Snap Shot
- The question: How complete and how accurate are witnesses’ spoken or written accounts of what a person looked like, including estimates of height, weight, age, and build, and what does a description say about a later identification?
- What the research shows: Descriptions of strangers are usually short and general, built mostly from sex, age, height, build, hair, and clothing, with few inner facial details. What witnesses do report is often broadly correct, but numerical estimates of age, height, and weight carry wide error, especially at a distance, and the quality of a description is only weakly related to whether a later identification is accurate.
- What it does not show: A brief or partly mistaken description does not show that a later identification is wrong, and a detailed, matching description does not show that it is right. Group research cannot say whether a particular witness’s description is accurate.
- The law: The accuracy of the witness’s prior description is one of the Neil v. Biggers reliability factors, applied in courts-martial through Mil. R. Evid. 321 and in Arizona at a Dessureault hearing and in RAJI (Criminal) Standard Criminal 45. Expert testimony on these limits is governed by Rule 702, Houser, and Chapple.
- Issues for each side: Prosecutors can show that a description that was recorded early and that matches the defendant on attributes the witness actually reported is consistent with an accurate identification, without treating the match as proof of it. Defense counsel can show how sparse, estimated, or changeable a description was, while recognizing that a thin description is common among accurate witnesses too.
A person description is a witness’s account, from memory, of what someone looked like: sex, age, height, weight, build, race, hair, clothing, and distinguishing marks. It usually precedes any identification and guides whom officers look for, which photographs they search, and which fillers go into a lineup. Later, the parties compare it with the defendant’s appearance, and courts treat its accuracy as one marker of an identification’s reliability.
Researchers study descriptions both in archives of what real witnesses told police and in experiments where the target’s true appearance is known. This paper covers how much witnesses report, which attributes they report, how accurate their estimates are, what affects completeness and accuracy, how descriptions are recorded, and how a description relates to a later identification. The companion paper, Memory for Persons and Faces, addresses the face itself: encoding, viewing distance, exposure time, familiarity, and appearance change. Lineup recognition is covered in Face Recognition, composites in Face Composites, and other-race faces in The Cross-Race Effect.
How Much Witnesses Report
Archival studies consistently find that descriptions of strangers are short. In a study of police reports of violent crimes (Kuehn, 1974, as cited in Lane & Houston, 2021, p. 28), victims gave about seven descriptors of the offender on average, most often sex, age, height, and build, and less often race, weight, complexion, and hair color. Kuehn also reported that how complete a description was depended on characteristics of the crime and of the victim.
Later archival work points the same way. In a Dutch study of 2,299 descriptions given by 1,313 witnesses to robberies, the median description contained eight descriptors out of a possible 43, most of them general permanent features such as sex, build, and skin color, or temporary ones such as headwear and jackets (Van Koppen & Lochun, 1997). Laboratory witnesses tend to say more. Lindsay et al. (1994) compared descriptions from staged crimes with descriptions of real offenders published in newspapers and found averages of 7.35 and 3.94 features, respectively. The newspaper descriptions had passed through police and reporters, so the comparison partly measures what reached the public; even so, laboratory conditions may overstate how much a frightened witness to a brief real event can report.
The general conclusion drawn in reviews is that descriptions are usually vague, tend toward population averages (“medium build,” “average height”), and would match a large share of the people living near the scene (Meissner et al., 2007; Lane & Houston, 2021, p. 28).
Which Attributes Witnesses Report
The content of descriptions follows a pattern. Witnesses describe the body, the hair, and the clothing far more often than the inner face. In Van Koppen and Lochun’s (1997) sample, face shape appeared in about 12% of descriptions, but eye color in 5%, the nose in 3%, the mouth in 2%, and the chin in almost none. Lindsay et al. (1994) likewise found that facial features were rarely described in real cases. Clothing and hair are easy to put into words but can be changed within minutes, while the stable facial features that would narrow the field most are the hardest to describe.
The rarity of facial detail does not mean the witness did not see the face; it reflects in part the difficulty of putting a face into words, examined in Memory for Persons and Faces, and in part what police forms and questions ask for. Face Composites addresses the separate question of how well witnesses can reconstruct a face visually rather than verbally.
