Snap Shot
- The question: How well do facial composites (sketches and computer-generated faces built from a witness’s memory) resemble the person they depict, and can making or seeing a composite change a witness’s later identification?
- What the research shows: Composites from traditional feature-by-feature systems are correctly named by people who know the target only a minority of the time, especially after a delay of a day or more; holistic systems such as EvoFIT, and a holistic cognitive interview, have substantially raised naming rates in research. Several experiments found that building or viewing a composite lowered later correct identifications, but others, including a study with a live staged crime and police operators, did not.
- What it does not show: The research does not show that composites are useless as investigative tools, that a composite’s resemblance to a defendant proves identity, or that building a composite harmed any particular witness’s identification.
- The law: A composite is usually offered as a witness’s prior identification or description. Federal, military, and Arizona rules treat a testifying witness’s statement identifying a person perceived earlier as not hearsay (Rule 801(d)(1)(C)); courts differ on whether the sketch itself is hearsay (United States v. Moskowitz). The witness’s prior description is a Biggers reliability factor, and Mil. R. Evid. 321 governs identifications in courts-martial.
- Issues for each side: Prosecutors should treat a composite as a lead and document how and when it was made, without presenting its likeness as proof of identity; defense counsel can examine how the composite process may have affected a later identification, without asserting that composite construction necessarily corrupts memory.
When a crime is committed by a stranger and there is no suspect, police may ask the witness to help produce a picture of the offender’s face. The resulting facial composite, whether drawn by a sketch artist or assembled on a computer, is released to the public or circulated among officers in the hope that someone who knows the person will recognize the face and provide a name. Composites have solved cases, and they remain a standard investigative tool.
Composites reach court when one side points to the resemblance, or the differences, between a composite and the defendant, and when the composite process preceded the witness’s identification of the defendant from a lineup, raising the question whether building or viewing the composite affected that identification.
This paper describes the main composite systems and how they differ, research on how often composites are correctly recognized and how interviewing affects their quality, why producing a likeness from memory is difficult, the effects of composite construction on a witness’s later identification, and recent research on factors such as weapons. It then reviews the federal, military, and Arizona law governing composites as evidence.
Composite Systems
Composites are made in three main ways. A forensic artist sketches a likeness while the witness describes the offender, then revises the drawing as the witness comments on it. Feature-based systems, which descend from the mechanical Identikit and Photofit kits of the mid-twentieth century and now include software such as E-FIT, PRO-fit, and FACES, ask the witness to describe the face and then choose individual features (eyes, nose, mouth, hairline, face shape) from a library, which an operator assembles and adjusts. Holistic systems such as EvoFIT and EFIT-V work differently. The witness is shown arrays of complete faces and repeatedly selects those that most resemble the offender; the system combines the selected faces to generate new arrays, and the face gradually “evolves” toward the witness’s memory (Frowd et al., 2004).
The holistic approach was designed to address a mismatch. People generally recognize faces as integrated wholes, using the relationships among features, but traditional systems ask witnesses to recall and select features one at a time. Reviewers have proposed that this mismatch helps explain why feature-based composites often produce poor likenesses, even of faces the constructor knows well (Wells & Hasel, 2007).
How Often Are Composites Recognized?
The standard laboratory test of a composite is whether people who are familiar with the target, such as fans who know a celebrity or classmates who know a fellow student, can name the person from the composite. A meta-analysis of 23 studies that followed procedures used by police practitioners, comprising 6,464 naming responses from 1,069 participants, found that composites from the holistic EvoFIT system were named correctly far more often than composites from feature-based systems or sketches, with average estimated naming rates of about 56% for EvoFIT, 22% for sketches, and 15% for feature systems across the conditions studied (Frowd et al., 2015). Two of the authors of that meta-analysis, Frowd and Hancock, also developed EvoFIT (Frowd et al., 2004), and the rates come from conditions in which the people naming the composites were familiar with the target and the delays before construction varied. Delay mattered greatly: average naming fell from about 42% when the composite was built immediately to about 12% when it was built one to two days after the face was seen, and feature-based composites made after such delays were typically named only about 5% of the time.
