HALLIGAN, J.

Thomas Leith, for appellant.

Bradley W. Oastler, for respondent.

OPINION

HALLIGAN, J.

This case concerns an increasingly prevalent issue: when may someone who is not an eyewitness to a crime testify to a jury that the defendant is the person depicted in a photo or video. We hold that such testimony may be admitted where the witness is sufficiently familiar with the defendant that their testimony would be reliable, and there is reason to believe the jury might require such assistance in making its independent assessment. Here, there was no showing that the proffered witness was sufficiently familiar with the defendant to render his testimony helpful, or that the jury faced an obstacle to making the identification that the witness's testimony would have overcome. Accordingly, we reverse.

I.

On June 10, 2015, police cameras in Syracuse captured a grainy video of a man running through the street and firing three shots into a van. The van promptly drove off, and the responding officers recovered two bullet casings but did not see the shooter.

A grand jury indicted defendant Farod Mosley for the shooting in September 2015, but the indictment was dismissed as legally insufficient. In July 2016, as the People were preparing to re-present the charges against Mosley to a grand jury, an assistant district attorney showed the video of the shooting to Detective Steven Kilburn. Kilburn identified Mosley as the shooter in the video and repeated his identification to a grand jury, which indicted Mosley.

The trial took place in February 2018, and the key issue was identification of the shooter in the video. Though one of the van's passengers testified that he remembered the shooting, he said that he had not seen the shooter. And though an analyst testified as to the recovered bullet casings, she stated that they had not been tested for either fingerprints or DNA. The video therefore was the only evidence tying Mosley to the crime.

At trial, the People played the video and relied on Kilburn to provide lay, non-eyewitness identification testimony that he believed Mosley was the shooter in the video . Kilburn first testified as to his familiarity with Mosley during voir dire, out of the presence of the jury. Kilburn, a homicide detective, explained that he met Mosley on January 5, 2016, when Mosley was brought into the precinct as a suspect in a different crime.

Kilburn thus estimated that by the time of the trial for the instant offense in February 2018, he had known Mosley for a year to a year and a half. When asked to describe his interactions with Mosley, Kilburn said he had "sat in rooms with Mr. Mosley," "walked side by side with Mr. Mosley," and "had the occasion to speak with him," as well as having viewed both police-related and Facebook photos of Mosley. He estimated he had been in the same room as Mosley on "a number" of occasions, but gave only one specific date-January 5, 2016, the first date he could recall seeing Mosley face-to-face. He could not ever recall having "street interactions" with Mosley. He also specified that at the time the assistant district attorney asked him if he could identify the shooter in the surveillance video in July 2016, he knew that Mosley had been arrested on a warrant for a "shots fired and weapons charge."

The judge concluded that Kilburn had "an extensive basis of knowledge" to identify Mosley in the video. To avoid airing prejudicial information about other police interactions before the jury, the judge instructed Kilburn to avoid mentioning the unrelated criminal investigation for which he had arrested Mosley.

Kilburn accordingly told the jury that he met Mosley about seven months after the video was captured during his routine "canvassing" of the Syracuse neighborhood where the shooting occurred, had known him for about a year and a half, and that he had "interacted" with Mosley, "walked" with him, and spoken with him on a "couple" of occasions. He testified that he was familiar with Mosley's "body language," "body type," and "build." He then viewed the video and identified the shooter as Mosley. He explained that though he did not know Mosley at the time of the shooting and did not know what Mosley had been wearing that day, he based his identification on his interactions with Mosley, his "build," the "shape of his nose," and "on previously viewing the video and being able to zoom in and stuff." Though he referenced Mosley's nose, he conceded when shown screenshots of the video "the face is a blur" and there was no nose apparent at all. Kilburn also opined that Mosley's appearance had not changed, and that Mosley "as he sits there now is the same as when I first encountered him, which is the same as he appears in that video."

During deliberations, the jury requested to see the video again and a readback of Kilburn's testimony. They also requested-and were denied-a magnifying glass with which to view the video. The jury ultimately convicted Mosley of two counts of criminal possession of a weapon in the second degree (Penal Law § 265.03 [1] [b], [3]) and reckless endangerment in the first degree (id. § 120.25).

