Notice: This opinion is subject to formal revision before publication in the Atlantic and Maryland Reporters. Users are requested to notify the Clerk of the Court of any formal errors so that corrections may be made before the bound volumes go to press.

DISTRICT OF COLUMBIA COURT OF APPEALS

No. 23-CF-1078

ANDREW C. PATRICK, APPELLANT,

V.

UNITED STATES, APPELLEE.

Appeal from the Superior Court

of the District of Columbia

(2022-CF3-005665)

(Hon. Lynn Leibovitz, Trial Judge) (Submitted June 5, 2025 Decided September 18, 2025)

Jason K. Clark was on the brief for appellant.

Michael E. McGovern, Assistant United States Attorney, with whom Edward R. Martin, Jr., United States Attorney at the time the brief was filed, and Chrisellen R. Kolb, Gregory Evans, and Omeed Assefi, Assistant United States Attorneys, were on the brief, for appellee.

Before DEAHL and HOWARD, Associate Judges, and WASHINGTON, Senior Judge.

DEAHL, Associate Judge: Caprice Nicholas was carjacked while filling up her car at a gas station one morning. As Nicholas stood outside her Chevy Suburban near the fuel pump, a man wearing a jacket and with his hood up approached her

2 with a gun and demanded her car keys. Their interaction lasted for about one minute, culminating with Nicholas handing over her car key and her assailant driving off with her car. Nicholas was predominantly focused on the gun during the robbery, worried that her assailant might shoot her, but at “the very end” of their interaction she turned her attention to his face.

Several hours later, and about four miles from where the carjacking occurred, police arrested Andrew Patrick driving Nicholas’s car. The officers summoned Nicholas to the arrest scene to see if she could identify Patrick as her assailant. A friend drove Nicholas and as she approached—now about three hours from the carjacking—one of the arresting officers told her to “do like a slow drive-by to see if [you] see the person . . . that stole [your] truck,” and to “just look to your right and see if you notice the guy.” As Nicholas pulled up, police had positioned Patrick on her right, in handcuffs, flanked by two officers, and sitting on the curb within a few feet of her stolen Suburban. Nicholas positively identified Patrick as her carjacker. She repeated that identification when officers did a second drive-by with her, and then again in court. Patrick was convicted of carjacking and related offenses.

Patrick now appeals, arguing that the trial court should have granted his motion to suppress Nicholas’s identifications of him. In his view, the show-up procedures were so suggestive of his guilt—and so unnecessary given the ready

3 availability of doing a lineup or photo array since he had already been arrested—that Nicholas’s identifications of him had no assurance of reliability and their admission violated his due process rights. We agree.

This was an extraordinarily suggestive show-up identification, compounded by the fact that it was entirely unnecessary given the hours that had passed since the carjacking and the fact that Patrick was already under arrest. There is no readily apparent reason why the officers could not have utilized far more reliable identification procedures that did not lend themselves so heavily to false positive identifications. See Neil v. Biggers, 409 U.S. 188, 198 (1972) (“Suggestive confrontations are disapproved because they increase the likelihood of misidentification, and unnecessarily suggestive ones are condemned for the further reason that the increased chance of misidentification is gratuitous.”). There were no countervailing indicia that this was nonetheless a reliable identification uninfected by the suggestivity: Nicholas and Patrick were complete strangers, she had only briefly observed him with his hood up during a stressful encounter in which she was primarily focused on his gun, and her fairly generic description of her assailant omitted the most distinctive aspects of Patrick’s appearance.

The trial court committed constitutional error when it admitted Nicholas’s identifications of Patrick. We conclude that error requires reversal of Patrick’s

4 convictions for armed carjacking and one other weapons offense, though we affirm Patrick’s convictions for the other weapons-related offenses that were clearly not impacted by Nicholas’s identifications.

I. Factual and Procedural Background

Caprice Nicholas was at a gas station in Southeast D.C. at around 7:30 a.m. She had just gone inside the station to pay for gas, and as she returned to her car and prepared to add some gas to it, a man approached her. He was wearing a black jacket with white stripes running down its sleeves, a gray hood, a cross-body satchel, blue jeans, and gray sneakers. The gas station’s surveillance footage showed the man get very close and pin Nicholas against the tailgate of her car, where they stayed for around fifty seconds, though there is no audio in the video and it is tough to see what’s happening in any detail. Nicholas later explained in her testimony that the man pulled a gun from his satchel and brandished it, demanding that she give up her car keys. Nicholas testified that as the man was speaking to her, she was looking at “[j]ust the gun” and not “anything else” because she did not know “if he was going to shoot me.” Once Nicholas had her keys in hand, she spent some time “fumbling” with them to remove her house and work keys from the keyring so she could hold onto them. After that, she gave the assailant her car key and the man went and sat in the driver’s seat of the Suburban to start it up. Nicholas followed him

5 momentarily “to get a better look at him”—she “glanced” into the window for about one or two seconds and “saw a portion of his face” or “most of [his] left side.” As the man started the car, Nicholas ran inside the gas station to call 911, and the man drove off.

Police arrived at the scene within a couple of minutes and found Nicholas, who was “shaken up” and “crying.” She indicated that her assailant was a man with a medium-brown complexion,1 who was wearing a hood that was “gray or something” and a “satchel thing.” She did not mention any distinctive features of her assailant and was unable to describe what color shirt or pants he had on because she “wasn’t looking,” but said her assailant was “probably like thirty-seven [years old], maybe,” and a little taller than her. Nicholas also gave the officers her license plate number, which officers put out an alert for.

After one of the automated license-plate readers positioned at intersections around the city spotted the Suburban’s plate number, officers converged on the area and eventually located the car in Northeast D.C., about four miles from the

1

Nicholas did not identify her assailant’s race in either her 911 call or to responding officers. But in response to an officer’s question, “was he my complexion or his complexion”—referring to herself and another Black officer with a darker complexion—Nicholas said “like yours,” and the officer interpreted that in her own words as “medium brown.”

