dissenting.
I would hold that the district court’s decision to exclude Dr. Fulero’s testimony should be affirmed on the basis of Smith-ers’s delay in proffering it in its specifics to the court and Government. If we are to reach the merits of the decision, however, I am not nearly so certain as the majority is that the court did not perform the proper legal analysis. Certainly we should make that decision on the basis of a review of the entire record and not, as does the majority, largely on the basis of a handful of unfortunate but irrelevant remarks by the district court. In any event, once we have decided, as the majority has, that the court did not perform the proper Daubert analysis, our response should be to remand the issue for a proper hearing. We should not proceed to do that analysis ourselves, nor should we issue what is essentially a blanket endorsement of expert testimony on a subject deserving of, at best, our careful and skeptical scrutiny, effectively warning the district courts in this circuit that in the future it will be an abuse of discretion not to accept such experts. For these reasons, I must dissent.
I. Delay
As the majority noted, the district court’s primary reason for denying the renewed motion to permit Fulero to testify was that it was made “too late in the day.” In reasoning that Smithers’s initial motion in limine put the Government on sufficient notice of Fulero’s testimony, the majority makes no mention of the paucity of detail which that motion contained. Furthermore, the legal foundation of the majority’s reasoning is, in my view, erroneous.
A brief overview of the appellate courts’ reception of expert testimony on the fallibility of eyewitness identifications is necessary in order to explain the inadequacy of Smithers’s initial motion. The majority correctly observes that for approximately the first decade or so in which such testimony was submitted, courts were “uniformly skeptical ... for a host of reasons.” These reasons included distrust of the science behind the testimony, a concern that the majority goes to considerable lengths to dispel. But this was hardly the only reason given for disallowing the testimony, and that skepticism rightly continues in the appellate courts today. The majority opinion in this case acknowledges some of these decisions, but sidesteps the unanimous hesitancy among appellate courts to open the door too far to this testimony. In many cases, the excluded testimony is either a generic, scholarly exploration of psychological theory, bearing little relation to the facts of the particular case, see, e.g., United States v. Brien, 59 F.3d 274, 277 (1st Cir.1995); United States v. Rincon, 28 F.3d 921, 925 (9th Cir.1994); Jordan v. Ducharme, 983 F.2d 933, 939 (9th Cir.*3191993); United States v. Blade, 811 F.2d 461, 464-65 (8th Cir.1987); United States v. Posher, 590 F.2d 381, 382-83 (1st Cir.1979), or else so specifically directed at the validity of a particular witness’s testimony as to usurp the jury’s role in determining credibility, see, e.g., United States v. Lumpkin, 192 F.3d 280, 289 (2d Cir.1999); United States v. Hall, 165 F.3d 1095, 1107 (7th Cir.1999); United States v. Kime, 99 F.3d 870, 884 (8th Cir.1996); United States v. Dorsey, 45 F.3d 809, 812 (4th Cir.1995); United States v. Moore, 786 F.2d 1308, 1311-12 (5th Cir.1986); United States v. Langford, 802 F.2d 1176, 1179 (9th Cir.1986); State v. Gaines, 260 Kan. 752, 926 P.2d 641, 645 (1996); State v. Sabetta, 680 A.2d 927, 933 (R.I.1996). In either situation, even though the testimony may have provided some measure of insight that the jury otherwise would not have possessed, the risk of the jury’s being unduly swayed by testimony with the imprimatur of scientific expertise has been deemed significant enough that the decision to exclude it could not be considered an abuse of the trial court’s considerable discretion with regard to evidentiary matters.. This is especially so in light of the fact that the more traditional methods of exposing the fallibility of eyewitness identifications' — -cross-examination, jury instruction and closing argument — are more efficacious and far less risky than expert testimony that can at best be only marginally relevant to the facts at hand. See Moore v. Tate, 882 F.2d 1107, 1110-11 (6th Cir.1989); Hall, 165 F.3d at 1107; United States v. Smith, 122 F.3d 1355, 1358-59 (11th Cir.1997); United States v. Hicks, 103 F.3d 837, 847 (9th Cir.1996); Kime, 99 F.3d at 884; United States v. Ginn, 87 F.3d 367, 370 (9th Cir.1996); Rincon, 28 F.3d at 925-26; Jordan, 983 F.2d at 938-39; United States v. Curry, 977 F.2d 1042, 1051 (7th Cir.1992); Blade, 811 F.2d at 464-65; Moore, 786 F.2d at 1311-12; Fosher, 590 F.2d at 382; State v. McClendon, 248 Conn. 572, 730 A.2d 1107, 1115-16 (1999); McMullen v. State, 714 So.2d 368, 370 (Fla.1998); Gaines, 926 P.2d at 642-43; State v. Buell, 22 Ohio St.3d 124, 489 N.E.2d 795, 803-04 (1986); Currie v. Commomvealth, 30 Va.App. 58, 515 S.E.2d 335, 339 (1999). The grounds on which these courts have explained their rulings vary — the testimony was unhelpful, the subject was within the jury’s common knowledge, the subject was not a proper one for expert testimony under Evidence Rule 702 or some analogous test, or the prejudice substantially outweighed the probative value pursuant to Rule 403- — but the results were the same.
