Eyewitnesses frequently experience criminal events with other people (e.g., Skagerberg & Wright, 2008).  After such events, eyewitnesses sometimes talk with one another about what they witnessed.  These conversations can introduce error into reports of the event and even increase the chance that an eyewitness will make a false identification (e.g., Wright, Memon, Skagerberg, & Gabbert, 2009). As a result of these potential effects, psychological scientists recommend that eyewitnesses be encouraged to avoid conversations with one another and be required to complete identification procedures separately.

The research supported by this grant focused on the potential for influence between witnesses even when recommended procedures are followed. In particular, we tested whether witnesses derive information from the speed with which a co-witness completes an identification procedure. This information might be available in a variety of co-witnessing scenarios. Consider one example: detectives investigating a crime in which witnesses are members of the same family might administer identification procedures to all witnesses in sequence at their home. Knowing that a co-witness completed an identification procedure quickly might communicate information about how easy an upcoming identification task will be. For example, knowing that a co-witness quickly made an identification might suggest that it will be easy to pick the culprit from a set of photos. Information of this sort might affect witnesses’ own decision-making which could have implications for identification accuracy if it increases witnesses’ likelihood of choosing from a target-absent photospread (i.e., when the culprit is not present).

In order to test whether co-witness identification speed affects witness’s identification decisions, three studies were conducted. In the first two experiments, participants were paired with a confederate witness whose identification speed was manipulated. In Study 1 (N = 200) participant-eyewitnesses watched a video with the confederate witness.  After the video was over, the confederate left the room, returning after a fast (10 sec) or slow (4 min) identification. Upon returning, the confederate definitively stated that she identified someone from the photos. Participants then attempted their own identification from a target-absent photospread (i.e., the culprit’s photo was not included). Participant-witnesses made significantly fewer incorrect picks from the photospread when the confederate witness was slow (23%) compared with fast (37%).

In Study 2 (N = 151), co-witness identification speed was manipulated as in Study 1 but the confederate co-witness said that the culprit was not present in the photospread instead of making an identification. Again, the fast co-witness increased false identification rates compared with the slow co-witness. Participant-witnesses made significantly fewer incorrect picks when the confederate witness was slow (30%) compared with fast (47%).

In both studies, the effect of the speed manipulation was present for the female stimulus materials, but not the male stimulus materials. For each set of materials, a set of fillers was generated in which each photo was similar to the respective culprit. The highest similarity photo to the culprit was included in each photospread, along with five fillers selected from a set of photos rated as similar to the culprit.  We tested whether the discrepancy in results for the two sets of videos might be due to the fact that the female high similarity photo was a better proxy for the female culprit than was the male high similarity photo for the male culprit. To obtain ratings of similarity for both sets of materials, a third study (N = 755) was conducted online using Amazon’s Mechanical Turk. In this study, participants watched the male and female stimulus videos (in counterbalanced order).  After each video, participants rated how confident they were that each photo (fillers and culprit) was the person they saw in the video.  The pattern of ratings indicated that the female high similarity photo was a more viable alternative for the culprit than was the case for the male high similarity photo/male culprit. For example, 60.66% (458) participants provide a 0% confidence rating that the male high similarity photo was the culprit; only 40% (302) gave that same rating for their confidence that the female high similarity photo was the culprit.  These ratings suggest that the manipulation of co-witness identification speed might be particularly pernicious when an innocent suspect is a viable alternative for the culprit.

In both experiments manipulating the confederate co-witness’s identification speed, recommended procedures for conducting lineups and photospreads were followed. For example, witnesses were warned that the culprit might not be in the photos, the photospread was presented on a computer so the administrator could not influence the witness’s decisions, and the photospread was not biased toward any individual photo.  That co-witness identification speed still affected witness decisions reinforces calls to separate witnesses for identification procedures.  In addition, these data suggest that witnesses should be informed of a standard duration of any identification procedure so that they cannot be influenced by a co-witness’s identification speed.

Last Modified: 12/24/2019
Modified by: Amy B Douglass