6 carjacking. Officers followed the Suburban and pulled it over within a few minutes. They then ordered the driver, the vehicle’s lone occupant, out of the car.

Patrick exited the car from its driver’s seat. He was a Black man with a full beard and mustache, two long braids, one dangling down to each shoulder, and was wearing a gray hoodie, gray New Balance sneakers, light blue jeans, and a red bandana on his head. He also happened to be thirty-seven years old, though that of course was not evident from his mere appearance. Officers handcuffed Patrick and sat him on the curb a few feet away from the Suburban, while Patrick claimed that somebody “just gave me this truck.” In the Suburban, police found Patrick’s phone and a satchel bag on the floorboard of the passenger seat, which contained a loaded gun, an additional magazine, and Patrick’s bank cards.

MPD Detective Cornel Kelemen, one of the officers on the scene, called Nicholas and asked her to come to their location. Nicholas arrived around forty minutes later—now nearly three hours after the carjacking itself. According to Nicholas’s testimony, Kelemen told her over the phone just before she arrived to “do like a slow drive-by to see if [you] see the person . . . that stole [your] truck,” i.e., to “come past and ride slowly, but just look to your right and see if you notice the guy.” At this point, Patrick was still in handcuffs and seated on the curb next to the Chevy

7 Suburban. Police officers were flanking him, but “not holding him.” Nicholas’s reaction upon driving by and seeing Patrick “was like, Yes, that’s him.”

Nicholas then exited the car and said “that’s him” to Kelemen. Kelemen then told her, “[j]ust for the record, get in the car with me, and we’re going to put you in the backseat to do another drive-by to see what it is.” A few minutes later, Nicholas got into the backseat of a police cruiser with Kelemen and another officer in the front, and the cruiser drove by Patrick for a second “ride-through.” As they approached, Kelemen told Nicholas “[h]e’s going to be on your left side.” Patrick remained handcuffed and was flanked by two officers at the time of this second— and in Kelemen’s words, more “formal”—part of the show-up. The officer pulled the vehicle to within about fifteen feet of Patrick, and Nicholas identified him again, although she noted on this go-around that she “thought he had short hair.”

Patrick was arrested and later indicted for armed carjacking and related firearms offenses. These included possession of a firearm during a crime of violence, unlawful possession of a firearm, carrying a pistol without a license, possession of an unregistered firearm, and unlawful possession of ammunition.

Patrick moved to suppress Nicholas’s on-scene identifications of him, arguing that the show-up process was “highly suggestive” of his guilt and thus led to “unreliable” identifications. Patrick emphasized that officers had placed him in

8 handcuffs and sat him—with nobody else but the officers who were flanking him in the area—right next to her vehicle. He further complained that officers specifically directed Nicholas to look at the side of the street where they had positioned Patrick in what he broadly described as an overly “police assisted” identification procedure. Those suggestive procedures and on-scene identifications also irrevocably tainted any prospective in-court identification, so that in Patrick’s view any in-court identification should have been precluded as well.

The trial court denied the suppression motion, concluding that the show-up was not “unduly suggestive in the circumstances” and was reliable. The trial court noted that although Nicholas was instructed to look at someone “seated near her car,” and that the circumstances of doing “two” show-ups was “somewhat more suggestive than one,” Nicholas nevertheless had a “strong” “opportunity to observe” the carjacker at the time of the offense. This came from the facts that Nicholas had “face to face” contact with her assailant for “a very long minute” while “[i]t was daylight,” and Nicholas had made a “deliberate effort at the very end to look at him” because “she knew she had spent some time focused on a gun” and “fumbling with her keys.” This “close opportunity to look at him” also offset any problem with respect to the show-up occurring “three hours” after the carjacking, which the trial court recognized was “longer than lots of show-ups in the case law.” The trial court further highlighted Nicholas’s certainty—she “showed no hesitation” and “clearly

9 recognized him”—as weighing against suppression, as did the fact that Patrick was precisely thirty-seven years old, just as Nicholas described her attacker’s age. The trial court thus permitted the government to introduce evidence about how Nicholas twice identified Patrick on the day of the carjacking, and the court likewise permitted Nicholas to identify Patrick as her assailant before the jury.

Aside from Nicholas’s identification testimony, and the rather incriminating fact that Patrick was found driving Nicholas’s car within hours of the carjacking, the government also presented fairly damning cell-site location information (CSLI) tracing Patrick’s cellphone. That CSLI data placed Patrick’s phone in the area of the gas station when Nicholas was carjacked and it showed that he was in the same location as the Suburban when the automated license-plate reader system first spotted the SUV at around 9:15 a.m. Police also identified Patrick’s DNA on the gun found in the satchel bag that they recovered from the Suburban, though his DNA was found amidst a “mixture of three individuals[’]” DNA.

The government’s closing arguments focused on this CSLI and DNA evidence as well as Nicholas’s identification of Patrick as the carjacker during the show-up and in court. The government stressed Nicholas’s identifications as one cornerstone of its case, highlighting that Nicholas had repeatedly identified Patrick as her assailant after “get[ting] a good look at him” during the carjacking itself:

10

The identification also include[s] that show up. You know

there was good weather. You can see that the sun was out;

you heard the testimony about that. No obstruction

whatsoever. MPD protocol is followed. She did not

waver with regard to her identification. She told

Detective Kelemen that’s him. He’s the one that took my

truck. And in court she told you the same; she identified

the defendant.

The jury convicted Patrick on all counts. The trial court sentenced him to sixteen years in prison and five years of supervised release. Patrick now appeals.

II. Analysis

Patrick argues on appeal that the trial court violated his Fifth Amendment due process rights in admitting the show-up identifications, and the in-court identification that flowed from it, because they were “impermissibly suggestive” and did not have sufficient independent indicia of reliability to overcome the taint of that suggestivity. We agree.