Accuracy of Descriptions and Estimates
Where accuracy can be checked, the findings are more favorable to witnesses than the brevity of descriptions might suggest. Of the descriptors that Van Koppen and Lochun (1997) could verify against police records, 59% were correct, 17% partly correct, and 24% wrong, and the median description was about 72% accurate. In a Norwegian study that compared 250 descriptions of armed bank robbers with the robberies’ video recordings, witnesses were generally accurate but gave little that would single out one person, and the witness’s role (teller or customer), the weapon, and the number of robbers modestly predicted description quality (Fahsing et al., 2004). Witnesses, in other words, tend to report a small amount of mostly correct general information.
The weak point is inner facial detail and numerical estimates. Among the few inner facial features Van Koppen and Lochun (1997) could verify, most descriptions of the nose, mouth, and chin were wrong, and eye color was fully correct in only about a third of cases. Height, weight, and age are judgments rather than observations, and research identifies two sources of systematic error. People anchor on themselves: estimates of a stranger’s height are pulled toward the observer’s own height, and error grows as the difference between observer and target grows (Twedt et al., 2015). People also anchor on categories, pulling estimates toward the average for the group to which the person seems to belong (Twedt et al., 2015; Meissner et al., 2007).
Distance and the witness’s age also matter. In a field study, 1,588 participants from young children to older adults viewed live targets at 5 to 110 meters and estimated each target’s sex, age, height, and weight. Sex was judged correctly most of the time at every distance, but even at 5 meters only about a quarter to a half of age estimates, and fewer height and weight estimates, fell within two units of the true value; serious errors increased with distance, and young children were the least accurate (Nyman et al., 2020). Archival data from real cases show similar error. In an analysis of 546 stranger sexual assault cases in England and Wales, victims’ estimates of the convicted offender’s age missed by about 4.8 years on average, were about as likely to be too high as too low, and were more accurate when the victim and offender were closer in age, consistent with an own-age bias (Thorley et al., 2018). In a laboratory study, estimates of age were the most accurate descriptor and estimates of height commonly missed by more than 5 centimeters; using age, height, and ethnicity together to narrow a set of photographs would have excluded the actual culprit about half the time (Handler & Frühholz, 2021). An estimate such as “about 5 foot 10, in his thirties” is best understood as a range, and a modest mismatch with the defendant’s height or age is common even among otherwise accurate witnesses.
Confidence in individual details has been studied less than confidence in identifications. Laboratory research on witnesses’ recall of crime details suggests that confidence can track the accuracy of a reported detail when it is measured soon after the event, but that the relationship weakens after long retention intervals (Spearing & Wade, 2022).
What Affects Completeness and Accuracy
Some conditions of the encounter bear on how much a witness can describe. In Van Koppen and Lochun’s (1997) data, descriptions were more complete when they were given sooner, when the witness was closer to the robber and had a better view, and when the threat was lower; better lighting was the one factor linked to a perfectly accurate description. Overall, however, these factors predicted accuracy poorly, so favorable conditions do not guarantee an accurate description and unfavorable ones do not guarantee an inaccurate one. Reviews group the main influences into the opportunity to view, stress, and intoxication (Meissner et al., 2007). Lane and Houston (2021, pp. 30–35) add that longer exposure generally helps, that no study sets a threshold for how long is “enough,” that people misjudge durations, and that alcohol at the levels studied has impaired the completeness of recall more clearly than it has impaired identification. Children give fewer and less accurate descriptors than adults, with particular difficulty estimating height, weight, and age (Lane & Houston, 2021, pp. 35–37).
Questioning matters as well. Open prompts and rapport generally elicit more correct detail than brief question-and-answer formats (see Improving Eyewitness Memory), but there is a limit. When participants were pressed to keep describing a face after they had run out of what they were confident about, the extra detail included more errors and their later identifications were less accurate than those of participants told to report only what they were sure of (Meissner et al., 2001). Lane and Houston (2021, pp. 42–43) conclude that research has not established an evidence-based best practice for how much description to request, so descriptions should be taken with care and interpreted with caution.
How Descriptions Are Recorded
The description that reaches court is often not the witness’s own words but an officer’s summary, a broadcast, a form with boxes for height and weight, or an investigator’s written statement, and each step can shorten, standardize, or alter what the witness said. A field study of 30 interviews by serving investigators found that less skilled interviewers obtained less accurate verbal accounts and produced written statements with more inaccuracies than more skilled interviewers did (Walsh et al., 2023). The study was small and not focused on person descriptions, but it supports a point both parties share: a recording or the witness’s own written first account is better evidence of memory than a later summary.
