Eyewitness Identification: A Systematic Investigation of Lineup Composition and Fairness, United States, 2019-2022 (ICPSR 38761)

Time Period(s) View help for Time Period(s)

Date of Collection View help for Date of Collection

Study Purpose View help for Study Purpose

The overall purpose of this project was to investigate photo array composition in order to improve eyewitness identification procedures. The original research questions were as follows (including in-text citations for the relevant publications):

1. What is the effect of varying levels of photo array fairness on empirical discriminability? (Carlson et al., 2021; Jones, 2021)
2. What are the boundary conditions of photo array fairness on discriminability? (Carlson et al., 2021; Jones, 2021)
3. Is the Diagnostic Feature Detection (DFD) hypothesis a useful theory of photo array/lineup fairness and empirical discriminability? (Carlson et al., 2021; Wooten et al., 2020)
4. How exactly does photo array composition (i.e., the nature of the fillers) drive fairness? (Carlson et al., 2019, 2021; Jones, 2021; Wooten et al., 2020)
5. Is it better to match fillers to the perpetrator's description or to the suspect? (Carlson et al., 2019)
6. How is the confidence-accuracy relationship affected by different levels of photo array fairness? (Carlson et al., 2019, 2021)

Study Design View help for Study Design

Within the project, experiments generally followed a similar design, depending on whether participants played the role of an eyewitness or a potential juror. Target/perpetrator faces were selected from a specified faces database, while fillers that matched the target's description were selected from various U.S. state prison databases. Target faces were typically male, 18-25 years old, White, brown- or blonde-haired, and clean-shaven.

For eyewitness-type experiments, participants would be presented with a crime vignette or mock crime video, with a facial image representing the perpetrator. Next, participants would engage in a distractor task (e.g., watch a YouTube video) and answer attention check items. Finally, participants would be presented with a randomly-assigned identification procedure and be asked to identify the perpetrator from the lineup and enter their confidence level.

For juror-type experiments, participants would be presented with lineups with an eyewitness's statement (condition-dependent) and were asked to rate the accuracy of the identification.

Users can refer to the publications under Data-related Publications for more details of each experiment set.

Jones et al. 2020: For each target, the team created lineups in which the target was present (TP) and absent (TA). To add a distinctive feature, the team used Photoshop to add a black eye on each target and recreated on all fillers. Experimental cells were created: black eye vs. no black eye, target-present vs. target-absent, and distinction matching between target and lineup (distinct/distinct, non-distinct/non-distinct, distinct/non-distinct).

Pleasant 2021: Several versions of targets and fillers were edited with Photoshop to add distinctions (e.g., tattoos, piercings) to internal (eyes, eyebrows, mouth, nose) or external (hair, ears, chin, neck) facial features. The researcher then created a target-present and target-absent lineup for each target and for each internal/external distinctive feature. Experimental cells were created varying on length of target presentation, presence of distinctive feature, and whether the lineup and target matched on distinction.

Carlson et al. 2023 (eyewitness sleep): Three separate experiments were conducted. In Experiment 1 (E1), participants watched a mock crime video, then completed the distractor task and items on sleep quality, and finally were randomly assigned to view a target-present or target-absent lineup. In Experiment 2 (E2), the team used the E1 design but with showups instead of lineups. In Experiment 3 (E3), the team used the E1 design , but a 48-hour delay was introduced in between participants viewing the mock crime video and completing the lineup identification.

Lockamyeir et al. 2023: Three separate experiments were conducted. In E1, participants were randomly assigned to one of eight conditions based on different confidence statements (low vs. high confidence, immediate ID vs. ID at courtroom trial). As mock jurors, they were presented with a lineup and eyewitness confidence statements and asked to rate the suspect's guilt and eyewitness's credibility. In E2, the team used the E1 design and added the presence of court instructions (instructions vs. no instructions) to the confidence statement condition. In E3, the team used the E2 design and added the presence of expert witness instructions to the instructions condition (none, court only, expert only, court + expert).

Carlson et al. 2019: Two separate experiments were conducted. In E1, target-present and target-absent lineups were created with a condition where the number of facial features (eyes, nose, mouth) varied between the target and fillers (1 different feature vs. 2 vs. 3). Participants were recruited from university psychology department subject pools. In E2, lineups were created where fillers either matched the suspect or the written description. Participants viewed a mock crime video. NOTE: E1 data is not available.

