Saving faces: Using eye movement, ERP, and SCR measures of face processing and recognition to investigate eyewitness identification

Research on eyewitness identification has repeatedly pointed to the unreliability of eyewitness memory, identifying many factors that can affect and alter what people remember and suggesting ways to question witnesses and conduct lineups to reduce the likelihood of memory distortions. These methods, however, involve direct questioning, and therefore are of little use if the witness does not have conscious access to the memory or is mistaken about what they remember.; Various behavioral and psychophysiological measures of memory can reveal differences in the processing of novel and previously encountered items, even in the absence of conscious recollection. This dissertation examines several measures of unconscious memory and investigates their possible application for face recognition.; Specifically, eye movements (EOG), galvanic skin response (GSR), and event-related brain potentials (ERPs) were recorded concurrently as individuals examined photographs of faces. Faces were studied for short (300–1000m) or long (3000–5000ms) durations; recognition memory was then tested after one of four delay intervals: ∼1/2 hour, one hour, one day, or one week.; Three measures in particular yielded differences between new and old faces regardless of recognition accuracy: (1) the amplitude of an early (N2) ERP component was larger for old than new faces, (2) fixations were more tightly clustered around the inner face features of old than new faces, and (3) old faces were associated with less left-bias fixations during the first second of viewing. All of these effects occurred even in the absence of conscious recollection, and, for the clustering of eye movements and the N2 ERP component, across all four test delays. Other measures tended to differentiate old from new faces, but were more sensitive to test delay and response accuracy. Some measures changed qualitatively across test delay, possibly reflecting the influence of memory consolidation processes. Despite its long use for lie detection, GSR measures were wholly unreliable for our purposes. To use various of these dependent measures for face recognition simultaneously is a feat that required certain technical advances; as some of these algorithms are of general interest (localizing eye fixations using EOG, removing EOG artifacts from the EEG), these will be mentioned in brief.