Children’s Face Recognition

Snap Shot

• The question: How well do children recognize a stranger’s face, and how much weight can a child’s identification from a lineup or photo array bear?
• What the research shows: When the culprit is in the lineup, school-age children identify him at rates close to adults’; an early meta-analysis found no significant difference after about age five, while a larger 2015 meta-analysis found a modest deficit, and preschoolers lag. The most consistent age difference lies in culprit-absent lineups, where children, and even young adolescents, choose someone more often than adults. Procedures that make “not there” an easy answer, such as the elimination lineup and the “wildcard,” have reduced that choosing in some studies.
• What it does not show: Group research cannot show whether a particular child’s identification is right or wrong, and a short description does not by itself predict accuracy or error. It does not show that children are unreliable identifiers as a class, and no procedure has been shown consistently across studies to remove the culprit-absent gap.
• The law: Children are presumed competent (Fed. R. Evid. 601; 18 U.S.C. § 3509). Identifications are screened under due process and, in courts-martial, Mil. R. Evid. 321; in Arizona through a Dessureault hearing and RAJI Standard Criminal 45. Experts may explain the research (Chapple) but may not opine on a particular child’s credibility (Lindsey).
• Issues for each side: Prosecutors can show that children often identify a present culprit accurately and that thin descriptions are typical, but should document unbiased instructions and the procedure. Defense counsel can show children’s elevated culprit-absent choosing, but should recognize that this is a probabilistic risk, not proof that a particular child chose wrongly.

Children witness crimes and are victims of them. When the offender was a stranger, a child may be asked to identify the person from a photo array or lineup, and the identification may become central evidence. A child’s mistaken identification is not a lie: a child who chooses an innocent person is usually trying to help and is often sincere. The research addresses not whether children tell the truth but how their face memory and lineup decisions differ from adults’, and under what conditions those differences matter.

This paper reviews the development of face recognition, children’s lineup performance, why children choose when the culprit is absent, child-friendly procedures, and confidence, including research through 2025, and then the federal, military, and Arizona law.

How Face Recognition Develops

Describing a face is a language-dependent recall task; identifying one is a largely visual recognition task. The two develop on different timetables and should be evaluated separately.

An influential early study proposed that young children recognize faces piecemeal, relying on features such as hair or glasses, and shift around age ten to processing the face as a configuration (Carey & Diamond, 1977). That account is now dated. A later critical review found holistic processing and other hallmarks of adult face recognition at the earliest ages tested, typically 3 to 5 years, and argued that the face-specific mechanisms mature early, with some apparent later gains reflecting the limits of earlier study methods (McKone et al., 2012). Measured face learning nonetheless keeps improving well beyond childhood: in more than 60,000 online participants, it improved until just after age 30 (Germine et al., 2011). How the two lines of work fit together remains debated. For legal purposes, children’s face memory improves gradually with age, as a matter of degree rather than a threshold.

Faces of people near one’s own age are generally recognized better than faces of people much older or younger. This own-age bias has been found in children and in younger and older adults (Anastasi & Rhodes, 2005), and a meta-analysis found it reliable across age groups, attributed largely to everyday experience with faces of different ages (Rhodes & Anastasi, 2012). A child identifying an adult stranger may therefore be at some disadvantage, although group data cannot estimate the size of the effect in a particular case.

Children’s Lineup Performance

Meta-analyses that pool many studies give the most reliable picture, and they show an asymmetry that depends on whether the culprit is in the lineup. A meta-analysis comparing children with adults found that, in culprit-present lineups, preschool children (around age four) made fewer correct identifications than adults, but children older than about five did not differ significantly from adults. In culprit-absent lineups, by contrast, children of every age examined were less likely than adults to make a correct rejection, that is, to say that the culprit was not there, and even adolescents of about 12 to 13 did not reach the adult rate (Pozzulo & Lindsay, 1998).