I will concede that the concept of expert testimony on the subject of eyewitness identification, and the scientific research behind the testimony, has gained some acceptance and respect in our courts since it was introduced. But the majority’s own recounting of the case law on this subject reveals that the appropriateness of using such testimony in court — instead of its traditional alternatives — to counteract the deficiencies of eyewitness identifications is still very much in controversy, for all of the reasons detailed above. The recent trend has been towards allowing the testimony in a limited number of “narrow circumstances,” but this merely reflects the liberality of Rule 702 and the gradual maturing of the research, not the “dramatic transformation” of judicial attitudes that the majority claims. See United States v. Smith, 156 F.3d 1046, 1052 (10th Cir.1998) (holding that cross-examination and common sense will presumptively suffice outside the “narrow circumstances [of] cross-racial identification, identification after a long delay, identification [ ... ] under stress, and [ ... ] the feedback factor and unconscious transference”); United States v. Harris, 995 F.2d 532, 535-36 (4th Cir.1993) (same); Currie, 515 S.E.2d at 338 (same); Brien, 59 F.3d at 277 (“a door once largely shut is now somewhat ajar”). Some of our sister circuits expressly retain their jaundiced view of this type of testimony. See Hall, 165 F.3d at 1104 (“This Court has a long line of cases which reflect our disfavor of expert testimony on the reliability of eyewitness identification”); *320Smith, 122 F.3d at 1357 (“This Court has consistently looked unfavorably on such testimony”). Every court to address the issue has left the admissibility of the testimony to the sound discretion of the district court on a case-by-case basis, either on the authority of Daubert, Rule 702, or an analogous state rule. No appellate court has adopted a presumption or per se rule in favor of admitting eyewitness identification expert testimony, something the majority’s opinion comes dangerously close to doing. Many courts have expressly disavowed such a rule. See Smith, 122 F.3d at 1359; United States v. Alexander, 816 F.2d 164, 169 (5th Cir.1987); Blade, 811 F.2d at 465; Langford, 802 F.2d at 1179; Sabetta, 680 A.2d at 933.
Moreover, the only federal appellate decisions finding the exclusion of this type of expert testimony to be an abuse of discretion are readily distinguishable from the instant case. In United States v. Stevens, 935 F.2d 1380, 1397 (3d Cir.1991), the Third Circuit reviewed a district court’s decision to admit the expert’s testimony as to some psychological theories but not others. The dangers of the expert’s testimony in general, then, were not at issue. The panel reversed because it found no reason why the excluded theories did not “fit” the facts of the ease as much as those that were admitted. In United States v. Downing, 753 F.2d 1224, 1242 (3d Cir.1985), the district court erroneously excluded the testimony per se instead of performing its gatekeeping function. In this circuit’s Smith decision, the Government conceded Dr. Fulero’s expertise, see 736 F.2d at 1105, and the proffer there specifically tied the theories of transference and cross-racial identification to the facts of that ease. See id. at 1106. We used this specificity to distinguish Fosher, which was representative of the cases finding expert testimony too removed from the particular facts to be helpful. See id. The lack of specificity in Smithers’s proffer likens this case to Fosher far more than to Smith. Apart from this distinction, the majority opinion’s characterization of Smith’s holding is troubling. At most, this court said in Smith that Dr. Fulero’s testimony on the reliability of eyewitness testimony might meet the Green criteria, and might have been improperly excluded. We did not, as the majority opinion claims, express our “acceptance of psychological studies as a scientifically sound and proper subject of expert testimony, noting, ‘[the] science of eyewitness perception has achieved the level of exactness, methodology and reliability of any psychological research.’ ” What we noted is that Dr. Fule-ro had testified to that effect, see id., and, in the final analysis, held that “even if it were error to exclude the expert’s testimony, such error was ‘harmless.’ ” Id. at 1106-07 (emphasis added). It is also worth noting that the analysis in Smith was not unanimous; the concurring judge did not find an abuse of discretion. Since the Smith case, no Sixth Circuit decision has reversed a district court’s exclusion of expert testimony on eyewitness identifications as an abuse of discretion.
I will address in a later segment of this dissent my view of this testimony’s utility, but for now it suffices to say that because the range of circumstances in which this testimony should be admitted is so narrow, the party offering it should be required as a threshold matter to make
an on-the-record detailed proffer to the court, including an explanation of precisely how the expert’s testimony is relevant to the eyewitness identifications under consideration. The offer of proof should establish the presence of factors (e.g., stress, or differences in race or age as between the eyewitness and the defendant) which have been found by researchers to impair the accuracy of eyewitness identifications.
Stevens, 935 F.2d at 1397 (quoting Downing, 753 F.2d at 1242).1 The Downing *321court remanded its case for a proper Rule 702 hearing on the proposed expert testimony, because the district court had merely held a brief sidebar on the issue on the tenth day of trial without a voir dire of the witness or any time for either party to present its view. See Downing, 753 F.2d at 1228.2 Here, however, the district court properly held a pretrial hearing on various motions in limine, including this one, but the content of Smithers’s supporting memorandum was woefully inadequate to enable the court to exercise its discretion in an informed manner. The 10-page supporting memorandum recited the applicable standards of Daubert and Rules 702 and 403, defended the legitimacy of Dr. Fulero’s field of study and academic qualifications, and included a few paragraphs indicting the reliability of eyewitness identifications in general. It contained absolutely no attempt to explain how the testimony would relate to the facts of the case or - which- of the psychological theories on memory3 (e.g., stress, “forgetting curve,” accuracy-confidence relationship, etc.) may -be applicable in the situation at hand. The memorandum’s attachments — Fulero’s vita and a selection of journal articles on the topic — plainly did nothing to provide the needed specificity. The Government made precisely this point in its response memorandum,4 and cited a number of authorities suggesting that cross-examination and jury-instructions were better alternatives. At the hearing, the court began the discussion by opining that “the government writes a pretty persuasive brief. You can argue to the jury people make mistakes all the time. You can bring out the discrepancies [through' cross-examination and a jury instruction].” Smithers responded by defending the scientific validity of the testimony. The court asked, “Has [the expert] rendered a report?” Receiving a negative response, the court continued: “I would *322have to go through a long voir dire ahead of time. I think if you’re going to have an expert you’ve got to have a report. You’ve given me his curriculum vitae.... I’d be happy to entertain [an instruction.]” It was after this exchange that Smithers asserted for the first time that Fulero would “testify to the specifics of the case and explain to the jury that there are scientific studies that have shown that eyewitness identification is flawed.” Smithers still did not, however, cite a specific theory or fact in the case to which these “scientific studies” would relate. The court then concluded that “[none of the cases] cited to me ... suggests that this is admissible evidence. The government’s brief is very persuasive, and I don’t have a report from the expert. No, I think ... you’re asking him to comment upon Debra White’s credibility.” (emphasis added). Smithers conceded the motion and asked permission to proffer the evidence. The court agreed, and although it offered several times to accept an oral proffer at that time, Smithers insisted on delivering it in writing.
