(dissenting). Our Court has made plain that it is an abuse of discretion for a trial court to preclude expert testimony on the reliability of eyewitness identifications where the case “turns on the accuracy of eyewitness identifications and there is little or no corroborating evidence connecting the defendant to the crime,” and the proposed expert “ ‘testimony is (1) relevant to the witness’s identification of defendant, (2) based on principles that are generally accepted within the relevant scientific community, (3) proffered by a qualified expert and (4) on a topic beyond the ken of the average juror’ ” (People v Santiago, 17 NY3d 661, 669 [2011], quoting People v LeGrand, 8 NY3d 449, 452 [2007]). A trial court cannot rest its determination to exclude expert testimony on alleged corroborating evidence that is itself unreliable (Santiago, 17 NY3d at 673).
In defendant’s case, the proposed corroborating evidence came from a coparticipant in the crimes and was highly suspect, with more than a speculative possibility that the jury would question the credibility of its source. I would not ignore, as the majority apparently does, the “several factors [that] call the corroborating identification[ ] into question” {id.). Instead, based on facts indicative of the untrustworthiness of this testimony, rendering doubtful its reliability, I conclude that the trial court abused its discretion when it denied defendant’s request to present expert testimony on matters potentially affecting the accuracy of the eyewitness identification. Therefore, I dissent and would affirm the Appellate Division order reversing defendant’s conviction and granting him a new trial.
I.
A jury convicted defendant of murder in the second degree, robbery in the first degree, and attempted robbery in the first degree, arising from the fatal shooting of Vincent Dotson, during a robbery in his barbershop. There was no forensic, DNA, or physical evidence connecting defendant to the crime, and the sole eyewitness was the surviving victim, “J.J.,” who observed defendant under conditions that defendant argued could potentially impact the reliability of the eyewitness identification.
Defense counsel filed a motion in limine to admit expert testimony concerning the influence of various factors on eye*1163witness recollection, including the level of violence during the interaction, the length of time of the incident, and the presence of a weapon. Defendant alternatively requested a hearing pursuant to Frye v United States (293 F 1013 [DC Cir 1923]) to determine whether his proposed expert testimony was generally accepted within the relevant scientific community. During argument on the motion, the People contended that under our case law no expert was necessary because J.J.’s testimony was corroborated by a separately charged codefendant, Willie Harvey, who would place defendant in the vicinity of the crime. The People stated that although Harvey did not go inside the barbershop and was not a witness to the robbery or shooting, he would testify that he drove defendant and three other men around the corner from where the barbershop was located, drove away with them, and later saw them with guns and the robbery proceeds.
The People also sought to “make the record very clear” about various credibility issues related to Harvey. The People described Harvey’s initial failure to identify defendant from a photo array, even after Harvey implicated his own brother and another person in the crimes. The People also explained that Harvey was initially charged with murder, robbery, and attempted robbery in the first degree, and entered a deal by which he was allowed to plead guilty to the robbery count in satisfaction of all charges, in exchange for testifying against defendant and the other participants in the crimes. The People alternatively requested a Frye hearing if the court was considering permitting the expert testimony.
In response, defense counsel argued Harvey could not provide the necessary corroboration for J.J.’s testimony because Harvey’s statements were self-serving, and his description of events was partially contradicted by other testimony. Moreover, Harvey admitted he had never seen defendant until the night of the crime.
The Judge denied the motion, explaining that he previously observed J.J. testify against another codefendant and considered him to be “a credible witness,” and that Harvey’s testimony appeared sufficient to corroborate J.J.’s identification of defendant. At the close of the People’s case, the Judge denied counsel’s renewed motion, without a Frye hearing, upon a finding that Harvey corroborated J.J.’s testimony.
*1164II.
At trial, J.J. testified that he observed a man exit a white Chevy Malibu and enter Dotson’s barbershop to get a haircut. Soon after, three other men entered, one of whom tried to lock the door behind them. J.J. described the first two men who walked into the barbershop as dark-skinned, and stated that the first man wore a dark coat and black winter hat. The third man was lighter-skinned and taller, and wore an orange coat and matching baseball cap.
The men ordered Dotson and J.J. to the ground, demanded money and drugs, and pistol-whipped them. After they had taken $200 from Dotson, the first of the three men to enter the barbershop fatally shot Dotson in the chest. The men ran out of the shop, but the shooter returned. J.J. closed his eyes and heard a clicking sound over his head. After he heard the shooter leave, J.J. called 911.
