Investigating the Reliability of Those Who Provide (and Those Who Interpret) Eyewitness Confidence Statements

Increasing evidence shows that high confidence at the time of the initial identification is a strong predictor of accuracy, so long as proper lineup administration procedures are followed (Wixted & Wells, 2017). This strong relationship between high confidence and accuracy is documented in many laboratory studies, using a variety of manipulations (e.g. weapon vs. no weapon, other-race identifications) and stimuli (e.g., identifications after viewing photos of faces, videos, and/or staged crimes). In this thesis, I present research from our lab that raises important caveats to the growing consensus about a strong relationship between eyewitness confidence and accuracy. Part I shows that individual differences in face recognition ability influence the rate of high confidence errors. Specifically, weaker face recognition ability corresponds to increased rates of high confidence errors in both a controlled eyewitness experiment using criminal lineups (Study 1A), and in an uncontrolled ‘real-world’ face recognition task of actors from the popular television show Game of Thrones (Study 1B). Part II shows that the probative value of eyewitness confidence statements depends on evaluators (e.g., police officers, judges, jurors) properly interpreting the level of certainty the witness intended to convey. In three experiments (Study 2A – C), participants systematically misinterpreted witnesses’ verbal confidence statements when they knew the identity of the suspect in a criminal lineup – a situation that is common in criminal justice decisions. Taken together, these studies suggest a degree of caution is warranted when using eyewitness confidence as an indicator of accuracy.