It was in the written proffer, which was not filed until after the Government had rested its case and immediately before Smithers rested his, that Smithers first made any colorable attempt to tie Dr. Fulero’s testimony to the facts of the case. Smithers identified the stress of the robbery, “detail salience” relating to Smith-ers’s scar, the length of time between the robbery and the trial, the “conformity effect” of subsequently received information, the photo spread methodology,5 and the relationship between the witnesses’ confidence and accuracy as relevant subjects for Fulero’s testimony. Smithers also took issue with the adequacy of a jury instruction in counteracting the fallibility of eyewitness identifications. Smithers had made none of these arguments before this point in the proceedings, either orally or in writing, despite several opportunities to do so. It was in this context that the court held another hearing on the motion, and remarked, “you finally got your act together with this latest filing_ Much different from the first filing,” to which Smith-ers responded, “Admittedly, Your Honor.” A lengthier conversation on the merits of the testimony ensued, followed by the court’s decision to continue to exclude the testimony, primarily because of the delay.
Federal Rule of Evidence 403 permits relevant evidence to be excluded “if its probative value is substantially outweighed by the danger of unfair prejudice, confusion of the issues, delay, waste of time, or needless presentation of cumulative evidence.” (emphasis added). A district court has “very broad” discretion in making this determination. See United States v. Hawkins, 969 F.2d 169, 174 (6th Cir.1992). A Daubert analysis includes a consideration of Rule 403, see Daubert, 509 U.S. at 595, 113 S.Ct. 2786; Rincon, 28 F.3d at 925, and several courts have held that Rule 702’s “helpfulness” inquiry incorporates Rule 403’s concern for undue prejudice. See Hall, 165 F.3d at 1104; Kime, 99 F.3d at 884; Curry, 977 F.2d at 1051. The district court was well within its discretion to refuse to require the Government to prepare a response to an expert witness when the first inkling of what the witness would testify to was not given to the Government until the middle of the trial, after the Government had rested its own case. There is no basis for the majority’s holding that Smithers’s initial motion — which did little more than introduce Dr. Fulero and his field of study — or his renewed motion at the start of trial, or his mid-trial brief, put the Government on sufficient notice of the substance or foundation of Fulero’s testimony so as to permit the Government to prepare a rebuttal, either to the motions or the testimony. The consequences of Smithers’s procrastination should rest on him, not on the Government. Other courts have held that initial notice of the intent to *323call an eyewitness identification expert witness only a few days before trial is grounds for exclusion. See Dorsey, 45 F.3d at 816 (remarking “the case law is clear that it is not an abuse of discretion ... to disallow expert testimony where a late proffer of evidence by the defense substantially prejudices the government in its ability to find its own expert and conduct similar testing” and upholding exclusion when notice given on first day of trial); Curry, 977 F.2d at 1052 (upholding exclusion when 4 days notice given); United States v. Dowling, 855 F.2d 114, 118 (3d Cir.1988) (upholding exclusion when 5 days notice given in trial held in the Virgin Islands); see also Hon. Robert P. Murrian, The Admissibility of Expert Eyewitness Testimony Under the Federal Rules, 29 Cumb. L. Rev. 379, 395-96 (1998-99) (instructing practitioners that “The offer of proof should establish the factors in the particular case which call for expert testimony, such as the extreme stress of the witness, differences in age or race of the defendant and the eyewitness, and suggestive line up techniques. If the factors necessitating expert testimony are not established, and the court excludes the expert testimony, the decision will likely be upheld on appeal”). These decisions and commentary contradict the majority’s blanket statement that delay is “not a proper basis for exclusion.”6
It is important to note that the majority relies solely on United States v. Collins, No. 87-5077, 1988 WL 4434 (6th Cir. Jan.25, 1988) (per curiam) (unpublished), for the proposition that tardiness is not a proper basis for exclusion of expert testimony. This use of Collins is both misleading and inappropriate. In Collins, which is not only unpublished but is pre-Daubert, the defendant proffered a psychologist who would testify that the tendency to fill in gaps in perception made the eyewitness identifications in the case unreliable. See id. at **1. The district court excluded the witness for only one reason — he had not been listed as a witness as instructed by a pretrial order. No admissibility determination of any kind was made. The witness was therefore excluded solely to punish the defendant for noncompliance with a discovery order. It was in this context that the court made the statement quoted by the majority here: “a criminal defendant’s relevant evidence may generally not be excluded on the basis of a discovery sanction.” The Collins court expressly distinguished the case from one determining whether such evidence was admissible pursuant to our then-recent Smith decision. In fact, the Collins court followed Smith in declining to rule that the testimony was admissible as a matter of law, and proceeded to find the error harmless in light of other evidence. See id. at **2. Collins, then, is completely inapposite to this case, which involves an admissibility determination and not a discovery sanction. Moreover, reliance on unpublished cases in a subsequent written opinion for purposes other than establishing preclusion or law of the case, unless the prior case is truly of such precedential value that it probably should have been published, does violence to the policy we have promulgated in 6. Cir. R. 28(g). This dubious use of Collins will only have the unfortunate side effect of encouraging lawyers to cite other unpublished decisions to us in the future, despite the clear intent of the rule.
II. The District Court’s Application of Daubert
The majority finds that the district court abused its discretion by failing to apply the evidentiary gatekeeping principles of Dau-*324bert. I am not convinced that the court committed this error, or that remand would be necessary even if it did.
