Eyewitness identification and lineups

Two days after the Northgate robbery, an anonymous caller to Crime Stoppers said that the robber was Ryan Pelletier, a 27-year-old who lived near the north end, had served a sentence for a robbery in 2019, and drove a grey hatchback. Detective Whitlock, Ferreira's partner, pulled Pelletier's booking photograph and suggested showing it to the clerk at the Pine Street gas bar, where a similar robbery had happened on January 24. The clerk there, Arjun Mehta, had seen the robber's face for several seconds before the man pulled up his face covering. 'Just show him the picture and ask if that's the guy,' Whitlock said. Ferreira said no. Desjardins tells the class that her refusal may have saved an innocent man from being charged with robbery.

Mistaken eyewitness identification is the most common factor in wrongful convictions later proven by DNA evidence. Canadian inquiries into the wrongful convictions of Thomas Sophonow, convicted three times of a Winnipeg murder he did not commit, and others found that suggestive identification procedures had played a central role, and the 2001 Sophonow inquiry led by former Justice Peter Cory made detailed recommendations for how lineups should be run. Honest, confident witnesses can be wrong, and a poorly run identification procedure can create a memory that feels completely real to the witness. This lesson covers how to collect identification evidence in a way that protects both the witness's memory and the suspect's right to a fair investigation.

Why identification is fragile

A face seen for a few seconds under stress is encoded imperfectly, and the memory fades quickly. When a witness looks at a lineup, they compare each face with that memory and often choose the person who looks most like their memory relative to the others, even if the offender is not there. This is called relative judgement. If the suspect is the only person who fits the description, or stands out in some way, the witness may choose him because he is the best available match. Once a witness has chosen someone, the memory of that face can replace the original memory, so that a later identification in court is really an identification of the person from the lineup. A witness who receives confirming feedback, such as 'Good, that's our suspect,' becomes much more confident, and later remembers having had a better view than they did.

For these reasons, the identification procedure is the most important system variable police control. A single-photo showing, like the one Whitlock suggested, is highly suggestive: it tells the witness that the police think this person is the offender. Courts in both countries treat it as unreliable, and it can taint any later identification by the same witness. A procedure designed to test the witness's memory fairly gives the witness a real chance to say 'he isn't there'.

Best practices for photo lineups

The Sophonow inquiry and the federal, provincial and territorial heads of prosecutions committee reports on preventing wrongful convictions recommended several practices that most Canadian services now follow. The core practices are similar in both countries and are supported by research.

A blind administrator is an officer who does not know which photo shows the suspect. Where no such officer is available, a blinded procedure can be used, such as the folder shuffle method, where each photo is placed in a separate folder and shuffled so that the administrator cannot see which one the witness is viewing. Canadian practice generally favours sequential presentation, showing photos one at a time, as the Sophonow inquiry recommended, with at least ten photos where possible. Research comparing sequential and simultaneous lineups has produced mixed results; what matters most is that the procedure is double-blind, fair, properly instructed and recorded.

Choosing fillers

Fillers should be chosen so that the suspect does not stand out. Their photos should match the witness's description of the offender: age range, sex, build, hair, skin tone and any feature the witness mentioned. All photos should be similar in format, background and quality, and none should have features such as a booking number or a different style that draws attention. If the suspect has a distinctive feature that the witness did not mention, such as a scar, it can be covered or added to every photo in the same way. A lineup's fairness can be tested by showing it to mock witnesses, people who never saw the crime, along with the witness's description. If the description alone leads most mock witnesses to the suspect, the lineup is unfair.

