Memorandum.

[1, 2] The order of the Appellate Division should be affirmed. It was an abuse of discretion for the trial court to admit the results of DNA analysis conducted using the forensic statistical tool (FST) without first holding a Frye hearing (People v Williams, 35 NY3d 24 [2020]; People v Foster-Bey, 35 NY3d 959 [2020]). Here, however, this error was harmless. The evidence of defendant's guilt was overwhelming. Video footage from a security camera inside the store was entered into evidence at trial, including footage from one camera trained on a display{**38 NY3d at 1012} shelf which captured a group of men holding defendant against the shelf. The other men then scatter, leaving the video frame, at which point defendant places an item on the shelf directly in front of him before he too runs out of the frame. After approximately two minutes and 15 seconds, during which no one approaches the shelf or the area where defendant placed the item, a police officer looks at the space on the shelf where the item was placed, walks over, and removes a gun. Rather than "mere physical proximity," the video shows that only defendant could have placed the item—the gun recovered minutes later—on the shelf, not "any of the several others in the same area" (dissenting op at 1016-1017). Therefore, there is no significant [*2]probability that the jury would have acquitted defendant had it not been for this error (Williams, 35 NY3d at 42; People v Crimmins, 36 NY2d 230, 241-242 [1975]). As a result, we need not reach defendant's remaining arguments concerning discovery of materials related to the FST.

Rivera, J. (dissenting).

A unanimous Court agrees with defendant that, under our prior holdings in People v Williams (35 NY3d 24 [2020]) and People v Foster-Bey (35 NY3d 959 [2020]), the trial court abused its discretion as a matter of law in admitting the forensic statistical tool (FST)-derived DNA results without first holding a Frye hearing. We part company on the impact of that admission on the jury's verdict and disagree with the majority that the error was harmless. The evidence of defendant's guilt of criminal possession of a weapon was not overwhelming and the FST DNA evidence was the strongest evidence of possession against him. Therefore, even under the nonconstitutional standard, there was a significant probability that the error infected the verdict and, accordingly, was not harmless.

Defendant was convicted by a jury of various counts of criminal possession of a weapon for a gun found between boxes on a deli store shelf during an attack on defendant by several unidentified individuals.^[FN1] According to the evidence at trial, shortly after defendant and another person entered the store, several individuals began shouting and attacking them. The assault was recorded on the store's video camera, which captured several men tackling, pushing, and punching defend{**38 NY3d at 1013}ant as they pinned him against the shelves in front of the deli counter. All the men were clustered close to defendant and the shelves. It also appears that, while defendant was held down against the shelves by several of his attackers, one of them attempted to stab and slash defendant several times.

Minutes after the attack commenced, in response to a store employee's 911 call that a group of men was in the store and that two of them had guns, the police arrived. The officers observed several men fighting with defendant and pinning him to the shelves. One man pushed defendant and ran out past the officers. Defendant fell back while the other men scattered.

The officers testified they could not see defendant's or most anyone else's hands. After isolating defendant, and after one of the assailants reentered and attempted to punch defendant in the face, one of the officers pulled a gun from between two boxes on one of the shelves where the men had attacked defendant. The officer testified that she believed that defendant had been in possession of the gun because he was dressed in black, which was consistent with the description of the person described in the employee's phone call and communicated via the police radio. The evidence also showed that at least two other men were also dressed in black, including the man who appeared to be stabbing defendant in the video recording. The officers arrested defendant, who was visibly bleeding and had suffered a cut to his hand and head which required medical attention.

The prosecution's theory was that defendant physically held the gun at some point and that he attempted to hide it on the store shelf. However, no eyewitness observed defendant in possession of the gun at any time in the store, and there were no fingerprints or blood on the weapon. For its physical proof, the prosecutor relied heavily on FST-derived DNA results.^[FN2] According to the criminalist's testimony, a standard contact/skin cell{**38 NY3d at 1014} DNA analysis^[FN3] determined that defendant was not a major contributor to the DNA collected from the gun, although he could not be excluded as one of three possible contributors. However, based on the FST analysis, the criminalist concluded that

"[t]he DNA mixtures found on the sample from the swab from the trigger [of the gun] is approximately 4.57 million times more probable if the sample originated from Levan Easley and two unknown unrelated persons than if it originated from three unknown unrelated persons. Therefore, there is very strong support that Levan Easley and two unknown unrelated persons contributed to this mixture rather than three unknown unrelated persons."