Several weeks after the shooting, J.J. saw defendant in a photo array and told the police that he looked like the shooter. Two months later, J.J. picked out defendant in a lineup, only this time he said defendant was the last man to enter the barbershop, notwithstanding the different descriptions he had previously provided of these two men. Defendant was the only person displayed in both the photo array and the lineup. J.J. made an in-court identification of defendant, again describing him as the last man of the three who entered the shop.
Harvey testified and also identified defendant as one of the men involved in the robbery. Harvey described how he first met defendant on the day of the crime, when Harvey, Harvey’s brother and cousin, defendant, and another stranger drove in separate cars to where the shooting occurred. Everyone but Harvey exited the cars and went around the corner. About 15 minutes later they returned, and again drove in separate cars to another location where Harvey saw the men put marijuana and guns on the hood of the car he had been riding in, a white Chevy Malibu. He then left in the Malibu with his brother and cousin, and rode with them until he was apprehended by the police later that evening.
As the prosecutor foretold, Harvey’s credibility was questionable. Defense counsel sought to establish Harvey’s penchant for fabrication. She cross-examined Harvey about his false statements to the police, including his initial denials of any knowledge about the crimes, and his attempts to minimize his role after he was identified as the driver. Although Harvey told *1165the court at his plea that he was the driver, he contradicted this statement at defendant’s trial and downplayed his involvement, saying he was a passenger and that his brother drove to the crime scene.1
The circumstances surrounding Harvey’s identification of defendant were particularly troubling. Harvey admitted that he initially failed to pick out defendant from a photo array a month after the shooting, only to identify him from a second photo array another month later, while Harvey was incarcerated and awaiting trial on charges of murder and robbery for the Dotson shooting. After Harvey identified defendant, Harvey entered a plea agreement whereby he would receive a 10-year sentence, in exchange for pleading guilty to robbery in the first degree and testifying against defendant and the other participants in the crimes.
The Appellate Division reversed defendant’s conviction and ordered a new trial, concluding that the trial court wrongly excluded the expert testimony on witness identification because there was little or no corroborating evidence connecting defendant to the crime, and Harvey’s testimony established that he was a person of dubious credibility, whose identification of defendant was itself unreliable. In my opinion, there is no merit to the People’s appeal.
III.
The People argue that the trial court did not abuse its discretion by denying defendant’s request to admit expert evidence on the reliability of eyewitness identification because Harvey’s observations of defendant were sufficient to corroborate J.J.’s identification, rendering the expert’s testimony unnecessary. In the alternative, the People request a Frye hearing to determine if the expert testimony is admissible, thus providing a basis for a new trial.
In the wake of the growing body of research on factors impacting the reliability of eyewitness identifications, and the data establishing a connection between misidentifications and wrongful convictions, we have encouraged courts to admit expert testimony on eyewitness identifications in appropriate cases, as a means to educate the jury about these factors (Santiago, 17 NY3d at 669, quoting People v Drake, 7 NY3d 28, 31 [2006], citing People v Young, 7 NY3d 40, 45 [2006]). This *1166type of expert testimony is particularly useful to the jury because research shows that jurors tend to find eyewitness identification extremely believable and treat certain factors as good indicators of accuracy, even though research suggests that some of these factors impact adversely on the reliability of eyewitness testimony (see John C. Brigham & Robert K. Bothwell, The Ability of Prospective Jurors to Estimate the Accuracy of Eyewitness Identifications, 7 Law & Hum Behav 19, 28 [1983]; Kenneth A. Deffenbacher et al., A Meta-Analytic Review of the Effects of High Stress on Eyewitness Memory, 28 Law & Hum Behav 687, 699 [2004]; Brian H. Bornstein et al., Effects of Exposure Time and Cognitive Operations on Facial Identification Accuracy: A Meta-Analysis of Two Variables Associated with Initial Memory Strength, 18 Psychol, Crime, & L 473 [2012]).
The admission of expert testimony on the accuracy of eyewitness identification rests within the discretion of a trial court, after “weighing a request to introduce such expert testimony ‘against other relevant factors, such as the centrality of the identification issue and the existence of corroborating evidence’ ” (Santiago, 17 NY3d at 668-669, citing People v Lee, 96 NY2d 157, 163 [2001]). Nevertheless, “there are cases in which it is unfair to deprive the jury of expert testimony about the reliability of eyewitness observations” (People v Oddone, 22 NY3d 369, 379 [2013]). In such cases it is an abuse of discretion to exclude expert testimony (LeGrand, 8 NY3d at 456).
