How the learning of unfamiliar faces is affected by their similarity to already known faces.

Background
While familiar and unfamiliar face processing share some characteristics, such as 'configural' processing [e.g. 1], there is also evidence that they rely on qualitatively different types of information [e.g. 2]. The robust observation of superior performance for familiar, compared to unfamiliar face recognition, suggests that familiarity is associated with a reliable processing benefit. Some authors propose that unfamiliar face matching should be conceptualised as image-matching, rather than involving any face-specific processes [e.g. 3]. Yet, virtually nothing is known about how unfamiliar faces become familiar. Current theoretical models [e.g. 4] suggest that faces are not encoded in isolation, but with respect to existing facial representations, often confined to an 'average' or 'norm' face. However, we are often struck by how similar a newly-encountered person is to someone whom we already know. I make the novel prediction that the encoding of unfamiliar faces is performed in relation to pre-existing face representations.
Design
This project aims to address major deficits in knowledge regarding how the learning of unfamiliar faces is affected by their degree of resemblance to already-known faces. I will manipulate unfamiliar faces' similarity to already known faces using various techniques, including morphing. Explicit learning will be measured by responses as to whether or not a face has been seen before. Critically, I will also measure implicit learning using adaptation designs, participant ratings for various characteristics (i.e. trustworthiness), eye-tracking (including pupil dilation), and event-related potentials (ERPs). Using these methods, I will be able to explore processing differences between familiar, similar-to-familiar, and unfamiliar face perception. As familiar face processing is likely to be confounded with conceptual information [e.g. 5] and ceiling effects, I will using training paradigms to manipulate familiarity.
Analyses
Correlational and regression analyses will be used to explore individual differences, and ANOVAs for group differences (i.e. stimuli that are familiar to some participants and stimuli that are unfamiliar to others) on explicit and implicit indexes of familiarity. Eye-tracking and ERP data will require pre-processing before responses are submitted to statistical testing.
Conclusions
I will explore what happens as an unfamiliar face becomes familiar. This work is theoretically informative. The 'Interactive Activation and Competition' [e.g. 6] model posits that identity and identity-related information are processed separately. While unfamiliar faces rely on visual representations, theoretically, familiar faces could also be processed through conceptual information. Speculatively, if the encoding of unfamiliar faces is performed at least partly in relation to pre-existing face representations, this could have interesting implications for how predictive coding [e.g. 7] might work. This work may be able to test particularly strong existing theories that suggest unfamiliar faces are not faces [3], and begin to develop a satisfactory account of face learning.
Implications
A better understanding of the perceptual, cognitive, and neural mechanisms involved in unfamiliar and familiar face learning will inform existing theories. In the longer term, this work may identify a collection of pre- and post-dictors of facial familiarity. This could have important implications in eye-witness settings