Match me if you can: Evaluating professional training for facial image comparison

Verifying the identity of an unfamiliar person by examining their face is a surprisingly error-prone task – approximately one in every four decisions made will be incorrect. Despite this high error-rate, unfamiliar face matching tasks form the basis of many important applied forensic and security procedures. To mitigate the risk of errors on face matching tasks, professional facial comparison staff undergo training to improve their ability. Although specifically designed to improve facial identification accuracy, the content, structure and effectiveness of these training courses is unknown and untested. The work reported in this thesis therefore investigates whether current professional training techniques successfully improve face matching accuracy.A review of professional training courses revealed that trainees are typically taught facial anatomy, principles of photography, and are encouraged to make featural comparisons. Practical exercises where trainees receive feedback on the accuracy of their decisions are also common. Although the psychology literature indicates that these methods could be effective, there is a lack of empirical evidence. Thus, a series of eleven experiments validated each of the aforementioned facial identification training techniques.These experiments indicated that facial anatomy and forensic photography training do not improve face matching ability, and neither does morphological classification – a type of featural processing where facial features are classified and compared according to pre-defined categories. Feedback training and facial feature similarity judgements were however, more successful. Feedback training resulted in a long-lasting improvement in observers’ ability to detect non-matching identities, and rating facial feature similarity resulted in an increased ability to detect matching identities. The feature similarity ratings themselves were predictive of identity, and when aggregated across individuals, resulted in almost perfect discrimination. Trained facial examiners also outperformed novices by using a qualitatively different strategy, characterised by an increased reliance on featural face processing.The findings in this thesis not only contribute to our understanding of how to improve performance on this difficult perceptual task, but are also informative of the cognitive processes underlying face matching decisions. The training methods identified in this thesis provide the opportunity for organisations to adopt evidence-based training procedures for staff performing security-critical face matching tasks