The majority pays passing obeisance to the abuse of discretion standard by which we review a district court’s decision to exclude expert testimony, but wholly fails to apply in this case the deference that standard requires. The factors listed in Daubert were meant to suggest to federal courts the relevant subjects of analysis when evaluating proffered experts under Rule 702, but they are “not holy writ” that the district court must invoke by name in order to pass our scrutiny. Kumho Tire Co. Ltd. v. Carmichael, 526 U.S. 137, 119 S.Ct. 1167, 1179, 143 L.Ed.2d 238 (1999) (Scalia, J., concurring). The Supreme Court has recently instructed that
The trial court must have the same kind of latitude in deciding how to test an expert’s reliability, and to decide whether or when special briefing or other proceedings are needed to investigate reliability, as it enjoys when it decides whether or not that expert’s relevant testimony is reliable. Our opinion in Joiner makes clear that a court of appeals is to apply an abuse-of-discretion standard when it reviews a trial court’s decision to admit or exclude expert testimony. That standard applies as much to the trial court’s decisions about how to determine reliability as to its ultimate conclusion. Otherwise, the trial judge would lack the discretionary authority needed both to avoid unnecessary “reliability” proceedings in ordinary cases where the reliability of an expert’s methods is properly taken for granted, and to require appropriate proceedings in the less usual or more complex cases where cause for questioning the expert’s reliability arises. Indeed, the Rules seek to avoid unjustifiable expense and delay as part of their search for truth and the just determination of proceedings. Thus, whether Daubert’s specific factors are, or are not, reasonable measures of reliability in a particular case is a matter that the law grants the trial judge broad latitude to determine. And the Eleventh Circuit erred insofar as it held to the contrary.
Kumho Tire, 119 S.Ct. at 1176 (internal citations, quotations and alterations omitted). The court’s failure specifically to cite Daubert as its basis for excluding Dr. Fulero does not itself mandate remand. See Greenwell v. Boatwright, 184 F.3d 492, 498 (6th Cir.1999) (“Although the trial court is not required to hold an actual hearing to comply with Daubert, the court is required to make an initial assessment of the relevance and reliability of the expert testimony. Because the district court did not hold a Daubert hearing we must review the record to determine whether the district court erred in its assessment of the relevance and reliability of the expert testimony”); see also Hall, 165 F.3d at 1102 (approving of the district court’s evaluation of the testimony in a hearing that did not explicitly cite the two Daubert prongs but frequently referenced the decision). Instead, our task is to review the district court’s performance of its gatek-eeping function in light of “the facts of [the] particular case,” Kumho Tire, 119 S.Ct. at 1175 (internal quotations omitted), granting “the trial judge [ ] considerable leeway in deciding in [this] particular case how to go about determining whether [this] particular expert testimony is reliable.” Id. at 1176. In so doing, we must be mindful of the principles behind Dau-bert, but “the factors it mentions do not constitute a definitive checklist or test.” Id. at 1175 (internal quotations omitted).
I would hold that the way in which the district court conducted its analysis of the admissibility of Dr. Fulero’s testimony was not abusive of the court’s discretion. The core holding of the Daubert decision was that admission of expert testimony is governed by the Federal Rules of Evidence and not the theretofore majority rule of “general acceptance” by the scientific community. See Daubert, 509 U.S. at 587, 113 S.Ct. 2786. The primary “locus” *325of the court’s power to evaluate experts rests in Rule 702. See id. at 589, 118 S.Ct. 2786. Rule 702 requires that the testimony be reliable and relevant to be admitted. Because the Government has chosen not to contest Dr. Fulero’s qualifications as a psychologist or the abstract scientific validity of the studies he proposes to testify from, either at trial or on appeal, we may assume that Smithers has satisfied the reliability requirement. See Greenwell, 184 F.3d at 498. Instead, the Government has consistently focused its challenge on the relevance aspect of Rule 702, which “further requires that the evidence or testimony assist the trier of fact to understand the evidence or determine a fact in issue. [ ... ] The consideration has been aptly described ... as one of ‘fit.’ ” “ ‘Fit’ is not always obvious, and scientific validity for one purpose is not necessarily scientific validity for other, unrelated purposes.” Daubert, 509 U.S. at 591, 113 S.Ct. 2786 (internal quotations omitted). The district court here examined briefs on the issue from both sides in preparation for the hearing on Smithers’s limine motion. Both briefs recited the applicable factors from Daubert and Rules 702 and 403. Both examined the leading cases on this type of testimony, both before and after Daubert, focusing especially on United States v. Rincon, 28 F.3d 921 (9th Cir.1994) and our decision in United States v. Smith, 736 F.2d 1103 (6th Cir.1984) (per curiam). The Government’s brief went further to explain why the testimony would be unhelpful and prejudicial in this case, and why cross-examination and jury instructions would better address Smithers’s concerns. The court demonstrated its reliance on these briefs when it began the relevant portion of the motion hearing by noting its belief that the Government’s arguments were persuasive, informing Smithers that he could discredit the eyewitness identifications through cross-examination, and asking for Smith-ers’s assistance in choosing an appropriate jury instruction. As recounted above, the court pressed Smithers for additional details on how Fulero’s testimony would relate to the facts of the case, but no.such details were forthcoming. This left the court with little basis upon which to conclude that -the Government was in error in its contention that the testimony would only confuse the jury and invade its province by commenting directly on the credibility of the witnesses. Whether Smithers was unable to demonstrate the relevance of the testimony at this hearing on the first day of trial or was simply procrastinating, the onus should fall on him; the court dealt appropriately with the information and' arguments presented to it.
The court also acted properly once Smithers — at the close of his defense— finally proffered the details of Fulero’s testimony. At one point during the hearing on the renewed motion to permit Fule-ro to testify, the court — addressing the prosecutor. — explained its reliance on Rin-con, and noted, consistent with the “reliability” element of Daubert, that “the good professor in his affidavit and in his background and in the literature that was cited to me suggests that the fragility ... of eyewitness testimony has been established scientifically and that he brings an expertise that may assist the jury.” How the majority can hold, in light of this statement and the Government’s decision not to challenge Fulero’s competence, that “the district court did not make any determination as to this expert’s scientific reasoning or methodology” is puzzling. And because reliability was never at issue, any further inquiry into the reliability of the testimony was unnecessary and, indeed, is precisely the kind of proceeding that Kumho Tire expressly gives the district courts the discretion to avoid.
