This is an appeal from a delinquency adjudication after a bench trial on charges arising out of an attempted carjacking. The case against L.C. rested on the complainant’s identifications of him shortly after the crime occurred and a month later at trial. L.C. claims the trial court erred in precluding him from presenting expert testimony on relevant psychological factors bearing on the reliability of such eyewitness identifications. We agree that the court erred in ruling that the proffered testimony was not “beyond the ken” of the average layperson without conducting the particularized inquiry required by our decisions in Dyas v. United States1 and Benn v. United States (Benn II).2 We cannot dismiss the error as harmless. We therefore vacate the judgment and remand the case for further proceedings to determine the admissibility of the expert’s testimony.
I.
According to the government’s evidence, the complainant, Adrienne Kinney, had just parked her car in the alley behind her home in the 300 block of Division Avenue, N.E., on the evening of November 22, 2009, when she and her mother (who was accompanying her) were accosted by two unknown men. One of the men pushed Ms. Kinney back into her car, yelled at her, and tried to wrest her car keys from her hands. She resisted, and as the women began screaming for help and honking the car horn, the two assailants gave up and fled without taking anything. Ms. Kinney and her mother immediately called the police, who arrived within two or three minutes. Ms. Kinney told police the man who grabbed for her keys was taller than she was, of medium build and complexion, and that he was wearing an open-faced ski mask and a light blue ski jacket with a *294white stripe or stripes. She said his companion also wore a mask and was dressed in black.3
Within a few minutes, the police stopped appellant and a second man on the street at a location approximately 200 feet from the scene of the attempted carjacking, in the area toward which the perpetrators had run. Appellant was wearing a light blue jacket with white stripes; in his pocket police later found an open-faced ski mask. The man accompanying him had on a black jacket and blue jeans; the police did not find a mask in his possession. The police arranged for Ms. Kinney to view the two men. Sitting in a police car at a distance of approximately fifty feet, she identified appellant as her attacker and his companion as the second would-be carjacker. Ms. Kinney stated that she recognized appellant’s jacket, his complexion, his build, and his face, and she commented, “That’s why his hair ... looked puffy in the ski mask — he has dreads.”
Four weeks later, at appellant’s trial, Ms. Kinney made an in-court identification of appellant. She testified that she first noticed him and his companion when she drove her car into the alley, and that she paid close attention to them because they made her nervous. The two men were strangers whom she had never seen before that night. As she got out of her car and proceeded to gather her things, she was hoping they would walk on by, which they did — but then they abruptly turned and “started to rush” her. Ms. Kinney claimed she had a good look at appellant’s clothing; motion lights on the building behind her enabled her to see the colors of his jacket. She was face to face with appellant as they struggled over her keys. Although he was wearing a ski mask, it had a wide opening, through which Ms. Kinney said she could see the shape of his face and his nose, lips, eyes, teeth and skin. She professed to be certain of her identification. On cross-examination, though, Ms. Kinney agreed that she would describe appellant as having a dark complexion rather than a medium complexion. She also acknowledged that during the assault, she was screaming, her heart was racing, and she was afraid for both herself and her mother.
Appellant’s defense at trial was misiden-tification. He called two witnesses — his older cousin, with whom he and his mother resided, and William Locust, the person whom the police had stopped with L.C. They testified that L.C. was on the front steps of his neighbor’s house at the time of the attempted carjacking, and that he had just left there and was walking to a nearby convenience store and a gas station when the police stopped him.
