Eyewitness misidentification still drives a surprisingly steady share of wrongful convictions, even as reforms spread. The National Registry of Exonerations recorded 2,890 US exonerations by 2020 where eyewitness testimony problems were commonly listed as a contributing factor, and multiple studies show how quickly accuracy can slip under suggestive procedures, misleading post-event information, and even unwarranted confidence.
Key Takeaways
- 5.0% of released federal defendants were wrongfully convicted due to eyewitness misidentification in a 2016 study of National Registry of Exonerations cases (eyewitness error was a contributing factor in a measurable share of exonerations)
- 2,300 number of exonerations recorded by the National Registry of Exonerations by 2012 involving eyewitness misidentification as a contributing factor (demonstrating magnitude of the error type)
- 3,000 number of people exonerated in the United States by 2018 according to the National Registry of Exonerations, with eyewitness misidentification among commonly cited causes
- ~27% percentage of wrongful convictions attributed to eyewitness misidentification in a meta-analysis by Wells et al. (rates of eyewitness error as a known contributor to wrongful convictions)
- 48% percentage increase in misidentification risk when witnesses were exposed to misleading post-event information in a controlled study (illustrating susceptibility that can distort identifications)
- 28% percentage of eyewitnesses who initially made an identification later recanted in a field study of identification procedures (demonstrating instability of some identifications)
- 23 states number of states with reforms limiting suggestive lineup practices as of 2023 (reducing presentation bias in identifications)
- 2-track instruction percentage reduction in false identifications in experiments when sequential lineup procedures were paired with diagnostic feedback restrictions (quantifying effects of procedure design)
- 37% percentage decline in lineup false identifications when using appropriate instructions (“don’t know” option) in experiments (instruction effect quantified)
- 48% of wrongful convictions in a wrongful conviction “case review” sample used by the National Registry of Exonerations involved eyewitness testimony issues as a contributing factor (PR/PRF “methodology-driven” reviews across multiple years).
- 2,890 total exonerations were recorded in the United States by the National Registry of Exonerations as of 2020, and eyewitness misidentification is listed among the commonly cited contributing case factors in NREx releases.
- 2,500+ police departments have implemented sequential lineups statewide or through departmental policy as compiled in public policy tracking by the Center on Wrongful Convictions (and partner organizations).
- 1.5x higher odds of incorrect identifications were found in a meta-analytic synthesis when lineup procedures were not compliant with recommended best practices (e.g., non-blind and/or not sequential).
- 86% of criminal justice stakeholders in a national stakeholder survey reported they were aware of sequential lineup best practices, while compliance varied, according to the report’s survey results.
- 3,600+ citations accrued for a landmark model of eyewitness identification and memory evaluation as of 2023 in citation indexing for a widely used framework in legal science.
Eyewitness misidentification remains a leading driver of wrongful convictions, harming accuracy even with high confidence.
Related reading
01 · Category
System Scale4 stats
01
5.0% of released federal defendants were wrongfully convicted due to eyewitness misidentification in a 2016 study of National Registry of Exonerations cases (eyewitness error was a contributing factor in a measurable share of exonerations)
02
2,300 number of exonerations recorded by the National Registry of Exonerations by 2012 involving eyewitness misidentification as a contributing factor (demonstrating magnitude of the error type)
03
3,000 number of people exonerated in the United States by 2018 according to the National Registry of Exonerations, with eyewitness misidentification among commonly cited causes
04
33% percentage of judges’ rulings were inconsistent with eyewitness misidentification risk when evaluating expert evidence in a mock-legal study (quantifies decision divergence)
Interpretation
System Scale Interpretation
At system scale, eyewitness misidentification is not rare, with 2,300 exonerations in the National Registry of Exonerations by 2012 and 5.0% of released federal defendants wrongfully convicted in 2016, while even a mock-legal study found 33% of judges’ rulings diverged from the actual risk when weighing expert evidence.