A later meta-analysis spanning the life span, with more than 20,000 participants in 91 studies, replicated children’s greater tendency to choose someone from culprit-absent lineups. Unlike the earlier review, it also found children less likely than young adults to identify the culprit when he was present, and signal-detection measures showed young adults better able than children to tell guilty from innocent suspects. The authors concluded that identifications by young adults were the most reliable but that identifications by witnesses of every age had probative value (Fitzgerald & Price, 2015). School-age children are often close to adults at picking out a culprit who is actually in the lineup, though the larger meta-analysis found a modest deficit; their most consistent vulnerability is a greater tendency to choose someone when the culprit is absent. Because investigators cannot know in advance whether the suspect is the culprit, that tendency is the central reliability concern.

Not every study shows parity, and procedure matters. With video lineups, correct identifications by 5- to 6-year-olds, 9- to 10-year-olds, and adults did not differ in simultaneous or elimination lineups, but a sequential lineup reduced children’s correct identifications (Humphries et al., 2012). With sequential lineups presented in augmented reality, children averaging about 11 years were less accurate than adults (Price & Fitzgerald, 2024).

Choosing When the Culprit Is Absent

Because the main age difference appears when the culprit is absent, three questions follow: what kind of error such a choice is, why children make it, and what a child’s description reveals.

Two Kinds of False Positive

A choice from a lineup that does not contain the culprit can take two forms. The witness may pick the innocent suspect, a false identification that can lead to prosecution of an innocent person. Or the witness may pick a filler, whom the police know to be innocent; that choice leads to no prosecution but is a known error that bears on the weight of the witness’s other identifications. In a fair lineup, most culprit-absent choices will, by design, fall on fillers; when the suspect stands out, more will fall on the suspect.

Why Children Choose

Researchers attribute children’s elevated choosing to social and decision factors more than to poorer memory for the face itself. Children may infer that a police officer would not bother showing photographs if the offender were not among them, may feel that choosing is the cooperative response, and may find it harder to say “he’s not there” to an adult than to point to a photograph. Witnesses of all ages can make relative judgments, picking whichever lineup member best resembles their memory of the culprit relative to the rest, and that strategy produces a “best match” even when the culprit is absent (Wells, 1984). These explanations rest mainly on the pattern of results and on the effects of procedures designed to counter them; most studies do not measure them directly.

Description Versus Identification

Compared with adults, children usually describe a stranger more briefly and with fewer details, with accuracy problems for body descriptors such as height and weight, for interior facial features, and for age (Pozzulo & Warren, 2003; Lane & Houston, 2021, pp. 35–37). The research does not support treating a thin description as a strong sign that a child did not see the person well. In a study of youths aged 10 to 14 and adults, witnesses in both groups reported a similar number of descriptors whether their identification decisions were accurate or not (Pozzulo & Warren, 2003), and a meta-analysis of adult and child studies found only a weak relationship between description and identification accuracy (Meissner et al., 2008). A sparse description is typical of child witnesses; standing alone, it neither undermines nor supports a later identification.

Procedures and Instructions

Most proposed safeguards are procedural. The foundation is a set of unbiased instructions, warning the child that the person might not be in the photographs and that “not here” or “I don’t know” is an acceptable answer, combined with a fairly constructed lineup in which the suspect does not stand out by clothing, background, or appearance.

Joanna Pozzulo and colleagues developed the elimination lineup for children. The child first eliminates all but the face that looks most like the culprit and then decides whether that face is the culprit. Its developers reported that the procedure reduced children’s choosing from culprit-absent lineups without significantly reducing correct identifications (Pozzulo & Lindsay, 1999). Later results have been mixed: with video lineups, the elimination procedure increased correct rejections for adults, while children’s correct rejections remained below adults’ (Humphries et al., 2012). In the meta-analytic data, sequential presentation, which can reduce choosing in adults, widened the child–adult gap in correct rejections compared with simultaneous presentation, and giving children identification practice or training did not raise their correct-rejection rates (Pozzulo & Lindsay, 1998).

Other procedures make rejecting the lineup as concrete as choosing. In the “wildcard” procedure, the lineup includes a silhouette with a large question mark, and the child points to it if the person is not there; among 159 children aged 8 to 11, it increased accuracy on culprit-absent lineups without affecting culprit-present performance (Zajac & Karageorge, 2009). Among 432 children aged about 12, a “not sure” option alone did not change false or correct identifications, though combined with instructions emphasizing accuracy it reduced false identifications under some conditions; the authors concluded that no manipulation had yet been shown to reliably reduce children’s false identifications without lowering correct ones (Brewer et al., 2010). Procedures can reduce, but have not eliminated, children’s choosing from culprit-absent lineups.

