The Role of Within-Person Variability Information in Unfamiliar Face Identification

<p dir="ltr">Unfamiliar face identification is known to be highly error prone, yet it is relied upon in various high-stakes practical contexts including border security and the criminal justice system. Conversely, people are highly adept at identifying familiar faces and thus familiarity transforms our ability to identify faces. One factor that makes familiar face identification easier is the encoding of how the face can vary in appearance, or within-person variability information. An understanding of within-person variability could form a crucial part of face learning. Based on this logic, researchers hypothesized that unfamiliar face identification could be improved by providing within-person variability information in a recognition task. Since this idea was first proposed, many studies have found evidence that this method improves unfamiliar face identification. The presentation of within-person variability information in the form of image arrays containing multiple naturalistic images of an identity has been shown to improve people’s ability to match unfamiliar faces relative to single images. However, some studies have failed to replicate this multiple image advantage, and recent research has produced directly conflicting results regarding the circumstances under which such an advantage is observed. This has led to the suggestion that the multiple image advantage may be dependent upon the way information from an array is encoded for identification and thus will only be observed in tasks that include a memory component or require observers to abstract an identity representation from the array. This thesis explored the role of within-person variability information from multiple image arrays in unfamiliar face identification, with a novel focus on clarifying the circumstances under which a multiple image advantage will occur in unfamiliar face matching tasks and understanding how people are encoding information from multiple image arrays. Studies 1 and 2 tested various experimental conditions and their effect on the multiple image advantage. Study 1 manipulated level of variability in the images, the presentation style of images, and decision types, while Study 2 manipulated the positioning of images on the screen. Together, these studies found that task conditions do influence the presence and magnitude of a multiple image advantage. In particular, task conditions that encourage observers to abstract information from an array are conducive to a multiple image advantage. To investigate the role of an abstracted identity representation in the multiple image advantage, Study 3 tested how people encode information from a multiple image array. The results of this experiment found that participants were best at matching targets to arrays when the targets were averages of all the array images or images from the array. This suggests that when observers are presented with a multiple image array in a sequential matching task, they create an averaged representation and also retain information about individual images (exemplars). Finally, Study 4 tested whether how information from multiple image arrays is encoded can be manipulated by task instructions. Results showed that task instructions which directed participants to create an averaged identity representation from an array were not effective in producing a multiple image advantage. The findings of this thesis support the idea that abstraction of an identity representation for use in unfamiliar face identification is what drives the advantage for multiple image arrays over single images. Consequently, experimental conditions that encourage abstraction of identity information from an array are more likely to produce a multiple image advantage. This thesis demonstrated for the first time that this identity representation integrates information about variant and invariant features of a face and can be robust enough to facilitate both recognition and discrimination of unfamiliar faces. In this way, multiple image arrays can facilitate early face learning for the improvement of unfamiliar face identification.</p>