The interview that precedes construction also matters. A holistic cognitive interview adds to the standard cognitive interview a phase in which the witness thinks about the overall character of the face and makes judgments about it (for example, how friendly or intelligent the person appeared). In one study using a feature-based system, composites built after a standard cognitive interview were named about 9% of the time and composites built after the holistic interview about 41% of the time (Frowd et al., 2008), and the meta-analysis found that the holistic interview roughly doubled the odds of correct naming across system types (Frowd et al., 2015). Combining composites made independently by several witnesses of the same person, by digitally morphing them together, also tends to improve likeness (Wells & Hasel, 2007).
These figures cut both ways. Composites made with older methods and after a delay are rarely named even by people who know the person well, but composites made with current methods can be recognizable, which is why they continue to generate leads. Naming rates do not measure how often a composite resembles an innocent person.
Why Recalling a Face Is Harder Than Recognizing One
Judging whether a face one is looking at is familiar is a much easier task than calling up a face in enough detail to rebuild it, and composite construction demands the second of these. The difficulty is greatest for unfamiliar faces. People recognize familiar faces even from poor images, but they rely heavily on external features such as hair and face shape when processing unfamiliar faces and match unfamiliar faces poorly even when comparing a photograph to a face in view (Lane & Houston, 2021, pp. 44–47, 52–53). A witness to a crime by a stranger is usually working from a brief view of an unfamiliar face, which is the situation in which composites are most often requested.
Where memory is incomplete, a constructor must still produce a complete face, and gaps may be filled from expectation. In an early study, the character attributed to a target (a lifeboat captain or a murderer) influenced the impressions that the resulting Photofit composites conveyed to other observers (Shepherd et al., 1978). The study is dated and small, but it illustrates that a composite can reflect the constructor’s beliefs about a person as well as the person’s appearance.
Composite construction also involves describing the face. Research on verbal overshadowing shows that describing a face can, under some conditions, make it harder to recognize later (see the companion paper on Verbal Overshadowing). In a 31-laboratory Registered Replication Report of the original verbal overshadowing study, participants who described a robber they had seen on video were compared with participants who did not describe him. The effect was larger when the description was given after a delay following the event and the identification test followed soon after the description. Participants who described the robber were about 4 percentage points less likely than controls to identify him when they described him immediately and completed a 20-minute filler task before the lineup, and about 16 percentage points less likely when the filler task came first and the lineup followed the description (Alogna et al., 2014). Composite construction may involve a longer or more analytic description than the few minutes of writing in that study, although the studies reviewed here did not measure this.
Effects of Composites on Later Identification
A witness who has built a composite is often later asked to view a lineup. Whether the composite process affects that identification has been studied directly, and the results are mixed.
In two experiments, participants who built a composite with a feature-based system were less likely than control participants to identify the original face from a later lineup; the effect appeared with photographs of faces and with a simulated crime, and composite building did not increase mistaken identifications from culprit-absent lineups (Wells et al., 2005). Two later studies tested lineups made of faces blended between the original face and the participant’s composite. Constructing a composite, and even viewing a composite made by someone else, reduced identification of the original face, with correct identifications in one study falling from about half of control participants to about a quarter of constructors and viewers (Kempen & Tredoux, 2012). In the other, participants who built a composite or viewed one over a week’s delay more often chose a blend of the target and the composite, suggesting that the composite had become incorporated into memory for the face, in a manner similar to the misinformation effect (Topp-Manriquez et al., 2016).
Other studies have not found harm. Adults who built a holistic EFIT-V composite identified the target from a later video lineup as accurately as adults who did not, although children who built a composite performed worse than child controls (Davis et al., 2016). In a study designed to resemble police practice, with a live staged crime, trained police operators, a short delay between the crime and construction, and a long delay before the lineup, composite construction did not significantly affect identification accuracy, and a laboratory replication with a sequential lineup found the same (Pike et al., 2019). The differences among these studies, which vary in the composite system, the delays, the lineup format, and whether the lineup included blends of the composite, make it difficult to state a single conclusion. A fair summary is that composite construction can impair later identification under some conditions, especially when the witness also views the composite repeatedly and when the lineup contains faces resembling the composite, and that the effect has not appeared under several more realistic procedures.
