The effects of explicit “Not Present” and “Don't Know” response options on identification decisions in computer‐administered lineups

Abstract

Eyewitness researchers recommend that "not present" and "don't know" response options should be presented with police lineups. Although it is important that witnesses-most of whom are unlikely to be familiar with the identification task-are fully cognizant of all response options available to them, an understanding of how explicit non-identification options affect performance is lacking. Across four experiments, including 3,633 participants and 8 different stimulus sets, we tested the effects of including non-identification options in computer-administered lineups. When explicit non-identification options were presented, target-present and-absent choosing decreased. This decrease in choosing was characterized by a shift from filler identifications to lineup rejections. ROC analyses revealed that there was no overall difference in discriminability between guilty and innocent suspects depending on response option condition. On balance, the findings suggest that, in addition to informing witnesses about acceptable responses, displaying non-identification response options does not undermine identification performance.

Estimator variables can matter even for high-confidence lineup identifications made under pristine conditions.

“…Indeed, the high-confidence point on the ROC is slightly higher for the condition in which the opt-out option was not available. Examining the supplemental data of Lucas et al (2020b), we calculated that high-confidence accuracy was approximately the same for both the condition in which there was an opt-out option (93.48%) and the condition in which there was not an opt-out option (92.82%). At a theoretical level, according to Semmler et al (2018) and Stretch and Wixted (1998a), when participants are aware that the conditions are not conducive to good memory, they will tend to set a very strict high-confidence decision criterion.…”

“…It is plausible that the inclusion of such potential opt-out witnesses would primarily affect low-confidence identifications (Weber & Perfect, 2012). For instance, Lucas et al (2020a) found that including an optout option decreased filler identifications and increased the mean confidence in filler identifications, consistent with the opt-out option primarily affecting low-confidence identifications. This interpretation is also confirmed by examining the high-confidence point on their receiver operating characteristic (ROC; i.e., the low/left point), which was virtually identical for participants who were given an opt-out option and those that were not.…”

“…To address this issue, we borrow from survey research and oversample from the poor viewing conditions (Vaughan, 2017). Oversampling has been used occasionally in eyewitness identification research to equate samples with differential participant loss (Lucas et al, 2020;Palmer et al, 2013). However, our use of oversampling was somewhat different and for a different purpose.…”

The role of experimenter familiarity in children’s eyewitness identification

“…Across a series of experiments combining ‘not there’ and ‘not sure’ options with adults, Lucas et al. ( 2020 ) found that the ‘not sure’ option was rarely selected and that the inclusion of an explicit ‘not there’ option within the lineup decreased false identifications and improved accuracy on TA lineups.…”

Eyewitness recall, lineup identification, and verbal and visual memory among chronic cannabis users after a minimum of 24‐h abstinence

“…Additionally, the target‐absent lineup was designed using the “removal‐without‐replacement” method to reduce sources of potential bias (i.e., removing the culprit and inserting a new filler); however, this design has been found to lower participants' decision criterion when making a lineup selection (see, e.g., Smith et al, 2019), potentially influencing accuracy in the present study. The absence of an “I don't know” response may have also affected participants' decisions to choose an individual from the lineup (i.e., reduced filler identifications; Lucas et al, 2020), thereby influencing rates of accuracy; however, prior research evaluating the effects of cannabis on lineup identification had very few “do not know” responses, requiring their elimination from analyses (Vredeveldt et al, 2018). Testing fatigue is another potential confound; however, learning and memory performance were average in the present sample, and research shows that some individuals experience no performance change in response to fatigue while others demonstrate increased performance over time (Strober & DeLuca, 2013).…”