Records also matter because descriptions are used to build lineups. When fillers are chosen to match the description, the brevity of the description, and its silence on many features, shape how the lineup looks and whether it is fair (Lindsay et al., 1994). Filler selection is treated in Lineup Construction.
Descriptions and Later Identifications
Intuition suggests that a witness who describes an offender well should also identify him well; the research supports only a weak version of that idea. Wells (1985) found that faces that were easy to describe were also easier to recognize, but that witnesses who gave good descriptions were not reliably better identifiers. In a study of 62 witnesses to a staged incident, the number of descriptors was modestly related to identification accuracy (r = .28; Sporer, 1992). A meta-analysis found a small but significant relationship between the accuracy of a description and the accuracy of the later identification (r = .14 across 32 tests), and a somewhat larger one for the number of incorrect descriptors (r = .18), while the sheer number of descriptors given was not significantly related (Meissner et al., 2008). Most relationships varied across studies, and they were weakest for live staged events, the paradigm closest to real crimes. A recent laboratory replication again found no reliable relationship between the length or detail of descriptions and lineup accuracy (Fazlić & Deljkić, 2025).
In practice, the number of details a witness gave was weakly related to identification accuracy in one study (r = .28) and not significantly related in a meta-analysis, and the number of wrong details was related to somewhat more, though still weakly (r = .18). Describing and recognizing a face also draw on partly different processes, and describing can sometimes slightly impair later recognition, an effect treated in Verbal Overshadowing.
Limits of the Evidence
Archival studies use real witnesses but cannot check most descriptors, because the offender’s appearance at the time is often unknown and archived cases are those with a suspect or conviction. Experiments know the truth but use low-stakes events, healthy adult volunteers, and often video. Many foundational studies are decades old, and police forms, body-worn cameras, and interviewing have changed since. The description–identification correlations are small and heterogeneous, and the conditions that predicted accuracy explained only a modest share of the variation. Above all, findings about groups of witnesses cannot tell the factfinder whether a particular witness’s description, or a particular estimate, is accurate.
Questions to Ask About This Research
When Kuehn’s (1974, as cited in Lane & Houston, 2021) average of about seven descriptors is cited, counsel can ask what was measured: how much 1970s victims said in police reports, not whether it was correct. When the 7.35 versus 3.94 comparison (Lindsay et al., 1994) is offered, ask whose words were counted, because the real-case figures were taken from newspaper reports. When Van Koppen and Lochun’s (1997) finding that most verifiable descriptors were correct is cited, ask what the denominator was: only descriptors that could be checked against police records, which excludes much of the face. When a witness’s height or age estimate is said to be “off,” ask how far off it is compared with the error typical in studies such as Nyman et al. (2020) and Thorley et al. (2018), and whether the witness’s own height or age could explain the direction of the error. When the meta-analytic correlations (Meissner et al., 2008) are cited, ask whether they concern the number of details or their accuracy, and remember that a correlation of .14 describes a weak tendency across many witnesses, not a test that can be applied to one. These questions are explained in “Reading Research Claims: A Guide for Attorneys” on this site.
Person Descriptions and the Law
Federal and Military Standards
Under the due process framework of Neil v. Biggers (1972) and Manson v. Brathwaite (1977), a court deciding whether an identification obtained through an unnecessarily suggestive police procedure is nonetheless reliable considers the witness’s opportunity to view, degree of attention, the accuracy of the witness’s prior description, the level of certainty, and the time between the crime and the confrontation. Perry v. New Hampshire (2012) limits that screen to suggestion arranged by law enforcement and leaves other reliability concerns to cross-examination, jury instructions, and expert testimony. The research bears on the description factor: a matching description is some evidence of reliability, but because descriptions are brief and general and good describers are only slightly more likely to be accurate identifiers, a match on common attributes adds less than intuition suggests, and a minor mismatch subtracts less.