In summation, the prosecutor extensively discussed the FST DNA evidence, reiterating that the criminalist had testified that for "the DNA mixture found on the sample [it] is approximately 4.57 million times more probable that the sample originated from the defendant Levan Easley and two unknown unrelated persons than if it originated from three unknown unrelated persons."

The Appellate Division affirmed defendant's conviction (see 171 AD3d 785 [2d Dept 2019]). Thereafter, we held in Williams and Foster-Bey that it was error to admit low copy number and FST DNA evidence without first holding a Frye hearing to determine whether the methodology had been accepted as reliable by the scientific community. A Judge of this Court granted defendant leave to appeal (see 35 NY3d 1093 [2020]). We now conclude [*4]that the same error identified in Williams and Foster-Bey occurred here. The only question remaining is whether the error in defendant's case was harmless.^[FN4]

Under the nonconstitutional standard, an error cannot be harmless—even when "the proof of the defendant's guilt, without reference to the error, is overwhelming"—if "there is a {**38 NY3d at 1015}significant probability, rather than only a rational possibility, in the particular case that the jury would have acquitted the defendant had it not been for the error or errors which occurred" (People v Crimmins, 36 NY2d 230, 241-242 [1975]). Thus, in order for a nonconstitutional error to be harmless, a two-pronged test must be satisfied: first, the evidence of guilt must be overwhelming, and, second, there must be no significant probability that, without the erroneously admitted evidence, the jury would have reached a different verdict. Evidence is overwhelming if it compels a guilty verdict, rather than merely provide some support for the prosecution's case. As the Court explained when first articulating New York's standard for nonconstitutional error:

"That 'overwhelming proof of guilt' cannot be defined with mathematical precision does not, of course, mean that the concept cannot be understood and applied in individual cases, although not always without some difficulty. It surely does not invite merely a numerical comparison of witnesses or of pages of testimony; the nature and the inherent probative worth of the evidence must be appraised. As with the standard, 'beyond a reasonable doubt', recourse must ultimately be to a level of convincement. What is meant here, of course, is that the quantum and nature of proof, excising the error, are so logically compelling and therefore forceful in the particular case as to lead the appellate court to the conclusion that 'a jury composed of honest, well-intentioned, and reasonable [people]' on consideration of such evidence would almost certainly have convicted the defendant" (id. at 241-242 [emphasis added]).

In Williams, we concluded that the admission of the DNA evidence without first holding a Frye hearing was harmless error because the evidence of the defendant's guilt was overwhelming. In that case, the prosecution "presented video evidence of the shooting, . . . eyewitness testimony identifying defendant as the shooter, and . . . testimony of defendant's former girlfriend with respect to the events that followed th[e] incident—including the girlfriend's account of defendant's handling of the subject gun and the forced secretion of that device" (35 NY3d at 42). In Foster-Bey, we similarly concluded that the error "was harmless in light of the eyewitness testimony as to the shooting and defendant's admission to his involvement in that incident" (35 NY3d at 961).

{**38 NY3d at 1016}Here, the erroneous admission of the FST DNA results was not harmless. The evidence of defendant's criminal possession of the gun was not overwhelming. There was no eyewitness who saw defendant in possession of the gun, no admission of his guilt, and no video recording depicting him holding the gun at any time. The evidence merely established that defendant entered the store, where he was attacked by several men and pinned against the shelf where the police later found the gun. Although there was evidence that, during the fight, someone yelled that a man dressed in black had a gun, and there was evidence that two men might have had guns, no one identified defendant as the man in black with the gun, and there was no evidence of a second weapon in the store. No eyewitness, including the responding officers who arrived minutes after the fight began, testified that defendant was holding a gun before or after the attack. No fingerprints or blood were recovered from the gun, or the shelf and boxes where the gun was found, even though defendant was not wearing gloves, and, at the time of his arrest, the police observed that he was bleeding from lacerations to his head and hand.