The majority acknowledges that although the district court did not explicitly explain that it was doing so, it did conduct some inquiry into relevance when it decided that the jury was aware of its obligation to be skeptical of eyewitness testimony. The record of the second hearing, however, reveals that the district court in fact *326looked carefully at the issue of relevance. Even at this point, Smithers did not make Dr. Fulero available for voir dire by the Government, but the court initiated a lengthy discussion with Smithers’s counsel on Fulero’s familiarity with the facts of the case, including Smithers’s scar, the photo lineup procedure, and the stress of the robbery. These are precisely the questions the district court needed to ask to determine the relevance and “fit” of Fule-ro’s testimony to the particular facts of the case. After hearing Smithers’s answers, the court concluded that Fulero would have acted in this case as more of an advocate than a neutral, scientific expert— a characterization borrowed from Rincon. See 28 F.3d at 923. The majority fails to suggest any means whatever by which the court could have conducted a better inquiry under the circumstances. Instead it flatly pronounces that the court’s conclusion was “simply wrong” because it would lead to the “absurd” result of never allowing such expert testimony.
I suspect that the Seventh and Eleventh Circuits might take umbrage at the majority’s characterizing as “absurd,” their strong presumptions against expert testimony regarding eyewitness identifications, see Hall, 165 F.3d at 1103; Smith, 122 F.3d at 1357. More importantly, I think it is the majority’s conclusion that is simply wrong. The majority fails to explain how this extreme result would follow from the district court’s observation. Indeed, if the district court meant flatly to disallow expert testimony concerning eyewitness identifications, it would not have gone out of its way at this hearing to replace sua sponte the pattern jury instruction on eyewitness identifications to which Smithers had already agreed with what it saw as “a much stronger instruction” in order to alleviate the genuine concerns that Smithers had raised. The majority is resolute in its conviction that the district court failed to “apply Daubert,” but it fails to explain how that court could have done any better with no more information than Smithers provided.
This case presents very few of the “narrow circumstances” identified by other courts in which this kind of expert testimony can be relevant. See Smith, 156 F.3d 1046, 1052; Harris, 995 F.2d at 535-36. There was no problem of cross-racial identification. The passage of time has been found to be a relevant factor when the recalled event occurred forty years prior, see Krist v. Eli Lilly and Co., 897 F.2d 293, 297-98 (7th Cir.1990), but not when the time lapse was a “routine” one of “merely” six years. See Curry, 977 F.2d at 1052. Here, the time between the robbery and Ms. White’s positive identification of Smithers in the photo array was two days; the time between the robbery and the trial was only one and a half years. Moreover, although Smithers alleges that there was an unconscious transference of mistaken identifications among the witnesses, the court explicitly found that all the evidence presented at the hearing appeared to suggest otherwise.
Furthermore, the majority identifies its primary basis for finding an abuse of discretion as the court’s “experiment” comment, explaining that “[b]asing an eviden-tiary decision on personal curiosity rather than applicable case law and the rules of evidence is a patent abuse of discretion.” The fact that these offhand statements were made is unfortunate. We review them on the cold record, separated them from their context and texture, including the voice inflection and facial expressions of their delivery. But the proceedings described above make it clear that the district court did not base its exclusion of Fulero on the sinister whimsy that the majority imputes to it. The statements were made at the close of the second hearing, after the court had again denied the motion and instead awarded Smithers a strongly worded instruction. The court’s comment to Smithers’s counsel that she had “made an excellent record that I’ve abused my discretion” was not indifference to the law, but an assurance that she had *327done well in preserving a record of her objection for appeal. The observation that admitting Fulero’s testimony would have been “tantamount to the Court declaring the defendant not guilty as a matter of law” and that “absent the eyewitness testimony I don’t think there’s enough here to go to the jury” correctly describes the severely prejudicial effect that Fulero’s testimony likely would have had on the Government’s case. Finally, the court’s “experiment” remark, while perhaps inappropriate, was made well after the motion had twice been denied and was the last statement made on the record before Smithers rested his case. It did not form the basis for the court’s exclusion of Fule-ro, nor did it prejudice Smithers in any other way. I do not agree that this single comment can justify the majority’s finding of a patent abuse of discretion.
The majority ultimately concludes that this case must be remanded for a new trial that, presumably, will include “a Daubert test,”7 whatever that may be. Were that the extent of our holding, my difference of opinion with the majority would simply be a disagreement about what Daubert requires and how the district court should have proceeded here. But the majority does not stop there. Instead, it proceeds into a lengthy explanation of what the court might have found had it applied Daubert to the majority’s liking. This, in my view, is wholly improper. Not only does this exceed our function as an appellate court, but it is anathema to the law that the majority had theretofore laid out; if the gatekeeping function is truly in the district court’s discretion and requires a fact-finding hearing, and the district court in this case has failed to exercise that discretion as utterly as the majority concludes, then surely the record before us is inadequate to permit us to announce what facts about the testimony the court would have discovered in- a hearing. Instead, we should follow the lead of our sister circuits which, upon a finding that the district court has not assessed an eyewitness identification expert’s relevance in a manner consistent with Rule 702, have remanded the matter without further discussion. See, e.g., Hall, 165 F.3d at 1102; United States v. Amador-Galvan, 9 F.3d 1414, 1417-18 (9th Cir.1993); Downing, 753 F.2d at 1226. I take some comfort in the fact that the majority’s sua sponte application of Daubert and glowing praise for eyewitness identification expert testimony are dicta, .since they exceed the actual holding that the court abused its discretion. To the extent, however, that the opinion as a whole is seen as persuasive authority cementing the already-extant impression that our circuit is among the most receptive to this type of testimony, see Murrian, supra, at 392, it does our jurisprudence a disservice.