In courts-martial, Mil. R. Evid. 321 treats an identification process as unlawful if it is so suggestive as to create a substantial likelihood of misidentification, and on an objection the prosecution must show by a preponderance of the evidence that the identification was reliable under the circumstances (Mil. R. Evid. 321(c)(1), (d)(6)(B)(i)). In United States v. Criswell (2018), the Court of Appeals for the Armed Forces, over a two-judge dissent, held that a military judge did not abuse his discretion in admitting an in-court identification that followed a single-photograph display he assumed was unnecessarily suggestive. The court agreed that the witness’s first description, given before she saw the photograph, was not exceptionally detailed, but held that the judge could also consider her later, fuller descriptions, while noting that a judge may give earlier descriptions greater weight. Expert testimony on descriptions is governed by Mil. R. Evid. 702 and United States v. Houser (1993); the December 1, 2023 amendment to Fed. R. Evid. 702, requiring the proponent to show that it is more likely than not that each admissibility requirement is met, became part of the military rule through Mil. R. Evid. 1102 in 2025 (Joint Service Committee on Military Justice, 2025), and practitioners should confirm the current text.
Arizona
A defendant who challenges a pretrial identification in Arizona is entitled to a hearing outside the jury’s presence under State v. Dessureault (1969); if the pretrial procedure was unduly suggestive, the State must show by clear and convincing evidence that the in-court identification is not tainted by it. RAJI (Criminal) Standard Criminal 45 instructs jurors that the State must prove beyond a reasonable doubt that the in-court identification is reliable and lists factors drawn from Biggers and Manson, including “the accuracy of any descriptions the witness made prior to the pretrial identification.” State v. Chapple (1983) is Arizona’s leading case on eyewitness expert testimony; exclusion was an abuse of discretion on its facts, admission otherwise lies within the trial court’s discretion, and State v. Nordstrom (2001) approved limiting an expert to general factors without opinions on a particular witness. Arizona Rule of Evidence 702 adopted the Daubert standard effective January 1, 2012 and was amended effective January 1, 2024 to track the 2023 federal amendment (Ariz. Sup. Ct. Order R-23-0004).
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. Brief descriptions are the norm among real witnesses, including witnesses who later identify the right person, so a short description is not by itself a sign of a poor memory. The general attributes witnesses report are usually correct, and the description–identification relationship is weak in both directions, so a description that omits the face or misjudges height by a few inches is consistent with an accurate identification. Prosecutors can strengthen this evidence by preserving the first description in the witness’s own words, with when and how it was taken and what the witness said they could not see.
The research also counsels caution. A match between the description and the defendant on sex, age range, build, and hair color is a match shared by many people and is weak evidence of identity, and a match on clothing proves little if the suspect was located hours later. Because a good description predicts an accurate identification only weakly, a detailed, confident description should not be presented as confirming the identification. Where the later description has grown more specific, or has converged on the defendant’s appearance after the witness saw a photograph, the earlier version is the better measure of memory.
Special Issues for Defense
The research supports several propositions defense counsel can legitimately present. Descriptions of strangers are usually general and would often fit many people; numerical estimates of age, height, and weight carry wide error that grows with distance and with differences between witness and offender; inner facial features are rarely described and, when described, are often wrong; and a written summary may not reflect what the witness actually said. Expert testimony explaining these findings, applied to the conditions and records in a particular case, can help the factfinder decide how much a description narrows the field.
The research also counsels caution about the limits that apply to the defense. A sparse or partly inaccurate description is common among accurate witnesses and does not show that an identification is mistaken, and the meta-analytic correlations are too small to support that inference. The general attributes witnesses do report are usually right, so a clear mismatch on a basic attribute deserves more weight than a few inches of height. An expert cannot tell the factfinder whether a particular witness’s description is accurate, and testimony that implies otherwise invites exclusion or limitation under Rule 702, Houser, or Nordstrom.
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
Fahsing, I. A., Ask, K., & Granhag, P. A. (2004). The man behind the mask: Accuracy and predictors of eyewitness offender descriptions. Journal of Applied Psychology, 89(4), 722–729. https://doi.org/10.1037/0021-9010.89.4.722
Fazlić, A., & Deljkić, I. (2025). The face recall and face recognition: Description accuracy does not indicate eyewitness identification accuracy. Nauka, bezbednost, policija, 30(2), 96–109. https://doi.org/10.5937/nabepo30-54463
Handler, A., & Frühholz, S. (2021). Eyewitness memory for person identification: Predicting mugbook recognition accuracy according to person description abilities and subjective confidence of witnesses. Frontiers in Psychology, 12, Article 675956. https://doi.org/10.3389/fpsyg.2021.675956
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
Lane, S. M., & Houston, K. A. (2021). Understanding eyewitness memory: Theory and applications. New York University Press.