Confidence, Older Children, and Adolescents

For adults, high confidence at an initial, properly conducted identification is generally associated with high accuracy. For children the picture is less settled. In two experiments with more than 1,400 children averaging nearly 12 years, confidence was meaningfully related to accuracy among adults who chose from the lineup but not among children who chose, and the authors concluded that children’s confidence gave investigators no useful guide to a suspect’s likely guilt (Keast et al., 2007). A later reanalysis of basic and applied studies reached a more favorable conclusion, finding that confidence carried considerable information about accuracy from at least age 8. In the same paper’s experiment, with 2,205 children aged 4 to 17, a strong confidence–accuracy relationship appeared from about age 10 and an emerging one from about age 7, and high-confidence suspect identifications were usually accurate, with accuracy rising with age (Winsor et al., 2021). In a virtual-reality study of 1,055 children and adults, memory was weak for all groups and children’s high-confidence identifications were not highly accurate under either short or long exposure; adults made too few high-confidence identifications to analyze (Moore et al., 2025). In the augmented-reality study, children’s confidence did not distinguish correct from incorrect identifications, while faster decisions were more often accurate at all ages (Price & Fitzgerald, 2024). A child’s initial confidence may carry some information, particularly in older children, but less reliably than an adult’s.

Older children and adolescents are not simply small adults in this research. In the earlier meta-analysis, even adolescents of about 12 to 13 did not reach the adult rate of correct rejections (Pozzulo & Lindsay, 1998); the later meta-analysis replicated children’s elevated culprit-absent choosing (Fitzgerald & Price, 2015); and 12-year-olds still showed a tendency to choose (Brewer et al., 2010). At the same time, correct identifications in the 2021 study rose from about a third of culprit-present lineups at ages 4 to 6 to just over half at ages 10 to 17 (Winsor et al., 2021). Adolescent identifications warrant the same attention to culprit-absent risk and procedure, without assuming either adult-level or young-child performance.

Limits of the Evidence

Nearly all of this research uses staged events or videos, with low stakes, short delays, and no real victimization. A child victim may have a longer or more personally significant view of the offender, or a briefer and more stressful one, and the research does not settle how those differences change performance. Samples are often school-age children tested at schools or museums, with limited diversity; preschoolers are studied less often; and laboratory lineups are typically fair and well instructed, while real procedures vary. Meta-analytic averages conceal large differences across studies, and the benefits of the elimination lineup and the wildcard have not replicated uniformly. Recent studies with larger samples and immersive methods have refined the confidence findings without overturning the basic pattern, and they disagree with one another (Winsor et al., 2021; Moore et al., 2025). Most importantly, group research describes rates and tendencies; it cannot establish whether a particular child’s identification is correct.

Questions to Ask About This Research

When Pozzulo and Lindsay (1998) is offered for the point that children older than about five match adults in picking out a present culprit, counsel should ask what exactly was counted: that result comes from culprit-present lineups, the same meta-analysis found lower correct rejections in culprit-absent lineups, the larger later meta-analysis found a modest culprit-present deficit as well (Fitzgerald & Price, 2015), and investigators cannot know in advance which situation applied. When the same research is used to argue that children’s identifications carry little value, counsel should ask whether the later meta-analysis in fact concluded that identifications at every age had probative value. When the meta-analyses are used to argue that a child’s choice was probably mistaken, counsel should ask about this child’s age, the instructions, the lineup’s fairness, and whether a filler or the suspect was chosen. When a child-friendly procedure is said to make an identification reliable, counsel should ask how well its results have replicated (Humphries et al., 2012; Brewer et al., 2010). When a child’s high confidence is offered as evidence of accuracy, counsel should ask whether the conclusion follows for a child of this age under these viewing conditions (Winsor et al., 2021; Moore et al., 2025). These questions are explained in “Reading Research Claims: A Guide for Attorneys” on this site.