In LeGrand, the Court “established a two-stage inquiry” governing this exercise of a trial court’s discretion (Santiago, 17 NY3d at 669, citing LeGrand, 8 NY3d at 452). Under the first stage, the trial court must determine “whether the case ‘turns on the accuracy of eyewitness identifications and there is little or no corroborating evidence connecting the defendant to the crime’ ” (id., quoting LeGrand, 8 NY3d at 452).2 Where there exists sufficient corroborating evidence, it is reasonable *1167for a trial court to conclude that the eyewitness identification is “quite unlikely to be mistaken, and that [expert] testimony would be an unnecessary distraction for the jury” (Young, 7 NY3d at 46). However, if the corroborating evidence, or witness, presents reliability issues, a trial court has no basis to assume a jury will not benefit from expert testimony on factors affecting the accuracy of eyewitness identification, and the scales tip in favor of allowing the expert to testify, if otherwise qualified. This is especially so when the eyewitness testimony is subject to the type of factors that place its reliability in doubt.
For example, in Santiago, no physical evidence linked the defendant to an assault in a subway station, but the victim and two other eyewitnesses identified the defendant as the assailant. Several factors impacted the accuracy of the victim’s identification. The defendant was a stranger to the victim, and the assailant’s face was partially concealed during the attack by his clothing and hat, such that the victim could not see the assailant’s hair and his face was not visible “[f]rom the middle of his top lip, down, and from the top of his eyebrows up” (Santiago, 17 NY3d at 664). The Court held that the trial court abused its discretion in denying defendant’s request to admit expert testimony on eyewitness identification, rejecting the People’s argument that the non-victim witnesses connected defendant to the crime. The Court determined that these eyewitnesses were insufficient to support exclusion of the expert because “several factors call[ed] the corroborating identifications into question” (id. at 673).3 One eyewitness, like the victim, only partially viewed the assailant’s face, and picked *1168out defendant from a lineup with “only 80% confidence” (id.). The Court also recognized that his memory may have been tainted by having seen a newspaper photograph of the defendant linking him to the crime the day after he initially viewed the lineup. Similarly, the other eyewitness’s identification of the defendant “may have been influenced by his memory of the police artist’s sketch of the assailant, calling into question the independence of this evidence” (id.).
However, where the corroborative evidence has a “strong indicia of accuracy” (id. at 671), the trial court may exercise its discretion to exclude expert testimony which might confuse and distract the jury from principal matters in the case (Oddone, 22 NY3d at 379). Thus, this Court found in People v Lee that the victim’s identification of the defendant as the person who stole his car was sufficiently corroborated by evidence that the defendant was arrested driving the stolen vehicle (96 NY2d 157, 163 [2001]). In that case, the Court also noted that the circumstances under which the victim observed the defendant supported its reliability. In People v Young, a home invasion and robbery case, the eyewitness saw only part of defendant’s face and “retained a ‘mental image’ only of his eyes” (7 NY3d at 42). Yet the Court agreed that there was sufficient corroboration evidence to link defendant to the crime because the “stolen property was found in possession of two of defendant’s acquaintances; neither of [whom] could have been the robber,” and one of the acquaintances admitted she got the property from the defendant (id. at 46).
In People v Allen, eyewitnesses independently identified the defendant as the knife-wielding member of a two-man team of masked intruders in a barbershop robbery. The first eyewitness “quickly . . . recognized” the defendant because he regularly *1169encountered him in the neighborhood, was familiar with his voice, and also knew his nickname — the same nickname defendant provided to the police upon his arrest. When the police arrived at the barbershop, this eyewitness provided a description and information about the defendant and subsequently picked out the defendant from photos in mug books and a photo array. This identical photo array was shown the same day of the robbery to the second eyewitness who immediately identified the defendant’s photo, and who also knew defendant from the neighborhood (People v Allen, 13 NY3d 251, 262 [2009]; Santiago, 17 NY3d at 670 [discussing Allen: the “(t)wo witnesses recognized the robber with the knife as an individual whom they knew from the neighborhood”]). The second eyewitness had a close view of the defendant because the defendant had searched him for jewelry and stood near him during the robbery. Both eyewitnesses separately identified defendant from a lineup four months later. Under these circumstances, the Court concluded that the trial court did not abuse its discretion in denying expert testimony because “the corroborating identification possessed strong indicia of accuracy. In particular the defendant . . . was known to the second eyewitness, who recognized him during the robbery” (Santiago, 17 NY3d at 671, discussing Allen, 13 NY3d at 269).