III. The Merit of Expert Testimony on Eyewitness Identifications
The trepidation with which nearly all appellate courts have treated this subject is representative of a broader reluctance, which I share, to admit the expert testimony of social scientists with the same deference given to the testimony of those in the physical sciences. I do not seek to discredit the value of these researchers’ work; the ever-expanding psychological disciplines have done much in the past several decades to explode commonly held misconceptions and enrich our understanding of human behaviors. As even those courts most opposed to admitting the testimony in court have acknowledged, those benefits include an enhanced insight into the fallibility of eyewitness identification that can inform our trial procedures. See, *328e.g., Hall, 165 F.3d at 1104. The difficulty arises in treating psychological theories as if they were as demonstrably reliable as the laws of physics. Conclusions reached by applying the laws of all but the most theoretical of physical sciences to a particular set of facts are verifiable through replication; disagreements between dueling experts in the physical sciences (e.g., accident reconstructionists or DNA experts) typically focus on the data to which the scientific method is applied, which is subject to objective analysis. The certainty of the testimony of social scientists, however, is limited by the nature of their field. They typically base their opinions on studies of small groups of people under laboratory conditions; those studies are then interpreted and extrapolated to predict the likelihood that another person under similar but non-controlled conditions will manifest similar behavior. Each step of this analysis — the choice of sample and control groups, the conditions under which they are observed, the cause and nature of the observed behavior, and the likelihood that the observed behavior will be replicated by a different person in a non-controlled setting — is influenced by the personal opinion of the individual expert. Nor will there be much similarity between the persons typically studied by social scientists and the witnesses in any given criminal trial. The studies are virtually always based on college students or other readily available test subjects in a controlled environment (which are the most easily measurable), not individuals involved in real world incidents such as actual robbery victims. See, e.g., United States v. Hines, 55 F.Supp.2d 62, 72 (D.Mass.1999) (assessing relevance .of studies of college students); Brian L. Cutler and Steven D. Penrod, Assessing the Accuracy of EyeWitness Identifications, in Handbook of Psychology in Legal Contexts 193 (R. Bull and D. Carson ed. 1995) (Attachment E to Smithers’s Motion in Limine) (“Most of what is known about the psychology of eye-witness memory has been acquired through laboratory experiments”). The limits of social science testimony were aptly expressed in Gacy v. Welborn, 994 F.2d 305, 313-14 (7th Cir.1993):
Social science has challenged many premises of the jury system. Students of the subject believe, for example, that jurors give too much weight to eyewitness evidence and not enough weight to other kinds. Still, the ability of jurors to sift good evidence from bad is an axiom of the system, so courts not only permit juries to decide these cases but also bypass the sort of empirical findings that might help jurors reach better decisions. Juries have a hard time distinguishing “junk science” from the real thing, but aside from some tinkering with the expert testimony admitted at trial, this shortcoming has been tolerated. Jurors reach compromise verdicts, although they aren’t supposed to. Juries return inconsistent verdicts, representing irrational behavior or disobedience to their instructions. Juries act in ways no reasonable person would act. This is the standard for granting judgment notwithstanding the verdict in a civil case, or acquittal after verdict in a criminal case, or reducing an award of damages, and there are plenty of occasions for these post-verdict correctives. Yet for all of this, courts do not discard the premises of the jury system, postulates embedded in the Constitution and thus, within our legal system, unassailable. This shows up in a striking fact about the Supreme Court’s treatment of social science: of the 92 cases between 1970 and 1988 addressing issues of evidence and trial procedure, not one relied on the extensive body of evidence about jurors’ conduct.
(citations omitted).
No psychological study will ever bear directly on the specific persons making an eyewitness identification in court; psychological experts will always be forced to extrapolate from studies done on other people and opine on the relevance such data might have to the facts at hand. *329Cross-examination of the identifying witnesses, on the other hand, will always provide more relevant testimony, because by definition the inquiry is limited to what the eyewitnesses themselves saw and experienced. See Smith, 122 F.3d at 1359 (“defendants who want to attack the reliability of eyewitness recollection are free to use the powerful tool of cross-examination to do so”). Indeed, to a certain extent, lawyers are abdicating their own roles when they seek to rely on experts instead of cross-examination to discredit an eyewitness identification. See Amaral, 488 F.2d at 1153 (“Our legal system places primary rebanee for the ascertainment of truth on the test of cross-examination. [ ... ] It is the responsibility of counsel during cross-examination to inquire into the witness’ opportunity for observation, his capacity for observation, his attention and interest and his distraction or division of attention” (internal quotations and citations omitted)). The witness’s cross-examination testimony can then be framed as the defendant chooses in closing argument to maximize its potential to undermine the identification. See Currie, 515 S.E.2d at 339. What the defendant is unable to establish by these means — e.g., the counter-intuitive concept suggested by psychological research that confidence in one’s recollection does not necessarily reflect accuracy — can be ably communicated by the court in its jury instructions. Instructions have an advantage over experts in that they can be informed by advances in social science research while communicating only those theories that are relevant to the facts of the case, and avoiding the extra delay and expense of producing and rebutting expert testimony, all without the imprimatur of scientific rehability that accompanies expert testimony. Certainly the utility of jury instructions in these situations was aptly demonstrated in this case, where the district court skillfully addressed Smithers’s concerns by adopting an instruction specifically tailored to explain the possible deficiencies of the identifications in this case. In any event, given the utility of cross-examination and jury instructions combined, it is little wonder that the vast majority of appellate cases have found the choice of these mechanisms over expert testimony, even if the expert may have some particular insight that would not be otherwise revealed, not to be an abuse of the district court’s broad discretion under Kumho Tire, Daubert, and Rule 702. See Moore, 882 F.2d at 1110-11; Hall, 165 F.3d at 1107; Smith, 122 F.3d at 1358-59; Hicks, 103 F.3d at 847; Kime, 99 F.3d at 884; Ginn, 87 F.3d at 370; Rincon, 28 F.3d at 925-26; Jordan, 983 F.2d at 938-39; Curry, 977 F.2d at 1051; Blade, 811 F.2d at 464-65; Moore, 786 F.2d at 1311-12; Fosher, 590 F.2d at 382; McClendon, 730 A.2d at 1115-16; McMullen, 714 So.2d at 370; Gaines, 926 P.2d at 642-43; Buell, 489 N.E.2d at 803-04; Currie, 515 S.E.2d at 339.