Mosley argues that the trial court abused its discretion in admitting Kilburn's testimony. The Appellate Division (with two Justices dissenting) rejected this contention, holding that the People demonstrated Kilburn was more likely than the jury to correctly identify Mosley in the video. Because the People failed to establish that Kilburn's testimony would aid the jury in making an independent assessment regarding whether the person in the video was Mosley, we reverse.

II.

We have twice before considered the admission of lay non-eyewitness identification testimony: in People v Russell, 79 N.Y.2d 1024 (1992), and in People v Sanchez, 21 N.Y.3d 216, 225 (2013). In Russell, the trial court allowed four lay witnesses-the defendant's roommate, his roommate's mother, his landlord, and a friend-to identify the defendant as the person caught in surveillance photographs committing a bank robbery (People v Russell, 165 A.D.2d 327, 329 [1991], affd 79 N.Y.2d 1024). The People presented evidence that the defendant had disguised himself by growing an "uncharacteristic" beard that he sported during the robbery, and that he changed his appearance afterward by shaving it (id.). Stressing two considerations-the "personal knowledge" that the witnesses had "of defendant's appearance as of the time when the photographs were taken," as well as the fact that defendant had made the jury's task "more onerous" by creating impediments to identification (Russell, 79 N.Y.2d at 1025)-we held that the identification testimony was properly admitted. In reaching this conclusion, we referenced two federal cases, United States v Robinson, 804 F.2d 280, 282 (4th Cir 1986) and United States v Farnsworth, 729 F.2d 1158, 1160 (8th Cir 1984). Both of these cases construed Rule 701 of the Federal Rules of Evidence (lay opinion testimony) and focused on the two inquiries underpinning our conclusion in Russell: the extent of the witnesses' familiarity with the defendant, and whether the defendant had made the jury's job harder through use of a disguise or change in appearance. Similarly, in Sanchez we concluded that the admission of lay non-eyewitness testimony was proper because the proffered testimony was from two detectives who had known the defendant "from prior occasions" (People v Sanchez, 95 A.D.3d 241, 249 [2012], affd 21 N.Y.3d 216), and it was "undisputed" that the defendant's appearance had changed between the crime and the trial (Sanchez, 21 N.Y.3d at 225).

Due to the widespread deployment of surveillance in public spaces and the ubiquity of private cameras and video recording devices, lay opinion identification testimony has become an increasingly common form of evidence (see George Bach, Moderating the Use of Lay Opinion Identification Testimony Related to Surveillance Video, 47 Fla St U L Rev 445, 447 n 5 [2020] [tracking the multifold increase in the annual average of state cases discussing lay opinion identification testimony related to surveillance video since 2010]; Bennett Capers, Crime, Surveillance, and Communities, 40 Fordham Urb LJ 959, 961-962 [2013] [tracking the expansion of surveillance technology in New York City]).

Correspondingly, lay opinion identification testimony and the issues surrounding its admissibility have garnered increased attention from courts around the country (see e.g. State v Gore, 342 Conn 129, 269 A.3d 1 [2022]; State v Sanchez, 247 N.J. 450, 255 A.3d 1118 [2021]; Glenn v State, 302 Ga 276, 806 S.E.2d 564 [2017]; Lenoir v State, 222 So.3d 273 [Miss 2017]; People v Thompson, 2016 IL 118667, 49 N.E.3d 393 [2016]).

