heavily against the verdict that it would be unjust to enter judgment.' " Lam, 2010 WL 5178839, at *2 (quoting United States v. Arrington, 757 F.2d 1484, 1486 (4th Cir.1985)). The majority, quoting United States v. Perry, 335 F.3d 316, 320 (4th Cir.2003), repeats this language. Ante at 203. Perry, in turn, is quoting United States v. Wilson, 118 F.3d 228, 237 (4th Cir.1997), which itself quotes Arlington. A review of Wilson and Arrington, however, reveals that they are addressing motions "based on the weight of the evidence,” where, of course, the jury verdict should be upheld absent very strong evidence to the contrary. See Wilson, 118 F.3d at 237 (“We have held that a district court should exercise its discretion to grant a new trial ‘sparingly’ and that the district court should grant a new trial based on the weight of the evidence 'only when the evidence weighs heavily against the verdict.’ ” (emphasis added) (quoting Arrington, 757 F.2d at I486)); Arrington, 757 F.2d at 1485-86 (considering a Rule 33 motion based on the weight of the evidence). Perry involved a Rule 33 motion based on the weight of the evidence and on an inconsistent verdict. See Perry, 335 F.3d at 322-23 & n. 12.
We have found that a district court abused its discretion in denying a motion for a new trial on other bases without imposing a separate requirement that the defendant demonstrate the evidence weighs heavily against conviction. See Ibisevic, 675 F.3d at 353-55 (finding that the district court abused its discretion denying the defendant’s motion for a new trial on the ground of improperly excluded evidence because in applying harmless-error analysis, regardless of whether the government's evidence was sufficient to convict, the court could not say with fair assurance that the error had not substantially swayed the judgment). And, of course, "some constitutional errors ... are so fundamental and pervasive” that a new trial should be granted without regard for any other facts or circumstances. Delaware v. Van Arsdall, 475 U.S. 673, 681, 106 S.Ct. 1431, 89 L.Ed.2d 674 (1986).
*210In the context of prosecutorial misconduct, we have held that a defendant may obtain a new trial by demonstrating that the government made improper remarks that prejudicially affected his substantial rights so as to deprive him of a fair trial. See United States v. Higgs, 353 F.3d 281, 330 (4th Cir.2003). Although we have identified the strength of the government’s proof absent the improper remarks as one of several factors relevant to the determination of prejudice, we have been careful to note that no single factor is dispositive of the prejudice inquiry. See Lighty, 616 F.3d at 361. Indeed, we have not hesitated to grant a new trial on the basis of prosecutorial misconduct where a defendant has established prejudice, even if the government’s case is strong. See, e.g., Wilson, 135 F.3d at 302 (remanding for a new trial based on the prosecutor’s improper, prejudicial statements despite "fully recognizing that the drug trafficking evidence ... was strong”).
I do not believe that the weight of the evidence becomes a separate, dispositive factor in a Rule 33 motion based on prosecutorial misconduct. Although the Constitution does not guarantee a perfect trial, a defendant is entitled to a fair one. See Lighty, 616 F.3d at 336. Where he has been deprived of a fair trial by virtue of a prosecutor's improper remarks causing prejudice, I think our precedent establishes that a new trial is warranted in the interest of justice. It appears that neither the majority nor the district court has, in fact, treated the weight of the evidence as a separate element necessary for the granting of a new trial. Accordingly, I think this statement is properly read to require only that the facts establish the prosecutor’s improper statements caused prejudice based on the relevant factors, which include but are not limited to the strength of the government's case.
. Likelihood of confusion is properly adjudged from the perspective of the average purchaser based on the conditions encountered in the marketplace, where a close, side-by-side visual inspection of the goods is infeasible. See J. Thomas McCarthy, McCarthy on Trademarks and Unfair Competition § 23:58 (4th ed. 2011).
. This finding, of course, is not dispositive. In prior cases, we have not hesitated to order new trials due to improper remarks by prosecutors that cause prejudice, even where the government presented sufficient or strong evidence of guilt. Wilson, 135 F.3d at 302 (remanding for a new trial due to the prosecutor's improper and prejudicial statements during closing arguments even while "fully recognizing that the drug trafficking evidence against [the defendant] was strong”); United States v. Mitchell, 1 F.3d 235, 242 (4th Cir.1993) (finding that the prosecution introduced sufficient evidence other than the improper remarks to support a conviction, but nevertheless finding prejudice requiring reversal). Here, the government's case can withstand a sufficiency of the evidence challenge, but it was hardly overwhelming. Indeed, a prior trial of Appellants on these charges resulted in a mistrial due to a hung jury.
. As evidence that its instruction was sufficient to cure the prejudice, the district court cited the fact that the jury convicted Appellants of counterfeiting the Burberry Check mark but not a Gucci trademark. Lam, 2010 WL 5178839, at *11. It is notable, then, that Bartko and Abel testified only as to their confusion regarding goods bearing the Burberry mark, and the prosecutor presented and cited no similar evidence regarding the Gucci mark.