# Mock witness test: people who did not see the crime are given the
# witness's description and asked to pick the person who fits it best.
lineups = {
    "Lineup 1 (Pelletier in position 4)": {"photos": 6, "mock": 30, "picked_suspect": 4},
    "Lineup 2 (Marsh in position 2)":     {"photos": 6, "mock": 30, "picked_suspect": 6},
}
for name, d in lineups.items():
    fair_rate = d["mock"] / d["photos"]          # expected picks if all fit equally
    functional_size = d["mock"] / d["picked_suspect"]
    print(name)
    print("   expected picks if fair:", fair_rate, " actual picks of suspect:", d["picked_suspect"])
    print("   functional size:", round(functional_size, 1), "of", d["photos"])

Lineup 1 (Pelletier in position 4)
   expected picks if fair: 5.0  actual picks of suspect: 4
   functional size: 7.5 of 6
Lineup 2 (Marsh in position 2)
   expected picks if fair: 5.0  actual picks of suspect: 6
   functional size: 5.0 of 6

Functional size is the number of mock witnesses divided by the number who chose the suspect. In a perfectly fair six-photo lineup, each photo would be chosen by about one mock witness in six, giving a functional size of six. Both lineups in the Northgate investigation were close to fair: in the first, only four of thirty mock witnesses chose Pelletier, fewer than the five expected by chance, and in the second, six of thirty chose Marsh, slightly more than expected. A functional size near two or lower would show that the suspect stood out and the lineup should be rebuilt. Port Ardent's identification unit runs a mock witness check on lineups for serious cases.

Running the procedure

On February 11, Constable Kim Tremblay, who had no involvement in the case and did not know who the suspect was, showed Mehta a sequential photo lineup of six photos in shuffled folders, with Pelletier's photo among five fillers chosen to match Mehta's description of a slim white man in his twenties with short brown hair. Before starting, she read the standard instructions: the person who committed the crime may or may not be in the photos; you do not have to choose anyone; the investigation will continue whether or not you choose someone; and if you do choose someone, tell me in your own words how sure you are. The procedure was video recorded.

Mehta looked at each photo. He paused at photo 4, which was Pelletier, and said, 'He's similar, but I don't think it's him. The guy's face was thinner.' He did not choose anyone. Tremblay recorded his words and thanked him without comment. When Desjardins asks the class what that result means, Amara Okafor says it does not prove Pelletier was innocent, because memory is imperfect, but it is evidence pointing away from him, and it must be disclosed. Desjardins agrees and adds that, two days later, records from Pelletier's employer showed he had swiped into a warehouse across town at 5:52 p.m. on February 7 and out at 2:03 a.m., and video from the warehouse showed him working throughout the evening. Pelletier was eliminated, and the anonymous tip was recorded as unfounded, with the steps that showed it.

After Marsh became a suspect through the DNA hit, a second lineup was prepared with his photo, taken from a 2022 identification photograph, among five new fillers. On February 23, a different blind administrator showed it to Mehta. Mehta chose photo 2, Marsh, and said, 'I think that's him. I'm fairly sure, not a hundred.' Those words, not a number or a paraphrase, went into the record. Desjardins points out that 'fairly sure' from a witness who saw a face for a few seconds a month earlier is meaningful but limited evidence. Its value came from the fact that it fitted the DNA, the car and the other evidence, not from the identification alone.

Show-ups and other procedures

A show-up is the presentation of a single suspect to a witness shortly after a crime, usually at or near the scene, for example when officers stop a person matching the description a few blocks away. Show-ups are inherently suggestive, and courts in both countries accept them only when they are done promptly, usually within an hour or two, when there is a good reason not to wait for a lineup, such as the need to release an innocent person quickly. Officers should give the same instructions as for a lineup, avoid showing the person in handcuffs or the back of a police car if possible, show the suspect to one witness at a time, and record the witness's exact words. If the witness has already seen the suspect in a show-up, a later lineup with the same suspect adds little.

Other identification evidence includes voice identification, which is even less reliable than visual identification; identification from video or photographs by people who know the suspect, such as a parole officer or relative, which courts treat carefully; and identification in court, often called dock identification, which on its own carries little weight because the accused is obviously the person charged. The Supreme Court of Canada in R. v. Hibbert (2002) warned that an in-court identification adds little to an earlier one and that juries must be cautioned about the dangers of eyewitness evidence.