We first note that, despite the increasing prevalence of lay non-identification eyewitness testimony, its use presents several challenges. Courts have historically been wary of lay opinion testimony, which may "usurp the fact-finding function of the jury" (United States v Garcia, 413 F.3d 201, 210-211 [2d Cir 2005]). Although use of such testimony may be warranted in certain circumstances, courts must be vigilant that the testimony does not unnecessarily invade the jury's role (see United States v LaPierre, 998 F.2d 1460, 1465 [9th Cir 1993] [though "sometimes permissible," the lay witness identification "ran the risk of invading the province of the jury"]; see also Kulak v Nationwide Mut. Ins. Co., 40 N.Y.2d 140, 148 [1976] [in the context of expert testimony, noting testimony that "intrude(s) on the province of the jury to draw inferences and conclusions" is "improper" absent jury handicap]). Trial testimony by a law enforcement officer may pose additional concerns by drawing attention to a defendant's prior interactions with the police, and efforts to mitigate any prejudice by omitting reference to a law enforcement officer's occupation or providing a cryptic basis for familiarity may constrain the opportunity for vigorous cross-examination (see Russell, 165 A.D.2d at 335; see also United States v Walker, 974 F.3d 193, 206 [2d Cir 2020]; United States v Pierce, 136 F.3d 770, 776 [11th Cir 1998]; United States v Calhoun, 544 F.2d 291, 295 [6th Cir 1976]).

Whether the probative value of such identification testimony outweighs these drawbacks is a decision committed to the sound discretion of the trial court, and reviewable by this Court for abuse of discretion. As we held in Russell, the relevant question is whether the proffered testimony "serve[s] to aid the jury in making an independent assessment regarding whether the [person in the surveillance footage] was indeed the defendant" (79 N.Y.2d at 1025). To answer in the affirmative, there must be "some basis for concluding that the witness is more likely to correctly identify the defendant from the photograph [or video] than is the jury" (Farnsworth, 729 F.2d at 1160; see also Russell, 165 A.D.2d at 333).

We did not elaborate on the contours of this inquiry in Russell or Sanchez. Most other jurisdictions that have considered the question apply a totality of the circumstances test considering whether the witness is familiar enough with the defendant that it is reasonable to believe their testimony would be helpful and reliable, and whether there is reason to believe the jury might require that assistance (see e.g. Gore, 342 Conn at 150-151, 269 A.3d at 14 [surveying cases]). Our review of these cases, as well as the cases arising in the Appellate Divisions, indicates that the issue of lay non-eyewitness identification testimony arises in a variety of circumstances and often requires courts to make fact-intensive determinations. Accordingly, we too adopt a totality of the circumstances test, which preserves the discretion of the trial court-the court that, being closest to the facts of the case, is in the best position to make these evidentiary decisions . We note that this standard is fully consistent with our reasoning in Russell and Sanchez and the underlying Appellate Division opinions that we affirmed in those decisions.

First, the court must determine whether "the witness has had sufficient contact with the defendant to achieve a level of familiarity that renders the lay opinion helpful" (United States v Fulton, 837 F.3d 281, 297-298 [3d Cir 2016], citing United States v Beck, 418 F.3d 1008, 1014 [9th Cir 2005]). "If the witness lacks such familiarity, it is the province of the jury to draw [its] own conclusions regarding the identity of the person depicted without the witness's assistance" (Commonwealth v Vacher, 469 Mass. 425, 441, 14 N.E.3d 264, 279 [2014]), and that is the end of the inquiry.

In making this assessment, courts may consider the witness's general level of familiarity with the defendant's appearance (Russell, 165 A.D.2d at 329; see also Gore, 342 Conn at 151, 269 A.3d at 14; Thompson, 2016 IL 118667 ¶ 44, 49 N.E.3d at 15-16; United States v Dixon, 413 F.3d 540, 545 [6th Cir 2005]; Pierce, 136 F.3d at 774; Robinson, 804 F.2d at 282), and whether the witness's familiarity spanned an extended period of time and variety of circumstances (Russell, 165 A.D.2d at 329; see also Gore, 342 Conn at 152, 269 A.3d at 14-15; Sanchez, 247 N.J. at 470-471, 255 A.3d at 1129; Walker, 974 F.3d at 205; Beck, 418 F.3d at 1015; Pierce, 136 F.3d at 774). Courts may also take into account whether the witness was familiar with the defendant's appearance at the time the surveillance footage was taken (Russell, 165 A.D.2d at 336, 329; see also Gore, 342 Conn at 151, 269 A.3d at 14; Thompson, 2016 IL 118667 ¶ 46, 49 N.E.3d at 404; Sanchez, 247 N.J. at 472, 255 A.3d at 1130; Walker, 974 F.3d at 205; Beck, 418 F.3d at 1015; Farnsworth, 729 F.2d at 1160). Also relevant is the witness's familiarity with the defendant's customary manner of dress or clothing on the day of the surveillance footage (Russell, 165 A.D.2d at 329; see also Gore, 342 Conn at 151, 269 A.3d at 14; Thompson, 2016 IL 118667 ¶ 46, 49 N.E.3d at 404; Beck, 418 F.3d at 1015; Dixon, 413 F.3d at 545; United States v Jackman, 48 F.3d 1, 5 [1st Cir 1995]). Along similar lines, courts should also consider whether the witness references a specific trait the defendant has (such as a distinctive gait, scar, or tattoo), and importantly, whether they identify that trait in the surveillance footage (United States v Williams, 396 F App'x 516, 517 [10th Cir 2010] [identification based on "style of walk"]).