Despite the underwhelming nature of the prosecution's evidence, the majority concludes the evidence is "overwhelming" based solely on its review of the video footage. The majority makes three assertions in sequential order, and its analysis is contingent on the first of those assertions: that the recording shows "defendant plac[ing] an item on the shelf directly in front of him" (majority mem at 1012). The video does no such thing. Indeed, not even the [*5]District Attorney maintains such a view of the evidence. The District Attorney has consistently argued only that, during the chaotic scene, with several men fighting and pinning and pushing defendant up against the store shelf where the gun was later found by an investigating officer, the video shows defendant "reach[ing] between two of the boxes in front of the deli counter." When pressed, counsel admitted that he could not "see a gun in [defendant's] hand in the video."

Next in the majority's analysis is the statement that nobody came near the shelf after "[t]he other men" had "scatter[ed]" (id.) from the shelf where defendant had allegedly placed an item and before the officer later removed the gun. However, the video recording clearly shows several people near the location of the gun—namely, the group of men who assaulted defendant. The majority disregards that the District Attorney{**38 NY3d at 1017} argues that the relevant portion of the video depicts defendant having "reached" for the shelf "during the fight" when he was surrounded by the group of men assaulting him, not after the men scattered. But mere physical proximity of defendant to the weapon found, without evidence that distinguishes defendant as the person holding the gun from any of the several others in the same area, with access and opportunity to put the gun on the shelf, is not overwhelming proof that defendant possessed the gun.^[FN5]

Finally, to bridge the gap in direct evidence, the prosecution relies on an inference that "[n]obody else could have put the gun there, and that's why there is harmlessness in this case," as does the majority, holding that "only defendant could have placed the item—the gun recovered minutes later—on the shelf" (id.). While a jury may accept such an inference, our harmless error standard prevents us from calling such circumstantial evidence overwhelming. Indeed, this Court has no authority to make an independent assessment of the recording and so the majority cannot rely for its harmless error conclusion on its factual finding that defendant can be seen putting an object, later found to be the gun, on the shelf. Our jurisdiction is "limited to the review of questions of law" (NY Const, art VI, § 3 [a]). Our role is to accept the facts developed in the record below; "where there are conflicting inferences to be drawn from the proof, the choice of inferences is for the trier of the facts. And that choice is to be honored unless unsupported, as a matter of law" (People v Leonti, 18 NY2d 384, 390 [1966]).

Since there was no overwhelming evidence of guilt, the analysis ends here because an error cannot be harmless absent such proof. However, if we proceed to the second prong, the{**38 NY3d at 1018} harmful effect on the verdict is obvious. Still, the majority announces, without explanation, that there is no significant probability that the jury would have acquitted defendant without the FST DNA evidence (see majority mem at 1012). This conclusion disregards the impact of the FST DNA evidence in a case without any eyewitness to defendant's alleged possession and where the prosecution had no evidence of defendant's fingerprints or blood on the gun.

As for the other DNA evidence, the criminalist testified that, based on skin cell DNA testing of the gun's trigger mechanism, she could not exclude defendant as one of three possible contributors, but he was not the major contributor, meaning he was not the person whose DNA was present in amounts larger than the other two potential contributors. The prosecutor did not present DNA evidence of the attackers, leaving open the possibility that any of them could have contributed to the mixture obtained from the gun. Further, the gun was found on a shelf where several men were in close proximity as the chaotic, violent episode unfolded. To overcome the fact that the evidence easily suggested that any one of the attackers may have placed or dropped the gun on the shelf, the prosecutor [*6]emphasized for the jury the FST-derived DNA profile. Thus, the FST DNA results, presented rhetorically as a 4.57-million-to-one ratio that defendant's DNA was on the gun, transformed a thin case of unlawful possession based on supposition and inferences into a scientifically supported guilty verdict against defendant. With those numbers, how likely is it that a reasonable juror would find that defendant's DNA was not on the gun? It defies logic to conclude, as the majority apparently does, that the jury could ignore that this evidence left only an infinitesimally small likelihood that the DNA belonged to someone else.

Thus, I dissent from the majority's unsupported, terse holding that the error was harmless because it ignores the record before us and is contrary to our consistent application of the harmless error doctrine.

Chief Judge DiFiore and Judges Garcia, Singas and Cannataro concur; Judge Rivera dissents in an opinion, in which Judges Wilson and Troutman concur.

Order affirmed, in a memorandum.