Here, there is no question that J.J.’s identification was central to the People’s case since it was the only direct evidence linking defendant to the crimes. The sole question is whether Harvey provided sufficient corroborating evidence to “significantly diminish! ] the importance of the proffered expert testimony” (Young, 7 NY3d at 46). The corroborating evidence must be sufficiently reliable that a trial court may conclude with some confidence that the eyewitness identification is “quite unlikely to be mistaken, and that [expert] testimony would be an unnecessary distraction for the jury” (id.). The record does not support such a conclusion in defendant’s case.
In contrast to Lee, Young, and Allen, and similar to Santiago, there are factors suggesting that Harvey’s identification of defendant is unreliable, based on undisputed facts placing his credibility in question. Harvey initially failed to identify defendant from a photo array a month after the crimes, and only implicated defendant after Harvey had been charged, arrested and held in jail awaiting trial for the murder and robbery, which carried a possible life sentence. He then identified defendant after he was offered a 10-year sentence in exchange for *1170his plea, thus avoiding life in prison. Although he testified that the only condition of the plea was that he had to testify and tell “[his] side of the story” at defendant’s trial, nevertheless he identified defendant right before his plea, raising the specter of fabrication to secure a generous plea deal. He also admitted that while he was incarcerated, and before he implicated defendant in the crimes, he saw news reports that defendant had been charged with the same crimes as Harvey. Most damaging to Harvey’s credibility is the fact that he implicated his brother and cousin in the crime, but not defendant, even though he had seen defendant’s picture in a police photo array soon after the crimes. Harvey’s reason for not identifying defendant at the time was that he did not know everything about what happened. This was simply unbelievable given his role as the getaway driver. Even if it was a tenable excuse, it would not outweigh the other circumstances pointing to his unreliability, just as the explanation provided by one eyewitness in Santiago for failing to identify the defendant when the opportunity first presented itself was not enough to offset the insufficiency of the corroboration in that case. There, the eyewitness claimed that he did not tell the police he saw the defendant in the lineup because he wanted to remain unknown due to his immigration status (Santiago, 17 NY3d at 665). If that explanation was insufficient, Harvey’s certainly fares no better.
The fact that Harvey’s reliability is suspect based on his credibility, rather than the potential inaccuracy of his identification, does not affect the analysis.4 Where circumstances suggest the corroborating testimony is inaccurate — as was the case in Santiago — those circumstances weigh in favor of admitting the expert testimony, and failure to do so constitutes an abuse of discretion (Santiago, 17 NY3d at 673). Similarly, where the credibility of the corroborating witness is questionable, the trial court must consider this as tipping in favor of admission of the expert testimony. All the more so when the corroborating witness has an apparent incentive to misidentify the defendant, such as, for example, a plea deal where the witness’s testimony ensures that he avoids a murder conviction and a possible life sentence. Under the facts of this case, a jury could very well have found Harvey incredible, and relied solely *1171on J.J.’s identification testimony. This is exactly the type of uncorroborated, single witness case for which we have explained it is most appropriate that the jury hear expert testimony on factors impacting the accuracy and reliability of witness identification.
This does not mean that codefendant or coparticipant testimony can never provide corroboration, or that a beneficial plea deal renders such testimony “incredible.” “Whether a victim’s or other eyewitness’s identification of a defendant is sufficiently corroborated by other eyewitness identifications, so that the trial court need not proceed to the second stage of the LeGrand analysis, is dependent on the circumstances of the case” (Santiago, 17 NY3d at 671). Here, the facts undermining Harvey’s credibility compel the conclusion that his testimony lacks “the strong indicia” necessary to sufficiently corroborate J.J.’s eyewitness identification of defendant (see id.).