Weapons and Other Conditions at the Scene
Conditions that impair memory for a face also affect composites. In an experiment published in 2024, participants viewed an unfamiliar face alone or with a knife for 10 or 30 seconds and built an EvoFIT composite the next day. When viewing time was 10 seconds, composites of faces seen with the knife were correctly named less often than composites of faces seen without it (about 47% compared with 66%); at 30 seconds, the knife made no reliable difference (Erickson et al., 2024). The study involved few constructors (10 per condition) and a single object, but it suggests that weapon presence and brief exposure, which commonly occur together in real crimes, can reduce composite quality.
Limits of the Evidence
Composite research usually tests constructors who studied a photograph or video of a target under calm conditions and without consequences, and namers who already know the target. That is a reasonable test of investigative usefulness, but it does not measure how often a composite will be mistaken for someone else. Many studies have small numbers of constructors, and some of the most influential are more than fifteen years old. Studies of the effect of construction on later identification vary widely in design, and the studies closest to field conditions did not find harm. No study can say whether a particular composite resembles a particular person because that person is the offender, or whether building a composite affected a particular witness’s identification.
Questions to Ask About This Research
When an expert relies on naming rates, counsel should ask what those percentages measure and whose performance they describe: rates such as 15% for feature-based composites or 56% for EvoFIT (Frowd et al., 2015) are averages from laboratory studies in which namers already knew the target, and they depend heavily on the delay before construction and the interview used. Counsel should ask which system was used in this case, how long after the event the composite was built, and whether a holistic or standard interview was used.
When studies are offered to show that composite construction damaged a later lineup identification, counsel should ask whether the study designs support that inference: Wells et al. (2005) and Kempen and Tredoux (2012) found impairment, but Pike et al. (2019) did not under more realistic conditions, and some studies used lineups containing faces blended with the composite. When a composite’s resemblance to the defendant is offered as support for identity, counsel should ask how often composites resemble people other than the target, a question the naming research does not answer. These questions are explained in “Reading Research Claims: A Guide for Attorneys” on this site.
Face Composites and the Law
Federal and Military Standards
A composite is usually offered as a record of the witness’s earlier identification or description of the offender. Under Federal Rule of Evidence 801(d)(1)(C), a statement is not hearsay if the declarant testifies, is subject to cross-examination about the prior statement, and the statement identifies a person as someone the declarant perceived earlier. Courts have differed about how to classify the sketch itself. In United States v. Moskowitz (2d Cir. 1978), the majority held that a police sketch is not a “statement” and therefore not hearsay, that the eyewitnesses’ statements that the sketch resembled the robber were admissible under Rule 801(d)(1)(C) because both witnesses testified and were cross-examined, and that the sketch is relevant only if the identifying witness testifies; Judge Friendly, concurring, would have treated the sketch as part of the witnesses’ statements and admitted it under Rule 801(d)(1)(C). Some state courts treat composites as hearsay that is admissible as a prior identification (State v. Motta, Haw. 1983), and New York treats a composite as hearsay that is inadmissible to prove guilt, although it may be admitted for other purposes, such as rebutting a charge that the witness’s identification is a recent fabrication (People v. Maldonado, N.Y. 2002).
The witness’s prior description is one of the factors used to assess the reliability of an identification that follows an unnecessarily suggestive police procedure (Neil v. Biggers, 1972; Manson v. Brathwaite, 1977), and a composite may be part of the record of that description. Perry v. New Hampshire (2012) limited the due process screen to suggestive circumstances arranged by law enforcement. Building a composite does not by itself point the witness toward the defendant, so research on its effects is ordinarily presented through cross-examination, expert testimony, and argument about weight; whether a particular composite procedure contributed to a suggestive identification is a question for the court on the facts.
