Eyewitness Testimony

Eyewitness misidentifications are, unfortunately, extremely common in the courtroom. Often, high-impact and consequential cases are decided by the testimony of a single eyewitness. When an eyewitness testifies in the courtroom, they are almost always highly confident in their observation. However, many eyewitnesses will also report that they initially had very low confidence in their identification of the suspect. What causes this change in confidence?

In a court of law, eyewitness testimony is often seen as paramount in proving the guilt of a suspect. However, we know that memory is malleable. By the time of trial, an eyewitness’s memory has been impacted by myriad external factors and has almost certainly been contaminated. Typically, police lineups and other forms of eyewitness “testing” are not done in a vacuum; i.e., they often are conducted multiple times with the same eyewitness, and confidence statements are frequently collected at a later time. These practices introduce repetition effects and other forms of bias into an eyewitness’s testimony, rendering it largely unreliable for a court of law.

Empirically, confidence ratings (asking an eyewitness to report their confidence in their identification) can be a useful tool in assessing the accuracy of such identifications. However, these confidence ratings are only helpful under certain conditions. Empirically, for an initial lineup of suspects combined with an uncontaminated eyewitness, high-confidence ratings indicate high levels of accuracy, and vice versa. However, this relationship does not hold true for any repeated tests because identification and memory effects have already contaminated the eyewitness.

In the fall of 2022, I worked with Dr. Scott Gronlund in the OU Memory Lab to research eyewitness memory. Specifically, we used signal detection theory as our theoretical framework to assess how we might make the most of eyewitness reports. Once we have eyewitness data, we are unable to improve the quality of the report; however, we can analyze ROC curves to adjust our sensitivity and our specificity to achieve the ideal balance between false positives and false negatives.

To learn more about this topic, I encourage the reader to investigate the work of Dr. Scott Gronlund.