This was an extraordinarily suggestive identification procedure, with officers placing Patrick in handcuffs, immediately next to Nicholas’s stolen vehicle, and directing her to precisely where he would be. While we have tolerated similar levels of suggestivity when there was some necessity for a prompt on-scene identification that very well might have led to the suspect’s release (if the complainant indicated they were not the culprit), that is not the case here. Nearly three hours had already

11 passed, and Patrick had already been placed under custodial arrest long before Nicholas arrived on the scene and made her identification. There was no good reason for the officers not to have employed more reliable identification procedures, like compiling a photo array, so that Nicholas was not alerted to the fact that officers had found Patrick in her car and seemed fairly confident that he was the culprit. And Nicholas’s identification of Patrick did not have sufficient independent indicia for us to discount the suggestive procedures underlying her identifications—Patrick was a stranger to her, Nicholas got only a fleeting glance at his face by her own account, and her generic descriptions of her assailant omitted Patrick’s most distinctive features, to wit, his long braids and full beard. In the end, admitting these suggestive identifications offended due process.

We explain ourselves in three parts. First, we lay some groundwork and detail the relevant due process doctrine in this arena in greater detail. Second, we detail why this identification procedure was so unnecessarily suggestive. And third, we explain why Nicholas’s identification of Patrick did not have independent indicia of reliability sufficient to overcome the suggestivity of the identification process itself. After we’ve covered that constitutional question, we address the government’s argument that the erroneous admissions of these identifications were harmless. We agree with the government about that only in part.

12

A. Pretrial identifications under the Due Process Clause

The Fifth Amendment’s Due Process Clause protects criminal defendants from “the introduction of evidence” at a criminal trial that is “so extremely unfair that its admission violates ‘fundamental conceptions of justice.’” Dowling v. United States, 493 U.S. 342, 352 (1990) (quoting United States v. Lovasco, 431 U.S. 783, 790 (1977)). Eyewitness identifications of strangers pose a particularly high risk of such unreliability and unfairness. “The identification of strangers is proverbially untrustworthy,” and identification procedures are “peculiarly riddled with innumerable dangers and variable factors which might seriously, even crucially, derogate from a fair trial.” United States v. Wade, 388 U.S. 218, 228 (1967) (quoting The Case of Sacco and Vanzetti 30 (1927)). Despite this great risk of error, eyewitness identifications often hold tremendous sway over the factfinder and are commonly at the root of wrongful convictions. See id. at 236 (describing the “grave potential for prejudice” at trial stemming from pretrial lineups); see also Thomas D. Albright & Brandon L. Garrett, The Law and Science of Eyewitness Evidence, 102 B.U. L. Rev. 511, 515-16 (2022) (“Of the over 370 individuals in the United States who have been exonerated due to DNA evidence to date, approximately 70% involved eyewitness misidentifications.”).

13

The Supreme Court has developed a two-step framework for analyzing whether pretrial identification procedures comply with due process. Courts must first ask whether the procedure, under the totality of the circumstances, is “impermissibly suggestive.” Biggers, 409 U.S. at 197; see also Manson v. Brathwaite, 432 U.S. 98, 106 (1977). If it is, then courts must proceed to the second step of determining whether the circumstances nonetheless show that the identification was sufficiently “reliable” to disregard the suggestivity, an inquiry we will detail momentarily. Biggers, 409 U.S. at 199; Brathwaite, 432 U.S. at 106; see generally Morales v. United States, 248 A.3d 161, 170-72 (D.C. 2021) (describing Supreme Court jurisprudence on this issue). Let us elaborate on these two steps.

The first, suggestivity step of the analysis basically asks whether the police had unnecessarily telegraphed to the complainant who they believe the culprit is, thereby exerting some undue influence and pressure on the complainant to confirm police suspicions. Importantly, this inquiry bakes in considerable allowance for the necessities of the situation, and accounts for the fact that officers sometimes need to conduct on-scene identification procedures to speed along an investigation, a critical point we’ll elaborate on shortly. Sticking just to suggestivity for the moment, show- up identification procedures like the one employed here are one of “the most grossly suggestive identification procedure[s] now or ever used by the police.” Wayne R. LaFave, 2 Crim. Proc. § 7.4(g) (4th ed. Nov. 2024 update) (quoting P. Wall, Eye-

14 Witness Identification in Criminal Cases 28 (1965)). The suggestivity stems primarily from the fact that (1) show-ups convey a clear message that the police believe they have their guy, and (2) show-ups do not give the witness any other options to choose from, meaning all false positives will point to the suspect rather than to any filler participants in a lineup or photo array. Morales, 248 A.3d at 172-73 (“Asking a witness whether they can identify a standalone suspect—rather than picking him out from a lineup or photo array—is among the most suggestive and objectionable forms of pretrial identification because of the strong implication that the individual has been singled out as the suspect by law enforcement.”); Ex parte Appleton, 828 So. 2d 894, 899-900 (Ala. 2001) (Show-ups do not “give the witness a choice of identifying any other person as being the perpetrator of the crime charged.” (quoting Ex parte Frazier, 729 So. 2d 253, 254-55 (Ala. 1998))). Show- ups thus involve an intrinsic risk of generating misidentifications, particularly from witnesses who have a “faulty memory,” are “inclined to guess,” and/or are prone to believing the police would not be asking for an identification unless they were pretty sure they caught the culprit. Young v. State, 374 P.3d 395, 421 (Alaska 2016).

It is now well-established that “[i]t is inherently suggestive for police to present a single suspect in custody to a witness at a show-up.” Long v. United States, 156 A.3d 698, 708 (D.C. 2017). Decades ago, this court observed that showing a complainant a single picture of the suspect and asking whether they recognize the

15 guy—akin to a photographic show-up—was the “most suggestive” and “most objectionable method of pretrial identification” around. Patterson v. United States, 384 A.2d 663, 665-66 (D.C. 1978) (quoting United States v. Dailey, 524 F.2d 911 (8th Cir. 1975)). But we tend to think this show-up procedure was considerably more suggestive than that single-photo display. Officers did not merely show the complainant a picture of a potential suspect; they placed that suspect in handcuffs and sat him next to her stolen car with officers flanking him when they asked Nicholas to identify him.