The second consideration is whether the jury needs the witness's assistance (see e.g. Fulton, 837 F.3d at 297; Vacher, 469 Mass. at 442, 14 N.E.3d at 279). Here, courts should consider whether the defendant's appearance changed between the time of the surveillance footage and the trial or whether the defendant disguised himself in the surveillance footage (see Russell, 165 A.D.2d at 329 [defendant sported "an uncharacteristic growth of facial hair" the day of the crime and shaved immediately after]; see also Gore, 342 Conn at 151, 269 A.3d at 14; Sanchez, 247 N.J. at 472, 255 A.3d at 1130; Thompson, 2016 IL 118667 ¶ 47, 49 N.E.3d at 404; Beck, 418 F.3d at 1015; Dixon, 413 F.3d at 545; United States v White, 639 F.3d 331, 336 [7th Cir 2011]; Jackman, 48 F.3d at 5; Farnsworth, 729 F.2d at 1160). In addition, courts may consider the quality of the surveillance footage and the extent to which the subject is clearly captured in the frame (see e.g. Gore, 342 Conn at 151, 269 A.3d at 14; Thompson, 2016 IL 118667 ¶ 48, 49 N.E.3d at 404-405; White, 639 F.3d at 336; Dixon, 413 F.3d at 545; Jackman, 48 F.3d at 5). Where an image is clear enough-that is, not so crystal clear that the jury could identify a person as capably as any witness, and not so obscure that no witness could readily identify the individual-this factor may weigh in favor of admission (see e.g. Gore, 342 Conn at 164-165, 269 A.3d at 22, quoting Jackman, 48 F.3d at 5; Sanchez, 247 N.J. at 473, 255 A.3d at 1130-1131; Fulton, 387 F.3d at 298).

Finally, we note that before admitting lay non-eyewitness identification testimony, a court should inquire as to the basis of the witness's familiarity outside the presence of the jury in a separate hearing or voir dire, as the court properly did here. The party offering the witness-in most cases the People-bears the burden of establishing that their testimony would both be helpful and necessary. We note, however, that it is incumbent on both parties to create a thorough record to aid the court in its determination and to allow for meaningful appellate review. Additionally, as a best practice, it would be appropriate for the trial court to provide cautionary jury instructions, both at the time of the testimony and during the final charge, explaining to the jury that lay non-eyewitness identification testimony is mere opinion testimony that they may choose to accept or reject, and reminding the jurors that because they are the finders of fact, it is their opinion as to whether the defendant is depicted in the surveillance footage that matters (see Sanchez, 21 N.Y.3d at 225 [noting that "the trial court issued appropriate limiting instructions to the jury"]; People v Lee, 214 A.D.3d 457, 458 [1st Dept 2023] [noting court's "repeated instructions reminding the jury that it was its function to determine who was depicted on the video"]; People v Laroc, 203 A.D.3d 1176, 1177 [2nd Dept 2022] ["appropriate limiting instructions" given both during testimony and the court's final charge]).

III.

We now turn to the case at hand and conclude that it was an abuse of discretion to admit Kilburn's testimony.