To bolster his misidentification defense, appellant sought to call Dr. Steven Penrod to testify as an expert about research findings regarding psychological factors arguably present in this case that- had been shown to reduce the probativity of eyewitness identifications. The pertinent proffered factors included: an effect known as “clothing bias” that can result in a mistaken identification when the eyewitness is shown a suspect who is wearing clothing similar to that the witness had described;4 *295the poor correlation between the confidence an eyewitness expresses and the accuracy of the identification; studies showing that stress and emotional arousal negatively affect the accuracy of identifications by impairing the witness’s ability to perceive and to remember the perpetrator’s face;5 and the heightened suggestivity and unreliability of show-up identifications as compared to other identification procedures. The government moved in limine to exclude Dr. Penrod’s anticipated testimony on the grounds that it did not satisfy the first or third prongs of the three-prong test for the admission of expert testimony set forth in Dyas.6 Specifically, the government argued, the proffered testimony was not “beyond the ken of the average layman,” and the unsettled state of scientific knowledge did not permit a reasonable opinion to be offered by any expert. The motion asserted that this court’s case law supported the exclusion of expert testimony on the psychological factors affecting the reliability of eyewitness identification.7
After hearing argument on the motion at a pretrial hearing, but without either conducting a voir dire examination of Dr. Penrod (who was present and available for that purpose) or, so far as appears, evaluating the proffered psychological research on which Dr. Penrod relied, the trial court ruled his testimony inadmissible on the sole ground that none of it was beyond the ken of the average layperson. Rather, in the court’s view, the proposed areas of expert testimony were all matters within the reach of “common sense” that the defense could bring out in cross-examination and address in argument. The court did not rule on whether the other requirements of Dyas were satisfied.
In closing arguments, opposing counsel disputed whether the reliability of Ms. Kinney’s identification of appellant was undermined by clothing bias, stress, and the suggestiveness of the show-up procedure, whether there were significant inconsistencies in her previous descriptions of her assailant, and whether her confidence in her identification was entitled to weight. The court’s ruling excluding Dr. Penrod’s testimony precluded appellant’s counsel from supporting his arguments on these *296matters with scientific studies.8
After hearing arguments, the trial court credited Ms. Kinney’s identification of appellant and found him guilty of carjacking and assault with intent to commit robbery.
II.
As L.C. argues, and as the government does not dispute, our decisions in Benn II and Russell v. United States9 now make it clear that the trial court erred in summarily concluding that the proffered expert testimony was not beyond the ken of the average layperson. In those cases this court recognized that the insights of modern psychological research into the factors influencing eyewitness identifications are not matters of common knowledge or common sense and are, indeed, often counterintuitive.10 It makes no difference that the fact-finder in this case was an experienced trial judge whose background and knowledge in the area might exceed that of the average layperson. This court has rejected the argument that when the Dyas issue is presented in a bench trial, the proper standard for admissibility should be the “ken of the presiding trial judge” rather than that of the average layperson.11 The average layperson standard applies “even though it could be said that the trial [judge is] competent to resolve the issue without the aid of an expert.” 12
Seeking to avoid the remedy required by Russell and Benn II — a remand for a full Dyas inquiry that likely would include a voir dire examination of Dr. Penrod13 — the government urges us to affirm on an alternative ground: There was no reversible error, the government contends, because the exclusion of Dr. Penrod’s testimony “was justified by the corroborative evidence in this case, which demonstrated the reliability of Ms. Kinney’s identification of appellant.”14 We must reject this argument.
First, even if we were to assume arguendo that, in deciding whether to admit Dr. Penrod’s testimony, the trial court had discretion to consider the evidence corroborating the complainant’s identifica*297tion of appellant, the court did not exclude the expert testimony on that basis, and the court certainly was not obligated to do so. That suffices to foreclose the government’s argument, because this court may not affirm a discretionary ruling on a ground the trial court did not rely on and had discretion to reject.15
Second, and more fundamentally, the corroborative evidence was irrelevant to the question of the admissibility of appellant’s proffered expert testimony, for the answer to that question depended only on whether the expert testimony would satisfy the three Dyas criteria and, if so, whether any danger of unfair prejudice or the like would nonetheless mandate its exclusion.16 Even in a bench trial, the inquiry as to admissibility does not call for the court to assess the weight of pertinent and admissible expert testimony in light of the anticipated strength of the opposing party’s evidence.