Even still, and herein lies the rub, we have with some regularity approved of show-up identification procedures because they are justified by the necessities of fast-breaking investigations, and courts must consider the extent to which any suggestivity was necessary and justified under the circumstances. We have thus described “prompt” show-ups as serving the valuable if narrow purpose of allowing police to quickly “exonerate an innocent person who had been mistakenly apprehended” when the necessities of the situation call for it. Howard v. United States, 954 A.2d 415, 423 (D.C. 2008) (quoting Maddox v. United States, 745 A.2d 284, 292 (D.C. 2000)). When police have detained a suspect who they believe is the culprit, the complainant might indicate they have the wrong guy so that the suspect can be set free and the police can resume their search without delay. In short, it gives the police a chance “to determine whether they [are] on the right track” at the early

16 stages of an investigation when time is of the essence. Simmons v. United States, 390 U.S. 377, 384-85 (1968) (describing officers resorting to use a single-photo identification procedure in order to track down “at large” assailants). Even highly suggestive identification procedures can thus be justified by the necessities of the situation.2

But whatever allowance courts make for show-up procedures that are justified by the necessities of the situation, “unnecessarily suggestive [identification procedures] are condemned,” because in that case “the increased chance of misidentification is gratuitous.” Biggers, 409 U.S. at 198 (emphasis added); see also Perry v. New Hampshire, 565 U.S. 228, 238-39 (2012) (“[D]ue process concerns arise only when law enforcement officers use an identification procedure that is both suggestive and unnecessary.”).

2

The need for a prompt show-up might arise from factors aside from how early the officers are in their investigation. For instance, necessity might exist where the witness is about to become unavailable, as was the case in Stovall v. Denno, 388 U.S. 293 (1967). The Supreme Court approved the show-up identification in Stovall where officers brought the suspect to a complainant’s hospital bed. Id. at 302. As suggestive as that was, the Court explained that the complainant was the sole witness to the crime—“the only person in the world who could possibly exonerate” the suspect—and was not able to leave her hospital bed and was apparently near death, which rendered “the usual police station line-up . . . out of the question.” Id. “Faced with the responsibility of identifying the attacker” and “the need for immediate action,” the police could follow “the only feasible procedure” (taking the suspect to the hospital room for a show-up) consistent with due process. Id.

17

That brings us to the second, reliability step of the analysis. Even where an identification procedure is unduly suggestive, the underlying identification may nonetheless be sufficiently reliable for admission at trial when courts are confident that the identification did not flow from the suggestive procedures but independent sources. Morales, 248 A.3d at 181 (explaining that the inquiry is “akin to asking whether the identification has an independent source”). That is typically the case when (1) “the witness had previously identified” the suspect under non-suggestive circumstances, id. at 167, (2) the witness was “well-acquainted with the defendant before being exposed to the suggestive identification procedure,” id., or (3) when the victim has accurately described the suspected culprit in such peculiar detail that we can rest assured that they could identify them absent the suggestivity, see id. at 179; see also Cureton v. United States, 386 A.2d 278, 285-86 (D.C. 1978) (identification in highly suggestive circumstances was nonetheless reliable where victim had both previously identified the culprit and further accurately described his “race, sex, age, height and weight” and his “hair style, facial hair, skin color, clothing, and the type of weapon” used).

But we have cautioned that “‘the reliability inquiry may not include the consideration of evidence of the defendant’s guilt external to the identification’ itself.” Morales, 248 A.3d at 180 (quoting Long, 156 A.3d at 708). That is, even when the suspect is dead to rights and there is irrefutable proof of their guilt, such

18 extrinsic evidence is no point in favor of reliability of the witness’s identification of them. Id. That is because “[w]hether a particular piece of evidence is reliable is distinct from whether it happens to be accurate.” Id. Somebody entirely removed from the crime who has no basis whatsoever to identify a suspect as the culprit might nonetheless correctly guess who it is. That is not a reliable identification—you have no good reason to take that person’s word for it—just because it happens to be correct.

B. This identification procedure was steeped in unnecessary suggestivity

This was an extraordinarily suggestive identification procedure that, by any measure, was considerably more suggestive than even the standard show-up. Recall that the police had placed Patrick in handcuffs and sat him down right next to Nicholas’s stolen car in preparation for the identification. Detective Kelemen then called Nicholas and told her “to stop and . . . do like a slow drive-by” and “just look to your right and see if you notice the guy.” Patrick’s position in proximity to the car surely conveyed to Nicholas that police had found him driving the car and believed he was the culprit. Detective Keleman’s instructions further added to the suggestivity—they ran contrary to both MPD’s internal guidelines and generally accepted standards for speaking to witnesses before an identification procedure. MPD instructs their officers to “[t]ell the witness that the perpetrator may or may

19 not be present in the identification procedure.” Metropolitan Police Department’s General Order 304.07, Procedures for Obtaining Pretrial Eyewitness Identification (Apr. 18, 2013). And the New Jersey Supreme Court—in a widely cited opinion on the topic of pretrial identifications—has advised that “showup administrators should instruct witnesses that the person they are about to view may or may not be the culprit and that they should not feel compelled to make an identification.” State v. Henderson, 27 A.3d 872, 903 (N.J. 2011) (collecting recent social science research and issuing recommendations for conducting more reliable pretrial identification procedures).

The suggestivity of the show-up was also compounded by the fact the police conducted it twice, in close succession, as the trial court correctly recognized when it observed that “two” show-ups were “somewhat more suggestive than one.” That repetition created some added pressure for Nicholas not to change her mind after giving the police a positive identification. See Wade, 388 U.S. at 229 (“[I]t is a matter of common experience that, once a witness has picked out the accused at the line-up, he is not likely to go back on his word later on.” (internal quotation marks omitted)); see also Nancy K. Steblay & Jennifer E. Dysart, Repeated Eyewitness Identification Procedures With the Same Suspect, 5 J. Applied Rsch. in Memory & Cognition 284, 286 (2016) (“[W]itness confidence is elevated in eyewitnesses who see the suspect twice, regardless of whether the suspect is guilty or innocent.”).

