Unfamiliar face matching: Decision-making and improvement

Research has consistently found unfamiliar face matching to be a highly error prone task. Yet, little is known about the decision-making process that underlies this task. Furthermore, methods of training observers to improve accuracy have demonstrated mixed success. Therefore, the experiments reported in this thesis investigated how matching decisions to pairs of unfamiliar faces are made (Chapter 2), and evaluated a novel method for improving face matching accuracy (Chapters 3 and 4). Chapter 2 examined whether identifications are based on a series of smaller assessments for individual facial features and if so, how these evaluations are combined to reach an overall decision, by comparing decisions to whole faces with those to isolated feature regions. Individual facial features were found to influence the classification of the whole face disproportionately, but performance was best when all features were presented as an integrated whole. This thesis also explored whether matching performance could be improved by providing observers with clearly-labelled examples (Chapters 3 and 4). The benefit of examples was explored at an individual level and revealed that observers who were low-performing at baseline improved with the help of examples (Chapter 3). This examples advantage was maintained after the examples were removed, generalised to previously unseen stimuli taken from the same set as the target pairs, and also demonstrated some generalisation to stimuli from a new set with different characteristics. Chapter 4 then used eye-tracking to evaluate how examples were utilised during matching tasks, but did not reveal a clear improvement with the provision of examples. The different pattern of results may have been due to fundamental task differences introduced by the eye-tracking methodology. Thus, further research is required to fully explore the feasibility of the examples manipulation as a method for improving unfamiliar face matching