First, the People did not establish that Kilburn was sufficiently familiar with Mosley to render his identification helpful to the jury. When addressing this issue during voir dire, Kilburn testified only that he had "sat in rooms with Mr. Mosley," "walked side by side with Mr. Mosley," and "had the occasion to speak with him." Though Kilburn represented that he had interacted with Mosley "a number" of times, he failed to specify how many times, for how long, or what the nature of those interactions were. Indeed, it is unclear from the record whether Kilburn interacted with Mosley on more than one day (see People v Mosley, 200 A.D.3d 1658, 1161 [2021] [Whalen, P.J. and Lindley, J., dissenting] [characterizing Kilburn as having recalled "only a single day on which these interactions took place"]). Though Kilburn also referenced looking at photographs of Mosley, this is not a proper basis for familiarity (see Pierce, 136 F.3d at 774; see also Sanchez, 247 N.J. at 471-472, 255 A.3d at 1130; LaPierre, 998 F.2d at 1465). Far from having interactions with Mosley in a variety of circumstances, Kilburn testified that he had never had "street" interactions with Mosley, and it is unclear whether he ever interacted with Mosley outside of the police station. Kilburn was not familiar with Mosley's appearance at the time the shooting occurred; he did not meet him until seven months later. Nor did Kilburn know what Mosley was wearing on the day of the shooting.

Moreover, Kilburn did not connect any distinctive traits of Mosley's to the person depicted in the surveillance photo. Before the jury, Kilburn stated that he was familiar with Mosley's "build," "body language," "body type," and the "shape of his nose." Contrary to best practice, though, the People did not elicit these grounds for familiarity before the judge when laying the foundation for the admission of Kilburn's testimony, which would have avoided potential prejudice and allowed for robust cross-examination on that foundation. Even aside from this point, the assertions fail to support the utility of Kilburn's testimony. Kilburn did not specify what is unique about any of Mosley's features, or connect those features to his identification of the image in the video. Indeed, when asked to look at a screenshot of the video, Kilburn acknowledged that due to the poor quality, there was no nose apparent on the captured face at all.

Second, the People failed to demonstrate that the jury needed Kilburn's help. The shooter in the video did not wear a disguise, and there is no indication that Mosley's appearance changed. If, as Kilburn testified, Mosley "as he sits there now... is the same as he appears in that video," there was no reason to think that the jury would not be equally well positioned to identify the defendant based on his appearance in the courtroom. As for the surveillance footage, the testimony tends to suggest that the quality of the video here would make it difficult for anybody to reliably identify the subject. Kilburn testified that stills of the video were so blurry that it was impossible to make out a face, and that his identification was based on being able to "zoom in" and view the video multiple times-which, rather than establishing that he was better equipped to identify the defendant in the surveillance footage than the jury, demonstrated that his identification was largely based on exactly the same evidence that the jury had before it. In short, there was no basis to conclude, based on the totality of the circumstances here, that Kilburn would be more likely to correctly identify the defendant from the video than the jury.

The admission of Kilburn's testimony in this case illustrates the challenges of cross-examination inherent in lay opinion identification testimony from law enforcement. For example, during voir dire, Kilburn testified that he could not recall ever having "street interactions" with Mosley. Yet to avoid prejudicing Mosley at trial, at the judge's instruction, Kilburn told the jury that he knew Mosley from "canvassing" in the area around where the shooting took place. Not only did this statement overstate Kilburn's familiarity with Mosley by implying that he had interacted with Mosley in more natural settings, but the defense was unable to thoroughly cross-examine Kilburn about this aspect of his familiarity. As other courts have acknowledged, "the choice between full cross-examination and possibly revealing the defendant's criminal history is a difficult one" (Farnsworth, 729 F.2d at 1162), and so we too "caution trial courts to admit this kind of identification testimony only in limited and necessary circumstances with all appropriate safeguards" (Pierce, 136 F.3d at 776; see also Russell, 165 A.D.2d at 334-335; United States v Butcher, 557 F.2d 666, 670 [9th Cir 1977]).