20

We now turn our focus to the extent to which this suggestivity was reasonably necessary, or if instead police could have readily conducted a less suggestive identification procedure, such as a photo array or a lineup. See United States v. Brownlee, 454 F.3d 131, 138 (3d Cir. 2006) (finding show-up was unnecessarily suggestive when “there [was] no reason evident why . . . the witnesses could not have been taken to the police station for a less suggestive line-up or photo array”); Morales, 248 A.3d at 174 (finding identification procedure unnecessarily suggestive where “[t]he government had every opportunity to conduct an appropriate and timely pretrial photo array or a lineup procedure”); see also Perry, 565 U.S. at 238-39 (“[D]ue process concerns arise only when law enforcement officers use an identification procedure that is both suggestive and unnecessary.”); Biggers, 409 U.S. at 199 (rejecting “strict rule barring evidence of unnecessarily suggestive confrontations,” which may yet be admissible if sufficiently reliable).

The suggestivity here was entirely unnecessary because Patrick was already under custodial arrest long before this show-up identification procedure, and there was no chance that Patrick was going to be released at the scene even if Nicholas had failed to identify him as her carjacker. Nearly an hour before the first show-up identification, the police found Patrick with an unregistered and unlicensed loaded handgun while driving Nicholas’s stolen vehicle. The police immediately arrested him and thoroughly searched him incident to that arrest, at which point they further

21 uncovered what they believed to be “molly,” or MDMA, a Schedule I narcotic. So there was not a chance in the world that Patrick would be released at the scene, even if Nicholas had not identified him. The police were effectuating Patrick’s custodial arrest long before Nicholas’s arrival, and they then held Patrick captive and in handcuffs for the better part of an hour before any identification procedure took place.

When a suspect is already under custodial arrest like Patrick was, and there is no reasonable possibility that a witness’s failure to identify them will alter that, the central justification for permitting a show-up procedure—that it might result in the suspect’s release—vanishes. If Nicholas failed to identify Patrick, that might have conceivably dissipated the probable cause for Patrick’s arrest as to the carjacking charge in particular, but probable cause clearly would have persisted for Patrick’s serious felonies of possessing an unregistered and unlicensed firearm, and for the offenses of unauthorized use of a vehicle and the possession of illegal narcotics that officers had just found on him. That is not to say that there can be no necessity for a show-up when there is probable cause to arrest a suspect—our precedents are clear to the contrary. See Howard, 954 A.2d at 423-24 (rejecting appellant’s argument that a show-up was impermissible solely because police “already had probable cause to arrest”). The point is that Patrick was in fact under custodial arrest at the time of

22 this show-up, and nothing Nicholas could say was going to change that, so there was no meaningful necessity underlying this show-up procedure.

Rather than detaining Patrick on the roadside for the better part of an hour to await Nicholas’s arrival, officers instead could have invited or escorted her down to a nearby police station or precinct where they could have conducted a far more reliable identification procedure like a photo array. See Brownlee, 454 F.3d at 138. This show-up was not tethered to the types of necessity that our precedents have recognized as justifying show-up procedures in the past. The show-up was neither a “prompt” one performed in order to quickly “exonerate an innocent person who had been mistakenly apprehended” in the earliest stage of an investigation, Howard, 954 A.2d at 423 (quoting Maddox, 745 A.2d at 292), nor was it in response to a medical emergency such as a dying witness, see Stovall, 388 U.S. at 302. In these circumstances, “it [was] little to ask that law enforcement take some additional time,” assuming any additional time was required at all, to “conduct a less suggestive identification procedure.” Brownlee, 454 F.3d at 138 n.4 (concluding show-up was unnecessarily suggestive). Nothing necessitated or justified them in skipping more reliable methods.

The government’s main counter is that this court has previously found that show-ups “conducted under nearly identical circumstances” were not unnecessarily

23 suggestive. We disagree with that description. We have found no case in which we have approved of a show-up procedure that was so far removed in time and place from the underlying crime, see infra section II.C., nor have we found any other case in which the suspect was clearly under custodial arrest regardless of whether the witness identified him. The government’s best case is Diggs v. United States, which held that a show-up after a carjacking was not unnecessarily suggestive even though—fairly similar to the facts here—the police told the eyewitness that they “got” the perpetrators and the show-up occurred while the suspect was handcuffed and in view of the carjacked vehicle. 906 A.2d 290, 300-01 (D.C. 2006).

We do not find Diggs compelling because that identification was much closer in time to the offense and nothing about its facts or analysis suggests that both suspects’ custodial arrests were faits accomplis. The police in that case detained two suspects riding in an apparently carjacked vehicle, and neither of them appeared to have any illegal weapons or narcotics on them (or at least our opinion did not suggest any alternative grounds for arrest). Id. The police quickly summoned the complaining witness to the scene of the arrest and, within about thirty minutes of the carjacking itself, conducted a show-up. Id. at 294. If the victim indicated that either of those men did not participate in the carjacking, we see no obvious reason why

24 they might not have been released from the scene. So the necessity that animates the permissibility of show-up identifications existed in Diggs, but it is absent here.3

C. Nicholas’s identifications were not sufficiently reliable for admission

Even when an identification procedure is steeped in unnecessary suggestivity, however, the resulting identification may yet be admissible if it is sufficiently reliable. Brathwaite, 432 U.S. at 112 (when an “identification is reliable despite an unnecessarily suggestive [police] identification procedure,” exclusion “is a Draconian sanction . . . that may frustrate rather than promote justice”). So we must now assess whether Nicholas’s identification of Patrick was sufficiently reliable that