Furthermore, J.J.’s identification was subject to the types of factors that have an impact on eyewitness memory and accuracy. His observations of the four men involved in the crime were made under highly stressful conditions, which can affect memory (see People v Abney, 31 Misc 3d 1231[A], 2011 NY Slip Op 50919[U] [Sup Ct, NY County 2011] [on remand, permitting expert testimony on event stress]; Deffenbacher et al., A Meta-Analytic Review of the Effects of High Stress on Eyewitness Memory, 28 Law & Hum Behav at 699). J.J. watched as Dotson was murdered, and he himself survived apparently because the gun pointed to his head misfired. J.J. had his head down for most of the time and was pistol-whipped early during the robbery, minimizing the opportunity to observe his assailants. Also, defendant was a stranger to J.J. — as were the other intruders.
Since the People’s case depended on J.J.’s identification testimony, which was uncorroborated, the trial court should have proceeded to the second stage of the LeGrand inquiry (Santiago, 17 NY3d at 669). Here, there is no question that the testimony is “relevant to the witness’s identification of defendant,” “on a topic beyond the ken of the average juror,” and that defendant proffered an expert witness facially qualified (see id.). Indeed, in another case the same judge qualified the identification expert originally offered by defendant (see People v Norstrand, 35 Misc 3d 367, 373 [Sup Ct, Monroe County 2011]). Furthermore, because the court failed to grant a Frye hearing, I assume that the matters identified by the defendant *1172are encompassed within generally accepted scientific principles (Oddone, 22 NY3d at 379). Thus, the expert should have been permitted to testify.
Lastly, the error was not harmless, as the proof of defendant’s guilt was not overwhelming (People v Crimmins, 36 NY2d 230, 242 [1975]). Indeed, there was no physical evidence that tied defendant to the crime, and only J. J. identified defendant as one of the men inside the barbershop. For the reasons I have explained, the defendant is entitled to a new trial and the Appellate Division order should be affirmed.
Chief Judge DiFiore and Judges Pigott, Stein and Garcia concur; Judge Rivera dissents and votes to affirm in an opinion in which Judges Abdus-Salaam and Fahey concur.Order reversed and case remitted to the Appellate Division, Fourth Department, for consideration of the facts and issues raised but not determined on the appeal to that Court, in a memorandum.
. Harvey claimed that his attorney at the plea told him to say he was the driver.
. Contrary to the majority’s contention, LeGrand did not simply list factors to be considered or ignored by courts without guiding standards and structure (majority mem at 1161). Rather, LeGrand established a legal framework with specific areas of inquiry, to be applied by courts when determining the admissibility of expert testimony on eyewitness identification (see Santiago, 17 NY3d at 669 [LeGrand “set out standards governing the discretion of trial courts in regard to the admission of expert testimony on eyewitness identification (and) established a two-stage inquiry for considering a motion to admit such testimony”], citing LeGrand, 8 NY3d at 452).
. The defendant sought to admit expert testimony concerning the impact on eyewitness recognition of
“exposure time (the amount of time available for viewing a perpetrator affects the witness’s ability to identify the perpetrator); cross-racial and cross-ethnic inaccuracy (non-Hispanie Caucasian eyewitnesses are generally less accurate in identifying Hispanic people than in identifying other non-Hispanic Caucasians); weapon focus (a victim’s focus on the weapon used in an assault can affect ability to observe and remember the attacker); lineup fairness (similarity of fillers to the suspect increases identification accuracy); lineup instructions (police instructions indicating that the police believe the perpetrator to be in the lineup increase the likelihood of false identification); forgetting curve (the rate of memory loss for an event is greatest right after the event and then levels off over time); postevent information (eyewitness testimony about an event often reflects not only what the witness actually saw but also information the *1168witness obtained later); wording of questions (eyewitness testimony about an event can be affected by how questions put to the witness during investigation are worded); unconscious transference (eyewitnesses sometimes identify as the culprit an individual familiar to them from other situations or contexts); simultaneous versus sequential lineups (witnesses are more likely to make mistakes when they view simultaneous lineups than when they view sequential lineups); eyewitness confidence issues (an eyewitness’s confidence level is not a good predictor of eyewitness accuracy, but eyewitness confidence is the major determinant in whether an identification is believed by jurors), and confidence malleability (eyewitnesses’ confidence levels can be influenced by factors unrelated to identification accuracy)” (Santiago, 17 NY3d at 666-667 [emphasis omitted]).
. I do not agree with the Appellate Division’s conclusion that Harvey’s memory was unreliable based on factors impacting Harvey’s ability to view defendant and that defendant was a stranger. The record establishes that Harvey observed defendant for a period of time the evening of the robbery.