Children’s Face Recognition and the Law

Federal and Military Standards

Under Federal Rule of Evidence 601 and 18 U.S.C. § 3509, a child is presumed competent and a competency examination requires compelling reasons. In courts-martial, Military Rule of Evidence 611(d) and R.C.M. 914A permit remote testimony by a child under 16 only on findings that it is necessary to protect that child from more than de minimis trauma caused by the accused’s presence. The research therefore bears on the weight of a child’s identification, not on whether the child may testify.

Due process analysis does not change with the witness’s age. A court screens an identification for reliability when law enforcement arranged unnecessarily suggestive circumstances, weighing opportunity to view, attention, the accuracy of any prior description, certainty, and elapsed time (Neil v. Biggers, 1972; Manson v. Brathwaite, 1977); otherwise reliability is tested through cross-examination, instructions, and expert testimony (Perry v. New Hampshire, 2012). In courts-martial, Military Rule of Evidence 321 makes an identification inadmissible on timely motion if it resulted from an unlawful process, including one so suggestive as to create a substantial likelihood of misidentification; when the defense raises an unreliable identification, the prosecution must prove its reliability by a preponderance of the evidence, and a later identification is admissible only on clear and convincing evidence that it is not the result of the unreliable one. Expert testimony is governed by Federal Rule of Evidence 702, as amended December 1, 2023, and in courts-martial by Military Rule of Evidence 702 and United States v. Houser (1993); the 2023 amendment took effect in the military rule in 2025 through Military Rule of Evidence 1102 (Joint Service Committee on Military Justice, 2025), and practitioners should confirm current text. Daubert v. Merrell Dow Pharmaceuticals, Inc. (1993) makes the trial judge a gatekeeper, and Kumho Tire Co. v. Carmichael (1999) extends that obligation to all expert testimony, including non-scientific technical or specialized knowledge.

Arizona

A defendant who challenges a pretrial identification may request a hearing under State v. Dessureault (1969), at which the State must show by clear and convincing evidence that the procedure was not unduly suggestive or, if it was, that the in-court identification is not tainted by it; the hearing is not constitutionally required when the State did not arrange the suggestive circumstances (State v. Nottingham, App. 2012). RAJI Standard Criminal 45 tells jurors that the State must prove beyond a reasonable doubt that the in-court identification is reliable and lists factors drawn from Biggers, plus “any other factor” affecting reliability, which may include the witness’s age and the procedure used. In State v. Chapple (1983), Arizona’s leading case on eyewitness expert testimony, excluding such testimony was an abuse of discretion on its facts; admission otherwise rests in the trial court’s discretion (State v. Nordstrom, 2001). Arizona Rule of Evidence 702 adopted the Daubert standard in 2012 and was amended effective January 1, 2024, to track the 2023 federal amendment. Under State v. Lindsey (1986), an expert may explain general research on children’s identifications but may not opine on whether a particular child is accurate or credible.

Special Issues in Legal Practice

The principles described above apply with equal force to every party in a legal proceeding. The science operates the same way regardless of who is asking the questions, and the research summarized here does not by itself favor either side. What differs between the parties is not the science but the vantage point — the particular places where these issues tend to surface, and the particular errors each party is best positioned to guard against. The two summaries below identify those recurring issues. They are offered for orientation, not as litigation strategy; in every case the controlling question is what the evidence in that case will, and will not, support, and the expert’s role is to inform the trier of fact rather than to advocate for either side.

Special Issues for Prosecutors

The research supports several propositions prosecutors can legitimately present. School-age children often identify a culprit who is present at rates close to adults’, and the largest meta-analysis concluded that identifications by witnesses of every age had probative value; the evidence does not support treating children as unreliable identifiers as a class. A brief or vague description is typical of children and does not by itself show that the child failed to see the offender. In one large study, high-confidence initial identifications by children were usually accurate, and confidence was increasingly informative from about age 7. A child’s identification is strongest when the record shows unbiased instructions, a fair lineup, an easy way to say “not here,” a recorded initial response, and no confirming feedback.