3

We further observe that our cases justifying show-up identifications on necessity grounds are arguably anachronistic, where ever-advancing technology has made it far easier and faster to produce suitable photo arrays, even at the scene of the arrest, within minutes if not seconds. See Henry F. Fradella, A Synthesis of the Science and Law Relating to Eyewitness Misidentifications and Recommendations for How Police and Courts Can Reduce Wrongful Convictions Based on Them, 47 Seattle U. L. Rev. 1, 10 & n.42 (2023) (describing how “software programs” that can be installed on ubiquitous “personal computers, laptops, and tablets” make photo arrays “far easier to administer today than in the past”); see also id. at 10, 84 (noting how such software can sometimes perform “facial matching” to quickly “select photos of” filler participants who are not the suspect); Miriam Jones, Mug Shot Lineups Leave the Polaroid Era, Gov’t Tech. (last visited Aug. 27, 2025), (describing how back in 2011, a Virginia police department implemented an app to generate photo arrays within five minutes), https://www.govtech.com/public-safety/ mug-shot-lineups-leave-the-polaroid-era.html; https://perma.cc/L7YU-FVBF. Those technological advancements make it increasingly difficult to say show-up identification procedures are necessary or justified, given the seemingly ready alternative of promptly creating a photo array that is far less suggestive. But we leave that topic for another day.

25 we can rest assured it stemmed from her familiarity with him, rather than the suggestive show-up procedures themselves. Id. at 114 (“[R]eliability is the linchpin in determining the admissibility of identification testimony.”).

Our reliability analysis considers factors including, but not limited to, “[1] the opportunity of the witness to view the criminal at the time of the crime, [2] the witness’ degree of attention, [3] the accuracy of [her] prior description of the criminal, [4] the level of certainty demonstrated by [her] at the confrontation, and [5] the length of time between the crime and the confrontation.” Biggers, 409 U.S. at 199-200; Morales, 248 A.3d at 176 (describing list of factors as “not exhaustive”). The first and fifth Biggers factors “typically drive the reliability analysis,” in part because those have the most support from the scientific community. Morales, 248 A.3d at 177 n.15 (citing research concluding that those factors are “the only two of the five” that are “clearly related to accuracy” (internal quotation marks omitted)); see also State v. Hunt, 69 P.3d 571, 576-77 (Kan. 2003) (adopting “refin[ed]” Biggers factors that retain focus on the witness’s opportunity to view and the “[s]pontaneity” of the identification). “Against these factors is to be weighed the corrupting effect of the suggestive identification itself.” Brathwaite, 432 U.S. at 114.

26

Considering those factors and principles, we have no sufficient assurance that Nicholas’s identification of Patrick did not stem from the suggestive identification procedures themselves.

Regarding the first and second factors, Nicholas had no previous familiarity with Patrick4—a total stranger—and she did not have a good opportunity to memorize his features. Her interaction with him lasted roughly one minute, during which she was under serious distress, and save for the last couple of seconds she was staring at “[j]ust the gun” and not “anything else” but the keys that she had been fumbling around with. Nicholas was emphatic on this last point, reiterating during her testimony that she was looking only at the gun during her interaction with Patrick, because she “didn’t know . . . if he was going to shoot [her].” This was classic “weapon focus,” which is known to “impair” witnesses’ “ability to make a reliable identification and describe what the culprit looks like if the crime is of short duration.” Henderson, 27 A.3d at 904-05 (citing multiple empirical studies); see

4

The Biggers factors seem to assume that the witness and the defendant are strangers. When that is not the case, the witness’s prior familiarity with the suspect is perhaps the strongest factor supporting a reliability finding—i.e., that the identification did not stem from the suggestive procedures themselves—even where the opportunity to observe the culprit at the time of the crime is relatively minimal. See Green v. United States, 580 A.2d 1325, 1327 (D.C. 1990) (stressing that it was “most unlikely” that a witness’s identification of the defendant stemmed from a suggestive procedure rather than the fact that the defendant was “a man he had known on a family basis for a period of more than six months”).

27 also Young v. Conway, 698 F.3d 69, 78-79 (2d Cir. 2012) (recounting social science literature and noting that “the presence of a weapon” and “stress of the situation” can damage eyewitness accuracy). Nicholas in fact told responding officers right after the carjacking that she was generally unable to describe her assailant’s clothing—not even the distinctive black jacket with white stripes that he was wearing—because she “couldn’t pay attention” and “wasn’t looking.”

It is true that it was daytime and that Nicholas made an effort to go around the driver’s side of the car to “get a better look” at the guy before he sped off, giving her about one or two seconds to look at her assailant’s face. But that brief glance at her carjacker gives us no assurance that the suggestivity of the show-up is not what ultimately animated Nicholas’s identification of Patrick. This case is quite similar to Morales. In that case, we concluded that a police officer’s minute-long interaction with a perpetrator in which the officer had “no view of his face” for most of that minute “approaches the nadir of a witness’s opportunity to observe.” 248 A.3d at 177-78. The officer in Morales saw the person’s face only during a “two-to-three second backward glance” as the officer chased him. Id. at 168. So just as Morales concluded that the “closely tied” first and second factors weighed against the identification procedure’s reliability, we reach the same conclusion here. Id. at 178.

28

Factor three, the accuracy of the witness’s prior description, does not meaningfully contribute to a finding of reliability either. Nicholas’s description of the carjacker was fairly generic, describing him as a man who was a bit taller than her and had a medium brown complexion. She did impressively nail her assailant’s age (assuming it was indeed Patrick) at “like thirty-seven, maybe,” though there was clearly a fair bit of luck involved in that precision, presuming Nicholas is not a carnival barker expert in guessing ages. Even still, Nicholas could not identify any distinctive characteristics of her assailant, and she in fact omitted the two most distinctive characteristics that jump out from a quick view of Patrick: his full beard and his long braids. Nicholas’s description of her assailant’s attire likewise lacked much specificity. She did not mention the distinctive red bandana that Patrick had on when arrested, which the government maintains he was wearing during the carjacking, see infra note 6, and she more generally could not describe what her assailant was wearing because she explained she was too focused on the gun and her keys to pay much attention to that. There are the notable exceptions that Nicholas mentioned her carjacker was wearing a “gray or something” hood, and had a satchel, and indeed Patrick had a gray hoodie on and a satchel was in the car with him. But overall Nicholas’s descriptions of her assailant were generic and omitted the most distinctive features of Patrick’s appearance and attire, so they do not give us

29 confidence that her identification of Patrick was not influenced by the suggestivity of the show-up.