In general, if evidence is relevant, it should be admitted unless it is barred by some other legal rule.17 “There are two components to relevant evidence; materiality and probative value.”18 “[T]he fact sought to be established by the evidence must be material, which is to say that the party must establish that fact as a condition to prevailing on the merits of his case.”19 And the evidence must have probative value, meaning “the tendency of evidence to establish the proposition that it is offered to prove.”20 The probativity threshold for purposes of admissibility is low: Am item of evidence, to be relevant, need only “tend[ ] to make the existence or nonexistence of a fact more or less probable than would be the case without that evidence.”21 As Professor McCormick explains:
*298An item of evidence, being but a single link in the chain of proof, need not prove conclusively the proposition for which it is offered. It need not even make that proposition appear more probable than not. Whether the entire body of one party's evidence is sufficient to go to the jury is one question. Whether a particular item of evidence is relevant to the case is quite another. It is enough if the item could reasonably show that a fact is slightly more probable than it would appear without that evidence. Even after the probative force of the evidence is spent, the proposition for which it is offered still can seem quite improbable.22
“Ordinarily,” therefore, “any evidence [that] is logically probative of some fact in issue is admissible[,] and if the evidence offered conduces in any reasonable degree to establish the probability or improbability of a fact in controversy, it should go to the jury” or, in a non-jury trial, to the judge.23 Thus, evidence may not be rejected as irrelevant merely because it is contradicted by other evidence.24
The foregoing principles, applicable to evidence in general, apply equally to the admission of relevant expert testimony. The standard of relevance is the same for expert testimony as it is for other evidence; as we have held, “there is only one standard of relevance.”25 It is true that the trial judge must exercise discretion to decide whether proffered expert testimony is likely to assist the trier of fact in the performance of its duties — “that is to say, in understanding the evidence, determining the facts that must be found and rendering its verdict.”26 But such helpfulness is determined by the three criteria governing the admissibility of expert opinion testimony set forth in Dyas, and ultimately turns on “the relevance and probative value of the proposed scientific evidence.”27 In the end, therefore, the criterion of helpfulness is met if the expert testimony is material and if its probative value is not substantially outweighed by the danger of unfair prejudice or other legitimate concerns.
Thus, the criterion of helpfulness is not a grant of authority to the trial *299judge to exclude relevant and otherwise admissible expert testimony merely because it is against the expected weight of the evidence.28 Still less does Dyas authorize the trial judge to exclude expert testimony as unhelpful based on the perceived strength of the opponent’s evidence alone.29 The rationality of such a rule cannot be defended, for as the Supreme Court explained in Holmes v. South Carolina, “by evaluating the strength of only one party’s evidence, no logical conclusion can be reached regarding the strength of contrary evidence offered by the other side to rebut or cast doubt.”30
In taking the opposing view, the government relies on cases in which we have said that where corroboration of a challenged identification exists, the exclusion of proffered expert testimony on eyewitness identification generally does not constitute an abuse of discretion.31 But that does not mean the existence of corroboration is a legitimate reason for the trial court to exclude the expert testimony; it only means that the exclusion likely will not be so prejudicial as to necessitate reversal by the appellate court. Abuse of discretion is a standard of appellate review incorporating an assessment of prejudice. A trial court exercises its discretion erroneously when it relies on an improper factor, but “the reviewing court must weigh the severity of the error against the importance of the determination in the whole proceeding and the possibility for prejudice as a result.”32 It is only when the impact of the error is so serious we must reverse that we say the trial court “abused” its discretion.33 In some cases in which a challenged identification was amply corroborated, the appellate court may be able to conclude that the erroneously excluded expert testimony would not have undermined it significantly. When that is so, the court can conclude that the error did not affect the outcome of the trial, and hence that there was no “abuse” of discretion.34 But to say the trial court did not abuse its discretion is not to say the court exercised its discretion properly.