The presence of a person labeled an “expert” by the court in the witness stand inevitably carries the risk of jurors’ accepting that person’s testimony as scientifically irrefutable truth. This simple fact underlies the special importance given to the court’s gatekeeping function with expert testimony, and it is in the majority’s flat rejection of this concept that its reasoning is the shakiest. In its fifth footnote, the majority observes: “it appears the trial court thought the expert nature of the testimony would unduly impress the jury; this is an improper factor upon which to exclude expert testimony, for if this were the test, no expert could ever testify.” While it may be correct as a hypothetical matter that exclusion of a witness solely because he was an expert would be an abuse of discretion, that is simply not what occurred here. Rather, the court reasoned that cross-examination and a jury instruction were preferable to permitting the jury to hear testimony that was only marginally relevant and demonstrably prejudicial to the Government. The court was in good company in this conclusion. Daubert itself observed that “[e]xpert testimony can be both powerful and quite misleading be*330cause of the difficulty in evaluating it. Because of this risk, the judge in weighing possible prejudice against probative force under Rule 403 ... exercises more control over experts than over lay witnesses.” 509 U.S. at 595, 113 S.Ct. 2786. A number of other courts addressing eyewitness identification expert testimony have explicitly cited the expert’s “aura of reliability” as a prejudicial factor weighing against its admissibility. See Lumpkin, 192 F.3d at 289; Brien, 59 F.3d at 276-77; Blade, 811 F.2d at 465; United States v. Purham, 725 F.2d 450, 454 (8th Cir.1984); Fosher, 590 F.2d at 383-84 (collecting cases referencing the “aura of reliability”); Downing, 609 F.Supp. 784; United States v. Collins, 395 F.Supp. 629, 636-37 (M.D.Pa.1976). The majority’s citation-free asseveration on this subject is simply untenable.
Expert testimony on eyewitness identifications can also be unduly prejudicial when it is phrased so as to comment directly on the credibility of the eyewitness. No court in any context would allow one witness to testify to the credibility of another, because assessment of the credibility of witnesses in our legal system is the sole province of the jury. See Greenwell, 184 F.3d at 496; Gacy, 994 F.2d at 313-14; Murrian, supra, at 380. As illustrated above, a number of courts have cited this tenet as a basis for excluding eyewitness identification experts. That threat was also present in this case, as the majority points out, because Dr. Fulero proposed to testify, among other things, that the eyewitnesses “would have observed and been able to recall the large scar on Mr. Smithers’ neck. That deformity would have been more memorable to the witnesses.” The majority opinion says that the solution to this admittedly inadmissible testimony is simply to excise the offending language. This ignores the fact that at the second hearing, Smithers identified the scar as “the key issue that [Fulero] would address.” Removing this aspect of Fulero’s testimony would gut the remainder of the majority’s reasoning as to why Fulero’s testimony should have been admitted. Nor is this one sentence the only example of how Fulero’s testimony would have stepped over the line. Smithers argued in his renewed motion that “Fulero would testify regarding the perception of the bank robber by [the witnesses] and how [various factors] are directly related to the accuracy of their identification testimony.” (emphasis added). Again: “Fulero would thus testify that the photo spread procedures, and the witness’ numerous meetings with the police, FBI, and each other, would have directly influenced the witness’ ability to recall the particular characteristics of the bank robber with any degree of accuracy.” (emphasis added). Other portions of the motion are phrased in a more appropriate form, indicating that Fulero would testify to research data as it relates to particular conditions experienced by the witnesses, leaving the application of that information to counsel and the jury. But these examples more than adequately justify the district court’s conclusion that Fulero (who, incidentally, is also an attorney) would have acted as more of an advocate than a scientific expert in this case. The majority’s decision merely to excise the offending portions of the testimony not only leaves very little testimony that is even arguably relevant, but relieves Smith-ers of his burden of proving that the testimony he proffered is admissible. Once again, the blame for Fulero’s exclusion lies not with the district court’s legal analysis but with Smithers’s inadequate production.
The cases holding that expert testimony regarding eyewitness identification is too general and those finding that it comments too directly on witness credibility delimit the narrow range of circumstances in which this testimony is properly admissible. Unless a very small number of eyewitness identifications form the only evi-dentiary basis for a conviction, and the proffered testimony relates directly to the facts of the case without commenting on the eyewitnesses’ credibility, the need for this testimony will simply not be so great that alternative means of cautioning the *331jury on this subject will not suffice. See, e.g, Rincon, 28 F.3d at 923-26. The existence of other inculpatory evidence will usually render any error in excluding the expert testimony harmless. See Smith, 736 F.2d at 1107; Hall, 165 F.3d at 1107-08; Smith, 156 F.3d at 1053-54; Blade, 811 F.2d at 465; Moore, 786 F.2d at 1313. Here, the fact that three witnesses identified Smithers adds to the probability of their accuracy. Moreover, the Government presented the identification of Smith-ers’s car at the'bank, the photo analysis showing that Smithers and the robber shared the rare characteristic of being over 6’5” tall, and a series of lies Smithers told police regarding his whereabouts. While this is not overwhelming evidence, it does alleviate considerably any concern that Smithers was convicted solely on the basis of erroneous eyewitness testimony.