The research also counsels caution. Children and many adolescents are more likely than adults to choose from a lineup that does not contain the culprit, and investigators cannot know in advance which kind of lineup they have constructed. A child’s confidence, particularly a young child’s, is a weaker indicator than an adult’s, and confidence expressed later, after feedback or repetition, is generally less informative than confidence at the first procedure. Procedures such as the elimination lineup and the wildcard can reduce choosing but have not been shown to eliminate the risk, and a child’s choice of a filler on another occasion bears on the weight of that child’s other identifications.

Special Issues for Defense

The research supports several propositions defense counsel can legitimately present. The most consistent age difference lies in culprit-absent lineups, where children choose more often than adults, and the research suggests that tendency is worsened by instructions implying the culprit is present, by lineups in which the suspect stands out, and, for children, by sequential presentation; confirming feedback after a choice can also inflate later confidence. Children’s confidence is less reliably related to accuracy than adults’, particularly for younger children and under poor viewing conditions, and the own-age bias may disadvantage a child recognizing an adult stranger. Expert testimony on these principles, with evidence of how the procedure was conducted, can help the factfinder weigh a child’s identification.

The research also counsels caution about the limits that apply to the defense. Children older than about five often perform close to adults when the culprit is present, and identifications at every age carry probative value, so the research does not show that a particular child’s choice was wrong, and a thin description is not, by itself, evidence of unreliability. An expert may not tell the factfinder that a particular child’s identification is mistaken, and a sincere child who chose wrongly is not untruthful. The argument depends on the conditions of the identification, not on age alone.

This paper is for educational purposes only and is not legal advice. The author is a psychologist, not an attorney; summaries of law are general and must be verified against current authority or with a licensed attorney.

References

Anastasi, J. S., & Rhodes, M. G. (2005). An own-age bias in face recognition for children and older adults. Psychonomic Bulletin & Review, 12(6), 1043–1047. https://doi.org/10.3758/BF03206441

Brewer, N., Keast, A., & Sauer, J. D. (2010). Children’s eyewitness identification performance: Effects of a Not Sure response option and accuracy motivation. Legal and Criminological Psychology, 15(2), 261–277. https://doi.org/10.1348/135532509X474822

Carey, S., & Diamond, R. (1977). From piecemeal to configurational representation of faces. Science, 195(4275), 312–314. https://doi.org/10.1126/science.831281

Fitzgerald, R. J., & Price, H. L. (2015). Eyewitness identification across the life span: A meta-analysis of age differences. Psychological Bulletin, 141(6), 1228–1265. https://doi.org/10.1037/bul0000013

Germine, L. T., Duchaine, B., & Nakayama, K. (2011). Where cognitive development and aging meet: Face learning ability peaks after age 30. Cognition, 118(2), 201–210. https://doi.org/10.1016/j.cognition.2010.11.002

Humphries, J. E., Holliday, R. E., & Flowe, H. D. (2012). Faces in motion: Age-related changes in eyewitness identification performance in simultaneous, sequential, and elimination video lineups. Applied Cognitive Psychology, 26(1), 149–158. https://doi.org/10.1002/acp.1808

Joint Service Committee on Military Justice. (2025). Amendments to the Military Rules of Evidence since 1984 (G. E. Maggs, Ed.; 2025 ed.). https://jsc.defense.gov/Portals/99/MRE%20Amendments%202025.pdf

Keast, A., Brewer, N., & Wells, G. L. (2007). Children’s metacognitive judgments in an eyewitness identification task. Journal of Experimental Child Psychology, 97(4), 286–314. https://doi.org/10.1016/j.jecp.2007.01.007

Lane, S. M., & Houston, K. A. (2021). Understanding eyewitness memory: Theory and applications. New York University Press.