Lindsay, R. C. L., Martin, R., & Webber, L. (1994). Default values in eyewitness descriptions: A problem for the match-to-description lineup foil selection strategy. Law and Human Behavior, 18(5), 527–541. https://doi.org/10.1007/BF01499172
Meissner, C. A., Brigham, J. C., & Kelley, C. M. (2001). The influence of retrieval processes in verbal overshadowing. Memory & Cognition, 29(1), 176–186. https://doi.org/10.3758/BF03195751
Meissner, C. A., Sporer, S. L., & Schooler, J. W. (2007). Person descriptions as eyewitness evidence. In R. C. L. Lindsay, D. F. Ross, J. D. Read, & M. P. Toglia (Eds.), The handbook of eyewitness psychology: Vol. 2. Memory for people (pp. 3–34). Lawrence Erlbaum Associates.
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
Nyman, T. J., Antfolk, J., Lampinen, J. M., Korkman, J., & Santtila, P. (2020). The effects of distance and age on the accuracy of estimating perpetrator gender, age, height, and weight by eyewitnesses. Psychology, Crime & Law, 27(3), 231–252. https://doi.org/10.1080/1068316X.2020.1798425
Spearing, E. R., & Wade, K. A. (2022). Long retention intervals impair the confidence–accuracy relationship for eyewitness recall. Journal of Applied Research in Memory and Cognition, 11(3), 384–391. https://doi.org/10.1037/mac0000014
Sporer, S. L. (1992). Post-dicting eyewitness accuracy: Confidence, decision-times and person descriptions of choosers and non-choosers. European Journal of Social Psychology, 22(2), 157–180. https://doi.org/10.1002/ejsp.2420220205
Thorley, C., Almond, L., Gregory, A., McAlonan, V., & McLoughlin, A. (2018). An archival analysis of sexual assault victims’ age estimation accuracy when describing stranger offenders. Psychology, Crime & Law, 24(10), 1030–1049. https://doi.org/10.1080/1068316X.2018.1503662
Twedt, E., Crawford, L. E., & Proffitt, D. R. (2015). Judgments of others’ heights are biased toward the height of the perceiver. Psychonomic Bulletin & Review, 22(2), 566–571. https://doi.org/10.3758/s13423-014-0689-z
Van Koppen, P. J., & Lochun, S. K. (1997). Portraying perpetrators: The validity of offender descriptions by witnesses. Law and Human Behavior, 21(6), 661–685. https://doi.org/10.1023/A:1024812831576
Walsh, D., Areh, I., Minhas, R., & Menzies, S.-L. (2023). When law enforcement interview witnesses and write their statements. Police Practice and Research, 24(3), 346–367. https://doi.org/10.1080/15614263.2022.2116436
Wells, G. L. (1985). Verbal descriptions of faces from memory: Are they diagnostic of identification accuracy? Journal of Applied Psychology, 70(4), 619–626. https://doi.org/10.1037/0021-9010.70.4.619
Legal Authorities
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. Nordstrom, 200 Ariz. 229, 25 P.3d 717 (2001).
United States v. Criswell, 78 M.J. 136 (C.A.A.F. 2018).
United States v. Houser, 36 M.J. 392 (C.M.A. 1993).
Fed. R. Evid. 702.
Mil. R. Evid. 321.
Mil. R. Evid. 702.
Mil. R. Evid. 1102.
Ariz. R. Evid. 702.
Ariz. Sup. Ct. Order R-23-0004 (2023) (amending Ariz. R. Evid. 702, effective Jan. 1, 2024).
Revised Arizona Jury Instructions (Criminal), Standard Criminal 45 (Identification) (State Bar of Arizona 2022).
Related topic papers: Memory for Persons and Faces · Face Recognition · Face Composites · The Cross-Race Effect · Verbal Overshadowing · Lineup Construction
Related topics from Arizona Forensic Services: Courts-Martial