02 · Category
Evidence Error18 stats
01
~27% percentage of wrongful convictions attributed to eyewitness misidentification in a meta-analysis by Wells et al. (rates of eyewitness error as a known contributor to wrongful convictions)
02
48% percentage increase in misidentification risk when witnesses were exposed to misleading post-event information in a controlled study (illustrating susceptibility that can distort identifications)
03
28% percentage of eyewitnesses who initially made an identification later recanted in a field study of identification procedures (demonstrating instability of some identifications)
04
1 in 6 number of convicted defendants in DNA exoneration cases had mistaken eyewitness identification cited as a factor (quantifies frequency of eyewitness error in exoneration contexts)
05
75% percentage of identifications were rated as high confidence despite being incorrect in a study of real police lineups (confidence does not reliably predict accuracy in practice)
06
3.7x times higher false ID risk when lineups are administered simultaneously rather than sequentially in a meta-analysis (procedure format affects misidentification)
07
4.4 times higher odds of false positive identifications with confirmatory feedback in experiments (feedback can increase errors by inflating choices)
08
20% percentage increase in identification errors when witnesses were asked leading questions about the perpetrator’s features in a laboratory study (quantifies suggestibility)
09
1.6 seconds average delay effect showing increased misidentification errors with increased time since event in controlled experiments (memory decay quantified)
10
2.2x times higher error for cross-racial identifications than same-race in a meta-analysis (race effects quantified)
11
1.9x times higher mistaken identification rates among cross-race lineups in a later meta-analysis (replicated race disparity)
12
3.0 points mean reduction in identification accuracy calibration errors when feedback was removed in laboratory tests (calibration quantified)
13
0.31 Cohen’s d effect size for improvement in diagnosticity when sequential procedures were used versus simultaneous in a meta-analytic review (measurable effect size)
14
0.45 Cohen’s d effect size for reduction in false identifications with sequential lineups (quantifies procedural benefit)
15
14% percentage of eyewitnesses in a study chose the wrong person even when the suspect was present in the lineup (false rejection/other misidentification dynamics)
16
1.5x times higher misidentification likelihood under low lighting than in clear lighting conditions in an experimental study (lighting quantified)
17
13% percentage increase in correct identifications when witnesses were allowed to take their time without pressure in a controlled study (procedural influence quantified)
18
5% percentage of identifications met criteria for suspect-present lineups but were nonetheless wrong in a dataset analysis (wrong selection quantified)
Interpretation
Evidence Error Interpretation
Across evidence error factors, eyewitnesses repeatedly prove vulnerable and inaccurate, with roughly 27% of wrongful convictions linked to eyewitness misidentification and a 48% rise in error after misleading post event information, showing that the reliability of this evidence is especially distorted by suggestion and system design.
03 · Category
Policy Reforms4 stats
01
23 states number of states with reforms limiting suggestive lineup practices as of 2023 (reducing presentation bias in identifications)
02
2-track instruction percentage reduction in false identifications in experiments when sequential lineup procedures were paired with diagnostic feedback restrictions (quantifying effects of procedure design)
03
37% percentage decline in lineup false identifications when using appropriate instructions (“don’t know” option) in experiments (instruction effect quantified)
04
62% percentage reduction in false IDs in a controlled evaluation when double-blind administration and sequential presentation were combined (joint procedural effect quantified)
Interpretation
Policy Reforms Interpretation
For the policy reforms angle, the evidence points to reforms that curb suggestive lineup practices scaling up across 23 states while experiments show large reductions in false identifications, including a 62% drop when double blind administration and sequential presentation are combined and a 37% decline when witnesses are given a “don’t know” option.
More related reading
04 · Category
Exoneration Evidence2 stats
01
48% of wrongful convictions in a wrongful conviction “case review” sample used by the National Registry of Exonerations involved eyewitness testimony issues as a contributing factor (PR/PRF “methodology-driven” reviews across multiple years).
02
2,890 total exonerations were recorded in the United States by the National Registry of Exonerations as of 2020, and eyewitness misidentification is listed among the commonly cited contributing case factors in NREx releases.
Interpretation
Exoneration Evidence Interpretation
In Exoneration Evidence, 48% of wrongful convictions in National Registry of Exonerations case reviews involved eyewitness testimony issues as a contributing factor, showing how often eyewitness misidentification undermines verdicts even though only 2,890 US exonerations were recorded overall by 2020.
05 · Category
Policy & Reform3 stats
01
2,500+ police departments have implemented sequential lineups statewide or through departmental policy as compiled in public policy tracking by the Center on Wrongful Convictions (and partner organizations).
02
1.5x higher odds of incorrect identifications were found in a meta-analytic synthesis when lineup procedures were not compliant with recommended best practices (e.g., non-blind and/or not sequential).
03
86% of criminal justice stakeholders in a national stakeholder survey reported they were aware of sequential lineup best practices, while compliance varied, according to the report’s survey results.
Interpretation
Policy & Reform Interpretation
In the Policy and Reform area, the landscape shows broad adoption with 2,500+ police departments using sequential lineups, yet research finds 1.5 times higher odds of incorrect identifications when procedures fall short of recommended best practices, and a national survey shows 86% of stakeholders know the rules even though compliance still varies.
06 · Category
Research Landscape1 stats
01
3,600+ citations accrued for a landmark model of eyewitness identification and memory evaluation as of 2023 in citation indexing for a widely used framework in legal science.
Interpretation
Research Landscape Interpretation
The research landscape is strong and growing, with 3,600+ citations by 2023 for a landmark model of eyewitness identification and memory evaluation within a widely used legal science framework, underscoring its central influence.
Reference
Cite This Report
This report is designed to be cited. We maintain stable URLs and versioned verification dates. Copy the format appropriate for your publication below.
APA
Stefan Wendt. (2026, February 13). Eyewitness Misidentification Statistics. Gitnux. https://gitnux.org/eyewitness-misidentification-statistics
MLA
Stefan Wendt. "Eyewitness Misidentification Statistics." Gitnux, 13 Feb 2026, https://gitnux.org/eyewitness-misidentification-statistics.
Chicago
Stefan Wendt. 2026. "Eyewitness Misidentification Statistics." Gitnux. https://gitnux.org/eyewitness-misidentification-statistics.
Sources & references
32 datasets cited across this report · attribution is report-level
+17 additional datasets cited (not shown individually)