Finally, the error in admitting Kilburn's identification was not harmless. We start by asking whether "the evidence is overwhelming in the first instance," because "unless the proof of the defendant's guilt, without reference to the error, is overwhelming, there is no occasion for consideration of any doctrine of harmless error" (People v Mairena, 34 N.Y.3d 473, 484 [2019] [internal quotation marks omitted], quoting People v Crimmins, 36 N.Y.2d 230, 241 [1975]). Here, the evidence is far from overwhelming: The only evidence tying Mosley to the crime was the low-quality video and Kilburn's identification testimony (see People v Mosley, 200 A.D.3d 1658 [2021] [Whalen, P.J. and Lindley, J., dissenting]). Without Kilburn's testimony, the People presented no more evidence than they did in their first indictment, which was dismissed for insufficient evidence. The error thus cannot be deemed harmless and requires reversal for a new trial.

Accordingly, the Appellate Division order should be reversed and a new trial ordered.

RIVERA, J. (concurring):

Video surveillance recordings are regularly proffered in criminal cases to establish a defendant's guilt. Clear images of the perpetrator of course aid the trier of fact because, where identity is at issue, such images may be compared with the defendant in court. That was not the case in this appeal, where it is undisputed the video was low-quality. Indeed, nothing connected defendant to the crimes charged except the testimony of a police officer who had no personal knowledge of the crime or defendant's guilt, but who identified defendant as the person shown in a blurry video image shooting at a van. I concur with the majority that this testimony did not aid the jury in determining whether defendant was the shooter in the video, that admission of this testimony was not harmless error and that the Appellate Division should be reversed and a new trial ordered (see majority op at 2, 16). I also agree with much of the majority's analysis of the challenges and dangers that noneyewitness testimony poses, especially where, as here, the witness is a law enforcement official (id. at 7-8). However, I do not adopt the majority's totality of the circumstances test because it partially relies on irrelevant and subjective factors which may hinder the truth-seeking process.

Therefore, I write separately to suggest a better approach consistent with the notion that a video is of no assistance in the truth-seeking process where the image is of such poor quality that the trier of fact cannot discern the perpetrator's physical characteristics. First, such videos are inadmissible for identification purposes as a matter of law. Second, where the video quality supports its admissibility, the trial court still must impose strict limitations on noneyewitness opinion testimony describing the video to ensure that the trier of fact independently determines whether the video depicts the defendant. Such opinion testimony is inadmissible when: (1) the image is clear enough for the trier of fact to compare with their own observations of the defendant at trial; and (2) the defendant has not changed or disguised their appearance to avoid the comparison. However, a noneyewitness may provide testimony connecting the defendant to the image in the video only when the witness has personal knowledge of a unique feature of the defendant's appearance that is not obvious in court but which the witness can point to in the video.

* * *

To be admitted, evidence must be relevant, and it must be more probative than prejudicial (People v Telfair, ___ N.Y.3d ___, 2023 NY Slip Op 05965 at *3 [2023], citing People v Ely, 68 N.Y.2d 520, 538 [1986]). Because the trier of fact is solely charged with determining a defendant's guilt, evidence must assist the jury in its deliberative process (People v Russell, 79 N.Y.2d 1024, 1025 [1992]; People v Huertas, 75 N.Y.2d 487, 493 [1990]). Thus, where a photograph or video is of such poor quality that the images are blurry and indecipherable, a court should not allow admission for identification purposes even if the photo or video may be helpful to the jurors' understanding of the scene or the surrounding events of the crime. The majority apparently would allow testimony in those cases where a blurry, nondescript image is used for identification (see majority op at 11-12). I cannot agree because its totality-of-the-circumstances approach increases the risk of misidentification by increasing the risk that a jury will simply defer to the prosecution's suggestions that the image is that of the defendant.

When a court permits the use of a photo or video to identify the defendant as the perpetrator, a person who did not witness the crime is in almost all cases similarly-situated to members of the jury in observing that evidence and determining what it depicts. The person's familiarity with the defendant's appearance is irrelevant since the video and photo "speak for itself" and the jury can simply turn to the defense table a few feet away and compare the image to the defendant.