In courts-martial, Military Rule of Evidence 801(d)(1)(C) is identical in substance to the federal rule, and Military Rule of Evidence 321(a) provides that evidence of a witness’s previous identification is admissible to corroborate the witness’s in-court identification even if the witness’s credibility has not been directly attacked, subject to objection under the rule. Under Rule 321, an identification process is unreliable, and therefore unlawful, if it is so suggestive as to create a substantial likelihood of misidentification. Expert testimony about composites is governed by Federal Rule of Evidence 702, as amended December 1, 2023, Daubert v. Merrell Dow Pharmaceuticals, Inc. (1993), and Kumho Tire Co. v. Carmichael (1999), and in courts-martial by Military Rule of Evidence 702 and United States v. Houser (1993); under Military Rule of Evidence 1102, the 2023 federal amendment became part of the military rule in 2025 (Joint Service Committee on Military Justice, 2025), and practitioners should confirm current text.
Arizona
Arizona Rule of Evidence 801(d)(1)(C) uses the same language as the federal rule, treating a testifying witness’s prior statement identifying a person as someone the declarant perceived earlier as not hearsay. A published Arizona decision addressing whether a composite sketch itself is hearsay was not identified for this paper, and practitioners should research current Arizona authority. Challenges to pretrial identifications are heard at a Dessureault hearing (State v. Dessureault, 1969), where the accuracy of the witness’s prior description is among the reliability factors considered. Arizona Rule of Evidence 702 adopted the Daubert standard effective January 1, 2012, and was amended effective January 1, 2024, to track the federal amendment (Ariz. Sup. Ct. Order R-23-0004). State v. Chapple (1983), Arizona’s leading case on eyewitness expert testimony, held that excluding such testimony was an abuse of discretion on its facts; outside circumstances like those, admission remains within the trial court’s discretion (State v. Nordstrom, 2001).
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. Composites are a legitimate investigative tool, composites made with current holistic methods and interviews can be recognizable, and a composite that led to a suspect can be explained as part of the investigation. Several studies, including the one most closely resembling police practice, found that building a composite did not impair a later identification. Prosecutors are well positioned to document the system used, the interview, the time elapsed, whether the witness saw the composite again, and the sequence of composite and lineup.
The research also counsels caution. Composites are often imperfect likenesses, particularly when made with feature-based systems after a delay, and the naming research does not show how often a composite resembles someone other than the offender, so a resemblance between a composite and a defendant does not by itself establish identity. Some experiments found that constructing or repeatedly viewing a composite reduced later correct identifications. Presenting a composite as corroboration of an identification without acknowledging these limits goes beyond what the research supports.
Special Issues for Defense
The research supports several propositions defense counsel can legitimately present. Likenesses built from a brief view of a stranger’s face often resemble the target poorly, and delay, weapon presence, and short exposure reduce their quality. Some experiments found that building or viewing a composite reduced later identification of the original face and that a composite can blend into memory for the face. Expert testimony explaining these findings, tied to the documented conditions of the composite and the identification, can help the factfinder evaluate both.
The research also counsels caution about its limits. Composites can generate accurate leads, and current methods produce more recognizable likenesses than older ones. Studies of construction effects are mixed, and the study closest to field conditions found no effect, so the research does not establish that building a composite impaired a given witness’s memory. Differences between a composite and a defendant are expected even when the composite depicts the right person. An expert cannot say whether a particular composite depicts the defendant or whether a particular identification was affected.