That brings us to whether the erroneous ruling was harmless in this case. The burden is on the government to persuade us that “the judgment was not substantially swayed by the error”35 — a standard requiring us to find it “highly probable” that the error did not contribute *300to the verdict.36 We are not prepared to make such a finding here. The government’s case rested on a single, contested eyewitness identification of a stranger, the reliability of which the excluded expert testimony might have afforded a reasonable basis to doubt. We said in Benn that
[i]n a case grounded on eyewitness identifications of a stranger, without other corroborating evidence, and in which the defense depends entirely upon demonstrating that the identifying witnesses are not as reliable as they believe themselves to be, to preclude the defendant from presenting the scientific testimony of a qualified expert on research that is generally accepted and not known to lay jurors to prove this point is not harmless under Kotteakos v. United States, provided that the facts underlying the identifications establish a sound foundation for applying the principles expounded by the expert.37
Here, to be sure, the identification was corroborated to an extent: When L.C. was stopped only minutes after the attempted carjacking, a short distance from the scene, he was wearing a blue jacket with a white stripe and he had a ski mask in his pocket — both of which matched the complainant’s description of what her assailant was wearing. But without denigrating this evidence, we must acknowledge that we cannot be confident enough that it rendered the identification of appellant free from reasonable doubt that might have been instilled had Dr. Penrod’s testimony been permitted. As appellant argues, his presence in the vicinity was not unusual given that he lived in the neighborhood, his jacket was not uncommon, and a ski mask was an appropriate item of cold weather apparel to have in a coat pocket in November when the nights may get chilly. The government presented no independent corroboration of appellant’s guilt — no physical or scientific evidence, incriminating admissions or behavior by appellant, or third-party testimony implicating him in the carjacking.
Russell is instructive. In that case we held that the trial court’s erroneous preclusion of testimony from the defendant’s eyewitness identification expert required a remand for a Dyas hearing notwithstanding the presence of corroborative evidence comparable to, if not indeed stronger than, such evidence here.38 We deemed it significant that there was no scientific or physical evidence linking Russell to the crime, nor any incriminating admissions.39 And we recognized that “the excluded expert testimony was central to [Russell’s] misidentification defense;” without it, we emphasized, he “had no factual underpinnings for the scientific theories he sought to present that might cast doubt on the eyewitness’ testimony.”40 Much the same is true here. If the erroneous exclusion of the defense expert’s testimony was not harmless in Russell, it is difficult to see *301how we could reach a different conclusion in this case.41
For the foregoing reasons, we hereby vacate the judgment and remand the case for further proceedings to determine the admissibility of the proffered expert testimony in accordance with Benn II under the criteria set forth in Dyas.42 “If the trial court determines ... that the expert testimony should have been permitted, it shall order a new trial. If the trial court adheres to its ruling that the testimony is inadmissible, it shall re-enter the judgment subject to appellant’s rights to challenge the ruling in a renewed appeal.”43
So ordered.
. 376 A.2d 827, 832 (D.C. 1977).
. 978 A.2d 1257, 1269-70, 1273-74 (D.C.2009).
. Although Ms. Kinney’s mother also talked to the police and participated in the ensuing show-up identification procedure, she did not testify and neither party relied on her at trial. We therefore omit any further description of her involvement and the information she gave the police.
. For example, Dr. Penrod apparently was prepared to testify about a recent study in which subjects viewed a target individual wearing distinctive clothing. When, later in the day, the subjects were shown another person wearing similar clothing, half of the subjects incorrectly identified the person as the target. See also State v. Henderson, 208 *295N.J. 208, 27 A.3d 872, 903 (2011) (en banc) (noting research findings indicating that "showups increase the risk that witnesses will base identifications more on similar distinctive clothing than on similar facial features”) (citing Jennifer E. Dysart et al., Show-ups: The Critical Issue of Clothing Bias, 20 Applied Cognitive Psychol. 1009, 1019 (2006), and A. Daniel Yarmey et al., Accuracy of Eyewitness Identifications in Showups and Lineups, 20 Law & Hum. Behav. 459, 461, 470 (1996)).
.Among other things, Dr. Penrod was prepared to testify about a "meta analysis" he had performed of some twenty-seven independent tests on the effect of stress on the reliability of eyewitness identifications. He found that the average error rate in the identifications was 34% when the witness was under high stress as compared with 19% under low stress conditions.
. We held in Dyas that expert testimony must satisfy three criteria to be admissible: (1) the subject matter of the testimony must be so distinctively related to some science, profession, business or occupation as to be beyond the ken of the average layperson; (2) the witness must have sufficient skill, knowledge or experience in the field or calling as to make it appear that his opinion or inference will probably aid the trier of fact in its search for truth; and (3) the state of the pertinent art or scientific knowledge must permit a reasonable opinion to be asserted by an expert. Dyas, 376 A.2d at 832.