There are two narrow exceptions. Such testimony is admissible where the defendant's appearance has changed between the time of the crime and the time of trial, such as in Russell, where defendant shaved his beard immediately after the crime (79 N.Y.2d at 1025). The testimony is also permissible where the defendant has some defining feature either in appearance or gait that is not visible to the jury-such as, for example, a limp, or a distinctive item of clothing gifted to defendant by the non-percipient witness (People v Rivera, 170 A.D.3d 566, 567 [1st Dept 2019] [defendant changed his appearance and non-percipient witnesses "recognize(d) defendant's mannerisms and peculiar way of walking"]). In such cases, the noneyewitness must point to this defining characteristic in the video so that the jury can then decide what weight to give this testimony. To the extent the majority relies on these cases to support its proposed balancing test, these citations are inapt and should be limited to cases involving a change in a defendant's appearance.

For example, if the witness observed unique physical features of the defendant before and close in time to the crime, such as a distinctive bone structure, hair and eye coloring, facial hair, and tattoos which have changed over time, or which appear altered or intentionally masked in the footage to avoid detection, then the witness can testify to the defendant's appearance at the time of the offense even if it differs from their appearance at trial. Or, if the video showed the perpetrator had a particular gait, the witness could testify they had observed the defendant walk in the same unique manner. Also, if the image in the video shows the perpetrator wearing particular clothing known by the witness to have been worn by the defendant-such as a sweater knitted by the witness and which they gave to the defendant-then they can testify that the clothing in the video worn by the perpetrator is the same one they have observed defendant wear in the past. By contrast, a person cannot testify that, in their opinion, a blurry video depicts the defendant based on nothing more than their memory of having observed the defendant at some prior point in time and comparing that memory to the video image and the defendant's appearance in court. That is true regardless of their relationship and interaction with the defendant. To the extent the majority leans into this "factor," I disagree that the continuum of familiarity is particularly relevant (see majority op 9-10). Such a witness provides no additional information to jurors who can compare the video to their own in-court observations of the defendant.

When the testimony comes from a police detective, as here, their status as a law enforcement official creates an even greater risk of unduly influencing the jurors. Data shows that most jurors in criminal cases believe police testimony (Vida B. Johnson, Bias in Blue: Instructing Jurors to Consider the Testimony of Police Officer Witnesses with Caution, 44 Pepp L Rev 245, 246 [2017]). The jury may believe that the officer's investigative training means they have perceptive abilities superior to those of the average person. "Jurors are likely to recognize law enforcement officers' enhanced perception in discerning criminal behavior, and may therefore tend to defer to officers' opinions as to the defendant's criminal activity" (Deon J. Nossel, Admissibility of Ultimate Issue Expert Testimony by Law Enforcement Officers in Criminal Trials, 93 Colum L Rev 231, 246 [1993]). The jury may also speculate that an officer observed the defendant on prior occasions as part of the officer's law enforcement duties. "[T]he jury may infer that the officer has special knowledge about the defendant" and, "as a number of courts have acknowledged, the jury may be led to believe that the officer's testimony is supported by information known to the officer but not presented to the jury. The jury may therefore tend to give more credence to the testimony than is warranted" (Nossel at 245). In other words, the jury may assume that the police official has access to information suggestive of the defendant's guilt.

Non-percipient witness testimony from law enforcement also adversely impacts the defendant's right to present a defense (see Crane v Kentucky, 476 U.S. 683, 690 [1985]). As here, a defendant's cross-examination of the witness is limited and thus a defendant cannot fully explore weaknesses in the identification. Unlike with other non-percipient witnesses, the "cross-examination might be unfairly limited, for example, by the nature of the relationship between a criminal defendant and a law enforcement witness. In this context, the defense attorney's decision not to cross-examine fully is not a tactical choice; rather, it is the only choice-tantamount to no choice" (George Bach, Moderating the Use of Lay Opinion Identification Testimony Related to Surveillance Video, 47 Fla St U L Rev 445, 467 [2020]). The Sixth Circuit's observations in United States v Calhoun aptly describe the problem: "The main defect in permitting [defendant's parole officer] to testify was that his broad assertion could not be tempered or probed by cross-examination. The defendant could not explore the possible motives his parole officer might harbor in positively identifying him as the robber" (544 F.2d 291, 295 [1976]).