To be clear, we share the trial court’s assessment that Nicholas’s belief that her assailant had short hair, while Patrick had two long braids, was not so exculpatory given that Nicholas’s assailant was wearing a hood that might have hidden the braids. It was similarly not so exculpatory that Nicholas failed to notice or describe Patrick’s beard or the distinctive red bandana that he perhaps had on, infra note 6—witnesses in high stress situations understandably can miss a lot of details, or fail to relay all of the details that they did notice. But whether Nicholas’s descriptions of her assailant tended to exculpate Patrick, or undercut her later identification of him, are simply not the relevant questions. The relevant question is instead whether her description was so detailed in its particulars that it alleviates the plain threat that the suggestive procedures themselves led Nicholas to identify Patrick. See Morales, 248 A.3d at 179 (analyzing how prior description was “sparse on specifics” and missing “common suspect descriptors”). It was not.

As for the fourth factor, Nicholas’s “certainty,” we do not assign any weight to this factor on either side of the scale. Nicholas expressed some equivocation, explaining that she thought her assailant had short hair, but otherwise seemed pretty firm in her conviction that Patrick was her carjacker. And yet, we have explained

30 why this factor is of minimal importance when the certainty is “expressed only after the suggestive” identification procedure, and would thus “seem to be a byproduct of” that suggestivity. Id. Some states have more “formally abandon[ed]” the witness certainty factor altogether after considering the growing body of scientific evidence that expressions of confidence are not correlated with accuracy. See State v. Discola, 184 A.3d 1177, 1188-89 (Vt. 2018); see also, e.g., Brandon L. Garrett, Eyewitnesses and Exclusion, 65 Vand. L. Rev. 451, 468-71 (2012) (noting that “confidence is not highly correlated with accuracy” and “[t]he way that police construct the lineup can enhance confidence”); Benn v. United States, 978 A.2d 1257, 1267-71 (D.C. 2009) (discussing similar studies in expert witness context). But we need not go so far here, as it suffices to say, under the circumstances, that it does not have any weight in our reliability calculus.

Factor five, regarding “the length of time between the crime and the confrontation,” also weighs against any finding of reliability. As we noted in Morales, “[t]he longer the gap in time between the sighting and the identification procedure, the better the opportunity to observe must be to provide any assurance of reliability.” 248 A.3d at 177-78. And here, the gap of around three hours between the carjacking and the show-up is plainly significant relative to Nicholas’s one- to two-second glance into the driver’s window (or even as compared to the minute in which she was in her carjacker’s presence). Three hours is a longer period of time

31 than the gaps present in seemingly all of our show-up cases—nearly double the high- water mark for show-ups we have approved of in the past—indicating that this was anything but a prompt show-up. See, e.g., Singletary v. United States, 383 A.2d 1064, 1068-69 (D.C. 1978) (gap of “a few minutes”); Howard, 954 A.2d at 418 (fourteen minutes); Diggs, 906 A.2d at 294 (roughly thirty minutes); Kaliku v. United States, 994 A.2d 765, 782 (D.C. 2010) (thirty minutes); Maddox, 745 A.2d at 293 (one hour); Lyons v. United States, 833 A.2d 481, 486 (D.C. 2003) (hour and fifteen minutes); Long, 156 A.3d at 709 (hour and forty-five minutes); Parker v. United States, 333 A.3d 1162, 1184 (D.C. 2025) (hour and forty-five minutes).

We do not consider this gap in time in a vacuum under factor five. We consider it in the particular context of a show-up identification procedure, where studies have shown that once the gap in time between the incident and the show-up exceeds two hours, the reliability of the underlying identification drops dramatically. One experimental study found that although show-ups and six-person lineups resulted in a similar number of false identifications when performed within minutes of a sighting, show-up witnesses became four times as likely to make a false identification when the gap in time became longer than two hours. A. Daniel Yarmey et al., Accuracy of Eyewitness Identifications in Showups and Lineups, 20

32 L. & Hum. Behav. 459, 459, 465 (1996) (adding that “[m]ost psycho-legal researchers” consider show-ups to be “highly prejudicial” to suspects).5

All of this is to say that the show-up in this case—which was steeped in unnecessary suggestivity—would require far stronger independent indicia of reliability to permit its admission consistent with due process. The Biggers reliability factors weigh heavily against admission. It is further telling that the first and fifth factors weigh against admission, considering that those factors are the most closely tied to identification accuracy as an empirical matter. See Morales, 248 A.3d at 177 n.15. And the government, for its part, does not draw our attention to any additional, unenumerated factors that might otherwise demonstrate this show-up was reliable. All considered, the show-up conducted in this case was “unreliable at its

5

This is not to overstate the reliability of even prompt show-ups: One recent meta-analysis compared show-ups that were conducted immediately to lineups conducted after a delay of more than twenty-four hours, and discovered that those delayed lineups had fewer false identifications than the immediate show-ups. Jeffrey S. Neuschatz et al., A Comprehensive Evaluation of Showups, in 1 Advances Psych. & L. 43, 49 (Monica K. Miller & Brian H. Bornstein eds., 2016). So “[e]ven in those situations for which showups might be thought to have a memorial advantage . . . showups were worse.” Id. at 63.