Wooten et al. 2020: Participants were randomly assigned to view a lineup of varying size (1, 3, 6, 9, or 12 faces), with target-present vs. target-absent conditions.

Carlson et al. 2023 (memory strength): Fillers were created to match the description but were high-similarity or low-similarity to the target. Participants were randomly assigned to view either a showup, a 3-person lineup, or a 6-person lineup.

Lockamyeir et al. 2021: When initially viewing the target, participants were randomly assigned to view either one or two target faces. If there were two targets, they were either similar or dissimilar, and presented either simultaneously or sequentially. Target-absent and target-present conditions were also used for lineups.

Carlson et al. 2022: Three separate experiments were conducted. In E1, participants were randomly assigned to view an eyewitness statement that varied based on identification speed (slow vs. fast) and confidence (high vs. moderate vs. low), with no statement as the control. In E2, the team used the fair lineups and other design elements from E1, with the addition of biased showups/lineups. Participants were also asked to rate the lineup fairness when identifying the target. In E3, participants were presented with the same fair lineups as E1, but a condition of eyewitness decision was added (eyewitness chose filler vs. eyewitness chose no one) and participants were asked instead to rate suspect guilt.

Hemby (thesis): Lineups were created varying targets and fillers based on the presence of a distinctive feature (scar or tattoo), whether the target and fillers matched on having a distinctive feature or not having a distinctive feature, and if the distinctive feature itself was identical for both target and fillers (replication vs. variant).

Carlson et al. 2021: Target conditions included showing either a full face or a cropped version with only internal features (full vs. internal) at the initial presentation, full vs. internal at the presentation for participant rating, and showup vs. lineup.

Lockamyeir et al. 2020: Two separate experiments were conducted. In E1, participants viewed a mock-crime video, which was filmed at three different distances (3m, 10m, and 20m). Participants were recruited from university psychology department subject pools. E2 used the same design as E1 but utilized a SurveyMonkey sample and reduced the number of targets to two from three. NOTE: E1 data is not available.

Sample View help for Sample

Each set of experiments recruited adult participants across the United States via SurveyMonkey and had a different final analytic sample size. Demographics information for each sample is not available in the data but is available in the relevant publication.

Jones et al. 2020: n = 4,218

Pleasant 2021: n = 6,638

Carlson et al. 2023 (eyewitness sleep): n = 3,836 (E1); n = 2,714 (E2); n = 2,073 (E3)

Lockamyeir et al. 2023: n = 1,033 (E1); n = 1,579 (E2); n = 3,287 (E3)

Carlson et al. 2019: n = 1,965

Wooten et al. 2020: n = 10,433

Carlson et al. 2023 (memory strength): n = 13,728

Lockamyeir et al. 2021: n = 5,509 (E1); n = 3,014 (E2)

Carlson et al. 2022: n = 1,820 (E1); n = 4,169 (E2); n = 3,029 (E3)

Hemby (thesis): n = 5,094

Carlson et al. 2021: n = 19,414

Lockamyeir et al. 2020: n = 3,762

Time Method View help for Time Method

Universe View help for Universe

Adults living in the United States.

Unit(s) of Observation View help for Unit(s) of Observation

Data Type(s) View help for Data Type(s)

Mode of Data Collection View help for Mode of Data Collection

Description of Variables View help for Description of Variables

Variables common to multiple datasets (please see each data file's "Codes" sheet for experiment-specific variables and conditions):

• Hit = correctly identified the target
• False alarm (FA) or Filler = incorrectly identified a filler as the target
• Rejection = in a lineup where the target was present, participant determined the target was not present
• Confidence = participant's level of confidence (0-100) in identifying the target
• Target-absent (TA) = target was not included in a lineup
• Target-present (TP) = target was included in a lineup
• Exposure = number of seconds participants were shown the target face
• Internal features = features present on internal parts of face (eyes, nose, mouth)
• External features = features present on external parts of face or head (neck, ears, hair)

Presence of Common Scales View help for Presence of Common Scales

Used in Carlson et al. 2023 (eyewitness sleep):

St. Mary's Sleep Questionnaire

Stanford Sleepiness Scale

Original Release Date View help for Original Release Date

2025-06-26