McKone, E., Crookes, K., Jeffery, L., & Dilks, D. D. (2012). A critical review of the development of face recognition: Experience is less important than previously believed. Cognitive Neuropsychology, 29(1–2), 174–212. https://doi.org/10.1080/02643294.2012.660138

Meissner, C. A., Sporer, S. L., & Susa, K. J. (2008). A theoretical review and meta-analysis of the description-identification relationship in memory for faces. European Journal of Cognitive Psychology, 20(3), 414–455. https://doi.org/10.1080/09541440701728581

Moore, K. N., Zwemer, D. U., Lampinen, J. M., Pennekamp, P., Nyman, T. J., Santtila, P., Korkman, J., Antfolk, J., & Yu, C. (2025). Using live-action 360-degree video to assess the impact of exposure duration on eyewitness identification accuracy at high confidence in children and adults. Cognitive Research: Principles and Implications, 10(1), Article 85. https://doi.org/10.1186/s41235-025-00692-9

Pozzulo, J. D., & Lindsay, R. C. L. (1998). Identification accuracy of children versus adults: A meta-analysis. Law and Human Behavior, 22(5), 549–570. https://doi.org/10.1023/A:1025739514042

Pozzulo, J. D., & Lindsay, R. C. L. (1999). Elimination lineups: An improved identification procedure for child eyewitnesses. Journal of Applied Psychology, 84(2), 167–176. https://doi.org/10.1037/0021-9010.84.2.167

Pozzulo, J. D., & Warren, K. L. (2003). Descriptions and identifications of strangers by youth and adult eyewitnesses. Journal of Applied Psychology, 88(2), 315–323. https://doi.org/10.1037/0021-9010.88.2.315

Price, H. L., & Fitzgerald, R. J. (2024). Reflector variables in augmented reality lineups: Assessing eyewitness identification reliability in children and adults with confidence, response time, and proximity to the lineup. PLOS ONE, 19(9), Article e0308757. https://doi.org/10.1371/journal.pone.0308757

Rhodes, M. G., & Anastasi, J. S. (2012). The own-age bias in face recognition: A meta-analytic and theoretical review. Psychological Bulletin, 138(1), 146–174. https://doi.org/10.1037/a0025750

Wells, G. L. (1984). The psychology of lineup identifications. Journal of Applied Social Psychology, 14(2), 89–103. https://doi.org/10.1111/j.1559-1816.1984.tb02223.x

Winsor, A. A., Flowe, H. D., Seale-Carlisle, T. M., Killeen, I. M., Hett, D., Jores, T., Ingham, M., Lee, B. P., Stevens, L. M., & Colloff, M. F. (2021). Child witness expressions of certainty are informative. Journal of Experimental Psychology: General, 150(11), 2387–2407. https://doi.org/10.1037/xge0001049

Zajac, R., & Karageorge, A. (2009). The wildcard: A simple technique for improving children’s target-absent lineup performance. Applied Cognitive Psychology, 23(3), 358–368. https://doi.org/10.1002/acp.1511

Legal Authorities

Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993).

Kumho Tire Co. v. Carmichael, 526 U.S. 137 (1999).

Manson v. Brathwaite, 432 U.S. 98 (1977).

Neil v. Biggers, 409 U.S. 188 (1972).

Perry v. New Hampshire, 565 U.S. 228 (2012).

State v. Chapple, 135 Ariz. 281, 660 P.2d 1208 (1983).

State v. Dessureault, 104 Ariz. 380, 453 P.2d 951 (1969).

State v. Lindsey, 149 Ariz. 472, 720 P.2d 73 (1986).

State v. Nordstrom, 200 Ariz. 229, 25 P.3d 717 (2001).

State v. Nottingham, 231 Ariz. 21, 289 P.3d 949 (App. 2012), abrogated in part (cautionary-instruction holding) by State v. Bigger, 492 P.3d 1020 (Ariz. 2021).

United States v. Houser, 36 M.J. 392 (C.M.A. 1993).

18 U.S.C. § 3509.

Fed. R. Evid. 601.

Fed. R. Evid. 702.

Ariz. R. Evid. 702.

Mil. R. Evid. 321.

Mil. R. Evid. 611(d).

Mil. R. Evid. 702.

Mil. R. Evid. 1102.

Rule for Courts-Martial 914A.

Revised Arizona Jury Instructions (Criminal), Standard Criminal 45 (Identification) (2022).

Related topic papers: Face Recognition · Lineup Construction · Confidence and Accuracy · Suggestibility of Children’s Memory · Memory Development · Eyewitness Memory

Related topics from Arizona Forensic Services: Child Abuse Allegations · Courts-Martial