Facial composite production : development of a new technique to identify perpetrators in a more reliable way

Eyewitness misidentifications play an important role in wrongful convictions nationwide. Research suggests that this might be due to the mismatch between how faces are represented in memory (holistically) and how current facial composite systems attempt to retrieve this memory (by individual features). This study examined whether the reverse-correlation image classification technique proves to be a suitable tool for face composite production. Furthermore we are suggesting that factors such as prejudice or being the member of an in- or out-group have an impairing effect on the usability of the composite images. In the first part of the study we assessed the participants’ level of prejudice by means of an IAT and let them construct a composite sketch with the use of the reverse-correlation image classification technique. Resemblance judgments of participants of the second study show that the composite sketches for offenders of an in-group resemble the actual offender significantly more than those made for offenders of an out-group. Furthermore, whether an eyewitness is implicitly prejudiced does not have any effect on the usability of the composite sketches. Recommendations for future research are suggested.