. The government did not mention this court's opinion in Bern II, which had been decided three months earlier. Appellant cited Benn II to the trial court when the motion was argued.
. The court sustained an objection when defense counsel attempted to cite Dr. Penrod's research, on the ground that the research was “not common knowledge.” [12/31/09 Tr. at 105-06]
. 17 A.3d 581 (D.C.2011).
. See Benn II, 978 A.2d at 1277 ("Despite the fact that jurors may be familiar from their own experience with factors relevant to the reliability of eyewitness observation and identification, it cannot be said that psychological studies regarding the accuracy of an identification are within the ken of the typical juror.”) (internal quotation marks, brackets and ellipses omitted); id. at 1268 (noting that "jurors, as a matter of common sense, are not fully aware of the factors that influence eyewitness testimony,” and that research has disproved commonly held beliefs in the accuracy of eyewitness identifications); see also Russell, 17 A.3d at 588 (stating that the court will have to undertake a "more in-depth consideration” of the proffered testimony on remand before rejecting it "based upon previously accepted common sense notions of what matters are within the ken of the average lay person”).
. Girardot v. United States, 996 A.2d 341, 347 (D.C.2010) (brackets omitted).
. Id. at 348.
. See Russell, 17 A.3d at 588; Benn II, 978 A.2d at 1277-80.
. Brief for Appellee at 29. There was, of course, both evidence corroborating Ms. Kinney’s identification of appellant (e.g., the jacket he was wearing and the ski mask in his pocket when he was stopped) and evidence contradicting that identification (e.g., the testimony of appellant’s witnesses). The government’s argument discounts the worth of the latter evidence.
. See Randolph v. United States, 882 A.2d 210, 219 (D.C.2005).
. See Ibn-Tamas v. United States, 407 A.2d 626, 632 (D.C.1979); accord Robinson v. United States, 50 A.3d 508, 523 (D.C.2012) (“In general, expert testimony should be admitted if it is relevant and is likely to help the trier of fact in its search for the truth, that is to say, in understanding the evidence, determining the facts that must be found and rendering its verdict.’’) (internal citations and quotation marks omitted).
. See, e.g., Reavis v. United States, 395 A.2d 75, 78 (D.C.1978) ("Relevance, and the concepts it embodies, determines initially whether a proffered item of evidence will be admissible.”); 2 Clifford S. Fishman, Jones on Evidence § 11:1 at 258 (7th ed. 2000) ("Evidence that is relevant should be admitted, unless barred by some other rule.”); 1 Kenneth S. Broun et al., McCormick on Evidence § 184 at 728 (6th ed. 2006) ("[U]nless there is some such distinct ground for refusing to hear the evidence, it should be received.”) (footnote omitted); Fed.R.Evid. 402 ("Relevant evidence is admissible unless any of the following provides otherwise: the United States Constitution; a federal statute; these rules; or other rules prescribed by the Supreme Court. Evidence which is not relevant is not admissible.”).
. McCormick on Evidence § 185 at 729 (footnote omitted).
. Reavis, 395 A.2d at 78; see also McCormick on Evidence § 185 at 729 ("[Materiality] looks to the relation between the propositions that the evidence is offered to prove and the issues in the case.”).
. Id. § 185 at 730.
. Punch v. United States, 377 A.2d 1353, 1358 (D.C.1977); accord Plummer v. United States, 813 A.2d 182, 188 (D.C.2002) ("Evidence is relevant if it has 'any tendency to make the existence of any fact that is of consequence to the determination of the action more probable or less probable than it would be without the evidence.’ ") (quoting Street v. United States, 602 A.2d 141, 143 (D.C.1992)); Fed.R.Evid. 401 ("Evidence is relevant if: (a) it has any tendency to make a fact more or less probable than it would be without the evidence; and (b) the fact is of consequence in determining the action.”). To put it differ*298ently, evidence is irrelevant in the sense that it lacks probative value only if "the evidence does not justify any reasonable inference as to the fact in question. Cases involving such evidence are few and far between.” McCormick on Evidence § 185 at 735 (emphasis in original; footnote omitted).