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
Alogna, V. K., Attaya, M. K., Aucoin, P., Bahník, Š., Birch, S., Birt, A. R., Bornstein, B. H., Bouwmeester, S., Brandimonte, M. A., Brown, C., Buswell, K., Carlson, C., Carlson, M., Chu, S., Cislak, A., Colarusso, M., Colloff, M. F., Dellapaolera, K. S., Delvenne, J.-F., … Zwaan, R. A. (2014). Registered replication report: Schooler and Engstler-Schooler (1990). Perspectives on Psychological Science, 9(5), 556–578. https://doi.org/10.1177/1745691614545653
Davis, J. P., Thorniley, S., Gibson, S. J., & Solomon, C. J. (2016). Holistic facial composite construction and subsequent lineup identification accuracy: Comparing adults and children. The Journal of Psychology, 150(1), 102–118. https://doi.org/10.1080/00223980.2015.1009867
Erickson, W. B., Brown, C., Portch, E., Lampinen, J. M., Marsh, J. E., Fodarella, C., Petkovic, A., Coultas, C., Newby, A., Date, L., Hancock, P. J. B., & Frowd, C. D. (2024). The impact of weapons and unusual objects on the construction of facial composites. Psychology, Crime & Law, 30(3), 207–228. https://doi.org/10.1080/1068316X.2022.2079643
Frowd, C. D., Bruce, V., Smith, A. J., & Hancock, P. J. B. (2008). Improving the quality of facial composites using a holistic cognitive interview. Journal of Experimental Psychology: Applied, 14(3), 276–287. https://doi.org/10.1037/1076-898X.14.3.276
Frowd, C. D., Erickson, W. B., Lampinen, J. M., Skelton, F. C., McIntyre, A. H., & Hancock, P. J. B. (2015). A decade of evolving composites: Regression- and meta-analysis. Journal of Forensic Practice, 17(4), 319–334. https://doi.org/10.1108/JFP-08-2014-0025
Frowd, C. D., Hancock, P. J. B., & Carson, D. (2004). EvoFIT: A holistic, evolutionary facial imaging technique for creating composites. ACM Transactions on Applied Perception, 1(1), 19–39. https://doi.org/10.1145/1008722.1008725
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
Kempen, K., & Tredoux, C. G. (2012). “Seeing is believing”: The effect of viewing and constructing a composite on identification performance. South African Journal of Psychology, 42(3), 434–444. https://doi.org/10.1177/008124631204200315
Lane, S. M., & Houston, K. A. (2021). Understanding eyewitness memory: Theory and applications. New York University Press.
Pike, G. E., Brace, N. A., Turner, J., & Vredeveldt, A. (2019). The effect of facial composite construction on eyewitness identification accuracy in an ecologically valid paradigm. Criminal Justice and Behavior, 46(2), 319–336. https://doi.org/10.1177/0093854818811376
Shepherd, J. W., Ellis, H. D., McMurran, M., & Davies, G. M. (1978). Effect of character attribution on Photofit construction of a face. European Journal of Social Psychology, 8(2), 263–268. https://doi.org/10.1002/ejsp.2420080210
Topp-Manriquez, L. D., McQuiston, D., & Malpass, R. S. (2016). Facial composites and the misinformation effect: How composites distort memory. Legal and Criminological Psychology, 21(2), 372–389. https://doi.org/10.1111/lcrp.12054
Wells, G. L., Charman, S. D., & Olson, E. A. (2005). Building face composites can harm lineup identification performance. Journal of Experimental Psychology: Applied, 11(3), 147–156. https://doi.org/10.1037/1076-898X.11.3.147
Wells, G. L., & Hasel, L. E. (2007). Facial composite production by eyewitnesses. Current Directions in Psychological Science, 16(1), 6–10. https://doi.org/10.1111/j.1467-8721.2007.00465.x
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).
People v. Maldonado, 97 N.Y.2d 522 (2002).
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. Motta, 66 Haw. 254, 659 P.2d 745 (1983).
State v. Nordstrom, 200 Ariz. 229, 25 P.3d 717 (2001).
United States v. Houser, 36 M.J. 392 (C.M.A. 1993).
United States v. Moskowitz, 581 F.2d 14 (2d Cir. 1978).
Fed. R. Evid. 702.
Fed. R. Evid. 801(d)(1)(C).
Ariz. R. Evid. 702.
Ariz. R. Evid. 801(d)(1)(C).
Mil. R. Evid. 321.
Mil. R. Evid. 702.
Mil. R. Evid. 801(d)(1)(C).
Mil. R. Evid. 1102.
Related topic papers: Verbal Overshadowing · Face Recognition · Memory for Persons and Faces · Lineup Construction · The Misinformation Effect · Memory for Violence and Weapon Focus
Related topics from Arizona Forensic Services: Courts-Martial