Non-percipient identifications of defendants in low-quality video or photo evidence, absent particularized identifying evidence, also increase the risk of mistaken identification. When videos are of poor quality, non-percipient witness identification "surreptitiously tends to bolster the video itself. A grainy, pixelated, or otherwise poor video might be of little help at all to a jury, but adding the lay witness may have the effect of encouraging jurors to see more in the video than they really can" (Bach, 47 Fla St U L Rev at 459). "[M]isidentification results in more wrongful convictions than all other causes combined, which corresponds with studies showing juries believe the witness making an identification up to eighty percent of the time, regardless if correct" (Dakota Kann, Admissibility of First Time in-Court Eyewitness Identifications: An Argument for Additional Due Process Protections in New York, 39 Cardozo L Rev 1457, 1464-1465 [2018]). The potential risk is greater in cases of cross-racial identification (People v Boone, 30 N.Y.3d 521, 528 [2017], citing C.A. Meissner & J.C. Brigham, Thirty years of investigating the other-race effect in memory for faces: A meta-analytic review, 7 Psychology, Public Policy, and Law 3, 15 [2001] ["(S)ocial scientists have found that the likelihood of misidentification is higher when an identification is cross-racial" and a meta-analysis of studies of the phenomenon found that individuals were "1.56 times more likely to falsely identify" individuals in cross-racial identifications]).

These concerns are not limited to photographs and videos. Advancements in new technologies like facial recognition software to assist in identification of perpetrators of crime may increase the risk of misidentifications and wrongful convictions (see e.g., Aziz Z. Huq, Racial Equity in Algorithmic Criminal Justice, 68 Duke LJ 1043, 1079 [2019] [citing studies that show facial recognition technology varied in its accuracy based on the identity of subjects, including being less accurate in identifying Black people, women, and younger people]; Dorothy E. Roberts, Digitizing the Carceral State: Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor, 132 Harv L Rev 1695, 1718 [2019], citing Clare Garvile et al, The Perpetual Line-Up: Unregulated Police Face Recognition in America, Ctr on Privacy & Tech at Georgetown Law [2016] ["A comprehensive investigation of police department facial recognition... found that the databases likely contain a disproportionate number of images of (B)lack people and that the software may be especially inaccurate in recognizing (B)lack people's faces"]).

Here, applying the approach I have described, while the video may have provided the jury with some understanding of how the underlying events unfolded, it should not have been admitted for purposes of identifying the defendant as the shooter because of its low quality. The perpetrator's face is not clearly visible in the grainy video, and the image does not depict any unique physical features which the jury could compare to their observations of defendant at trial. Compounding the error, once the court permitted the video to be played for the jury, the court allowed the prosecution to bolster this low-quality video evidence with opinion testimony of a non-percipient, police detective witness that the indecipherable image shows defendant. There was no evidence defendant altered or masked his appearance to avoid identification, and the witness did not testify that he had personal knowledge of any unique characteristic of defendant that he could point to in the video which the jury could not observe in the courtroom.

Moreover, the facts of this case demonstrate how non-percipient identification by law enforcement substantially increases the risk of a wrongful conviction. The first grand jury indictment was dismissed for legally insufficient evidence. A few days before the second grand jury, the prosecutor asked the detective to look at the video to support making an identification. There was nothing entered into evidence to indicate that the identification procedures were non-suggestive. The detective's identification of defendant, at the request of the prosecution, transformed this case from being legally insufficient to indict as a matter of law into a conviction, based solely on the officer's opinion that he could clearly see defendant in a blurry, low-quality video image, when that was not possible with the naked eye. Indeed, the jury requested and was denied a magnifying glass to view that same image. Whether the video depicts defendant as a matter of fact was solely within the province of the jury to decide and the officer's lay opinion did not aid in that determination.

Order reversed and a new trial ordered.

Chief Judge Wilson and Judges Garcia, Singas, Cannataro and Barros concur.

Judge Rivera concurs in result in an opinion.

Judge Troutman took no part.