33 core,” id. at 180, and the trial court erred in failing to suppress it and the resultant in-court identification. 6

D. Harm

The government argues that even if the trial court should have suppressed Nicholas’s identifications of Patrick, the admission of those identifications was harmless. In making this argument, the government concedes, per our precedents, that we must apply the demanding harm standard articulated in Chapman v. California, 386 U.S. 18 (1967). See Ellis v. United States, 941 A.2d 1042, 1048-51 (D.C. 2008) (applying Chapman to question of whether unconstitutional show-up required reversal). Chapman mandates reversal unless the error is “harmless beyond a reasonable doubt.” 386 U.S. at 24. To meet this high bar, the government must show that there is no “reasonable possibility that the evidence complained of might have contributed to the conviction.” Id. at 23 (quoting Fahy v. Connecticut, 375 U.S. 85, 86-87 (1963)). This is an “exacting standard” that asks, in similar words, “whether the guilty verdict actually rendered in this trial was surely unattributable

6

Patrick argues that Nicholas’s in-court identification should not have been permitted because it stemmed from the problematic show-ups, and the government does not contest that point. So we conclude that none of the identifications could be admitted at trial. See, e.g., Morales, 248 A.3d at 172 (suppressing in-court identification that was the fruit of suggestive pretrial mugshot display).

34 to the error.” Ellis, 941 A.2d at 1048-49 (quoting Sullivan v. Louisiana, 508 U.S. 275, 279 (1993)).

We first readily conclude that the error here was harmless as to Patrick’s convictions for the unlawful possession of a firearm, carrying a pistol without a license, possession of an unregistered firearm, and unlawful possession of ammunition. Nicholas’s identifications of Patrick as her carjacker could not have possibly had any bearing on those convictions. The government presented unrefuted evidence that, prior to and independent of the show-up, Patrick possessed a firearm and an additional magazine when he was stopped in the vehicle. Most critically, police found Patrick with a loaded handgun and magazine within his reach (in a satchel containing bank cards in his name) and discovered that the handgun contained traces of his DNA. It was further undisputed that Patrick could not lawfully possess a firearm or ammunition due to a prior conviction and that he had never registered or obtained a license to carry the handgun. The harm from the show-up—which was the substantial likelihood that Nicholas misidentified Patrick as her carjacker—could not have affected the jury’s view of that distinct bucket of evidence. We therefore affirm this set of convictions.

Patrick’s convictions for carjacking and possession of a firearm during a crime of violence, or PFCV, present a closer question. As to them, the government’s harm

35 argument rests principally on the overall strength of its case, and its case was indeed very strong: (1) Patrick was found in the stolen car a couple of hours after the carjacking, (2) surveillance footage showed the carjacker wearing blue jeans and gray shoes, 7 similar to the ones Patrick was wearing when stopped, (3) Patrick’s cell phone location data placed him at the scene of the carjacking at the relevant time, and (4) Patrick was found with a gun, in a satchel, that had his DNA on it. This was close to a slam dunk prosecution, and Patrick did not mount much of a defense to it.

And yet, the question for us “is not whether the evidence was ultimately overwhelming against a defendant, but instead whether the government has established ‘beyond a reasonable doubt that the error complained of did not contribute to the verdict obtained.’” Morales, 248 A.3d at 184 (quoting Chapman, 386 U.S. at 24). Under that exceedingly high bar, we cannot conclude beyond a reasonable doubt that Nicholas’s identifications of Patrick did not contribute to the

7

The government asserts that surveillance footage depicts the carjacker wearing a red bandana as well (similar to the one Patrick was wearing when stopped), a point that the government and defense counsel disputed in their closing arguments. Notably, if the government were right about that, it would further cut against the reliability of Nicholas’s identification given that she failed to notice such a distinctive feature of Patrick’s attire that partially covered his forehead. But as to the harm question, we have reviewed that video exhibit multiple times, and we think it is fair to say that it is unclear whether it depicts the carjacker wearing a bandana under his hood, so we doubt it would have played any material role in the jury’s assessment of the evidence.

36 jury’s verdicts on those counts, for two related reasons: (1) the fact that juries find eyewitness identifications to be one of the most persuasive types of evidence; and (2) that the government relied substantially on Nicholas’s identification testimony as a centerpiece of its case.

First, the Supreme Court has described the “grave potential for prejudice” at trial that can stem from suggestive pretrial procedures. Wade, 388 U.S. at 236. Given the dramatic power of an eyewitness’s identification of the culprit, pretrial rulings on the admissibility of those statements “might well settle the accused’s fate and reduce the trial itself to a mere formality.” Id. at 224, 235-36 (“The trial . . . may well not be that in the courtroom but that at the pretrial confrontation.”). So even where the government presents an otherwise compelling case, some jurors might call it a day and rest their verdicts entirely on such positive identifications, without even considering the remainder of the evidence.

Second, and relatedly, the government featured Nicholas’s show-up and in- court identifications heavily when making its case in front of the jury, no doubt because it is aware of how heavily juries weigh those identifications. The government stressed the centrality of this identification testimony through the presentation of its case. For instance, it presented three separate witnesses— Nicholas, Detective Kelemen, and another officer—about how the show-up

37 procedures were conducted. It presented three permutations of Nicholas’s identification of Patrick, stemming from the first drive-by, the second drive-by, and the in-court identification processes. Importantly, in its closing argument, the government presented the carjacking case as coming down to the question of identification and highlighted Nicholas’s identification as one of its key pieces of evidence. It took the time in closing to describe the show-up in some detail and stated definitively that Nicholas “told Detective Kelemen that’s him. He’s the one that took my truck.” And the government added that “in court she told you the same; she identified the defendant.” We have “observed repeatedly” that “[a] prosecutor’s stress upon the centrality of particular evidence in closing argument tells a good deal about whether the admission of the evidence was meant to be, and was, prejudicial.” Smith v. United States, 175 A.3d 623, 631 (D.C. 2017) (quoting Gathers v. United States, 101 A.3d 1004, 1009 (D.C. 2014)).

While we consider it a close question given the overall strength of the government’s case—even sans Nicholas’s identifications of Patrick—we have sufficient doubts about the impact of this constitutional error on Patrick’s jury that we reverse his carjacking and PFCV convictions.

38

III. Conclusion

For those reasons, we reverse Patrick’s convictions for armed carjacking and PFCV and remand to the trial court for vacatur of those convictions and further proceedings with respect to those charges. We affirm the remainder of his convictions.

So ordered.