. McCormick on Evidence § 185 at 733 (footnotes omitted). See, e.g., Winfield v. United States, 676 A.2d 1, 4 (D.C.1996) (en banc) (holding that when the defendant seeks to introduce evidence of a third-party perpetrator, "there is no requirement that the proffered evidence must prove or even raise a strong probability that someone other than the defendant committed the offense”) (internal quotation marks and alterations omitted).
. Plummer, 813 A.2d at 188-89 (internal quotation marks and brackets omitted).
. Evidence that is relevant and not barred by some other legal rule may be excluded, in the trial court's discretion, if the court finds that its probative value is substantially outweighed by a danger of unfair prejudice, confusion of the issues, or other legitimate concerns. See (William A.) Johnson v. United States, 683 A.2d 1087, 1100 (D.C.1996) (en banc) (adopting Federal Rule of Evidence 403). The trial court in this case did not make such a finding, nor does the government suggest that it should or could have done so.
. Winfield, 676 A.2d at 3.
. Hager v. United States, 856 A.2d 1143, 1147 (D.C.2004) (quoting Steele v. D.C. Tiger Market, 854 A.2d 175, 181 (D.C.2004)) (internal quotation marks omitted), amended by 861 A.2d 601 (D.C.2004).
. Bern II, 978 A.2d at 1278. Perhaps it would be more precise to say that helpfulness turns on the materiality and probative value of the proposed scientific evidence.
. See, e.g.. Western Indus., Inc. v. Newcor Can., Ltd., 739 F.2d 1198, 1202 (7th Cir.1984) (Posner, J.) (holding that trial judge erred in excluding expert opinion testimony, because “a judge in our system does not have the right to prevent evidence from getting to the jury merely because he does not think it deserves to be given much weight,” and thus "he may not screen witnesses simply to decide whether their testimony is persuasive”).
. As noted previously, while there was evidence corroborating Ms. Kinney’s identification of appellant, there also was evidence contradicting it.
. 547 U.S. 319, 331, 126 S.Ct. 1727, 164 L.Ed.2d 503 (2006). Indeed, Holmes makes clear that a rule of evidence allowing a trial judge to exclude a defendant's relevant and otherwise admissible expert testimony when the prosecution’s evidence of the defendant’s guilt is strong would contravene the constitutional guarantee of "a meaningful opportunity to present a complete defense.” Id.
. See Benn II, 978 A.2d at 1280, and Hager, 856 A.2d at 1149.
. (James W.) Johnson v. United States, 398 A.2d 354, 365, 367 (D.C.1979).
. Id.
. See, e.g., Heath v. United States, 26 A.3d 266, 282-85 (D.C.2011), cert. denied, — U.S. -, 134 S.Ct. 898, 187 L.Ed.2d 774 (2014).
. Kotteakos v. United States, 328 U.S. 750, 765, 66 S.Ct. 1239, 90 L.Ed. 1557 (1946).
. Heath, 26 A.3d at 275 n. 18 (D.C.2011), (internal quotation marks omitted).
. Benn, 978 A.2d at 1283 (footnotes omitted).
. In Russell the defendant’s clothing and appearance when he was apprehended shortly after the carjacking matched the victim’s description of the carjacker as a man with his hair in dreadlocks wearing a black North Face jacket with a hood, blue jeans, and black boots. Moreover, his jeans were wet and muddy, which was consistent with the fact that a police officer had seen the perpetrator flee into the woods. And unlike in this case, in Russell there were two eyewitness identifications, each corroborating the other. 17 A.3d at 584, 589-90.
. Id. at 589.
. Id.
.The government argues that appellant was able to test the reliability of Ms. Kinney's identification of him through cross-examination. But "[a]ppellant’s proffered scientific theories could not be developed on the cross examination of lay witnesses.” Id. As in Russell, ”[w]e are not persuaded that the fact that appellant cross-examined vigorously the eyewitness[ ] and argued the shortcomings of the identificationf ] to the [court] to the extent that he could means that he had a meaningful opportunity to present his misidentification defense.” Id.
. See Benn II, 978 A.2d at 1278, 1280.
. Russell, 17 A.3d at 589-90.