The various failings in Fulero’s proposed testimony accentuate the jurisprudential danger posed by the majority’s opinion. Its tangible eagerness to find that the district court abused its discretion in excluding this testimony is likely to set a precedent requiring admission of evidence tending to erode further the jury’s responsibility for making credibility determinations. Other courts have recognized this danger and steered clear of it. See, e.g., Alexander, 816 F.2d at 169 (“Requiring the admission of the expert testimony proffered in Moore would have established a rule that experts testifying generally as to the value of eyewitness testimony would have to be allowed to testify in every case in which eyewitness testimony is relevant. This would constitute a gross overburdening of the trial process by testimony about matters which juries have always been deemed competent to evaluate”); United States v. Thevis, 665 F.2d 616, 641 (5th Cir. Unit B 1982) (“To admit such testimony in effect would permit the proponent’s witness to comment on the weight and credibility of opponents’ witnesses and open the door to a barrage of marginally relevant psychological evidence”); Sabetta, 680 A.2d at 933 (“it would, effectively invade the province of the jury and ... open a floodgate whereby experts would testify on every conceivable aspect of a witness’s credibility”). The logical conclusion of today’s holdings — if not its implicit intent — is likely to be precisely this type of snowball effect in our circuit.
Acutely aware of the dangers of permitting expert testimony without a rigorous performance of the gatekeeping function, Daubert observed:
It is true that open debate is an essential part of both legal and scientific anal-yses. Yet there are important differences between the quest for truth in the courtroom and the quest for truth in the laboratory. Scientific conclusions are subject to perpetual revision. Law, on the other hand, must resolve disputes finally and quickly. The scientific project is advanced by broad and wide-ranging consideration of a multitude of hypotheses, for those that are incorrect will eventually be shown to be so, and that in itself is an advance. Conjectures that are probably wrong are of little use, however, in the project of reaching a quick, final, and binding legal judgment — often of great consequence' — • about a particular set of events in the past. We recognize that, in practice, a gatekeeping role for the judge, no matter how flexible, inevitably on occasion will prevent the jury from learning of authentic insights and innovations. That, nevertheless, is the balance that is struck by Rules of Evidence designed not for the exhaustive search for cosmic understanding but for the particularized resolution of legal disputes.
509 U.S. at 596-97, 113 S.Ct. 2786. I fear that the majority’s opinion here will only undermine the balance between truth-seeking and fairness that the Rules have so carefully crafted, without adding much at all to- the efficacy — at least in this circuit — of criminal justice. Indeed, the majority here holds that “Expert testimony regarding eyewitness identification *332must be recognized as scientifically commensurate with all other psychological studies, and may often be a valid source of information to help jurors understand the factors that effect [sic] eyewitness identifications.” The effect of the majority’s opinion is to establish the district court as the gatekeeper with discretion only to admit, but not to exclude, expert testimony relative to eyewitness identification.
For all of the foregoing reasons, I respectfully dissent. Footnotes
. Contrary to the majority’s characterization of this citation, I do not offer this quotation as a "holy writ” or rigid "test” that the district court should have adhered to, but rather as a *321common-sense explanation of Smithers’s burden to establish the relevance of his proffered testimony to the case. For this reason, the Stevens court's reasoning — or, for that matter, the identical emphasis on specificity in our Smith decision — is made no less valid by the fact that it pre-dates Daubert.
. Tellingly, the district court on remand again dismissed the testimony, this time as unhelpful to the jury and more prejudicial to the prosecution than it was probative to the defense. See 609 F.Supp. 784 (E.D.Pa.1985). The judgment was affirmed without opinion. See United States v. Downing, No. 85-1359, 780 F.2d 1017 (3d Cir. Nov.25, 1985) (table).
. The only assertion made in this memorandum that could arguably be considered "specific” to Fulero’s testimony in this case is the reference to the "known rate of error.” This brief discussion actually originated from Smithers’s recitation of the Daubert analysis. Smithers was "unclear how the third step in the Daubert analysis, reviewing the rate of known error, would apply to this form of scientific testimony.” Not only is this rate-of-error inquiry not a "step” mandated by Dau-bert but simply one of its "general observations,” Daubert, 509 U.S. at 593-94, 113 S.Ct. 2786, it is also clearly inapplicable in this case. Daubert cited as an example of scientific testimony the Seventh Circuit’s treatment of spectrographic voice identification tech-ñique. One method of examining this technique’s reliability was to ask how often it produced an erroneous result. Here, the proposed "technique” is of the exactly opposite type; it seeks not to make an identification, but to explain the reasons why an identification may be incorrect. Hence, a proper analogy to this Daubert observation might be to ask how often this technique correctly ascertains that an identification is wrong.
Nevertheless, Smithers continued: "[T]he question of known rate of error is addressed by the [Handberg] article included as Attachment C.... This article analyzes in detail the effect that certain variables are likely to have on the ability of eyewitnesses to correctly identify persons they have previously seen, pointing out the rate of error in making identifications. [This] forms parts of the scientific basis of Dr. Fulero’s testimony.” This passing reference was Smithers’s entire treatment of the "rate of error” issue, and does not provide the needed specificity.
.The Government noted that "Fulero’s testimony ... would likewise be of dubious assistance to the jury. His testimony does not relate to a specific fact in this case, such as the efficacy of the photo spread. Instead, defendant will offer his testimony regarding the general problems arising from eyewitness identification, in contrast to the specific issues that were presented in the Smith case.”
. Smithers now proposed to have Ftilero testify to the efficacy of the photo spread, despite having no response to the Government’s observation in its prior brief that Smithers had thus far proposed no such thing.
. As I believe my discussion here makes clear, I understand "delay” to mean “prolonging of the length of the trial,” and not, as the majority suggests I mean, merely “filed late.” This certainly appears to have been the district court's understanding as well, since its ruling was made immediately before Smithers rested his case, and granting the motion would have required a “lengthy voir dire,” more preparation'by the Government, and the direct and cross-examination of Dr. Fulero.
. Of course, in order to perform a more detailed inquiry next time, the district court should have the discretion to require Smith-ers to present his witness for voir dire, or at least to make an effort to present a sufficiently detailed proffer in a timely fashion. Unfortunately, the majority’s opinion would appear to curtain that discretion considerably, if, indeed, the majority's opinion leaves any room for the district court to perform any further inquiry at all.