Face and Scene Learning: A Trade-off in the Context of Appearance Variability

In daily life, we meet individuals in varying contexts, and their appearance may change to varying degrees between encounters. Although variability in appearance may initially make face learning more challenging, prior research has demonstrated that it ultimately enhances recognition accuracy (Devue & de Sena, 2023, Cognition). Here, we investigated whether memory advantages from exposure to variable appearances come at the cost of memory for contextual information and for face-scene associations. We hypothesised that inner features of faces that were learned with variable appearances would be memorised in a robust but decontextualized manner, leading to improved recognition performance for faces but not for scenes or face-scene associations. In contrast, we expected that faces with a stable appearance and the surrounding scenes should be memorised in a global manner, yielding relatively poorer face recognition with better scene recognition and stronger face-scene associations. To test these hypotheses, we conducted a pre-registered study (N = 64) examining how variability in appearance (stable vs. variable hair colour and style) influences face and scene learning. Participants learned six images of eight new faces, each embedded in two distinct scenes. After a 5-min delay, face recognition was tested using pictures displaying only inner facial features. Scene recognition and memory for face-scene associations were also tested to investigate memory for the extrinsic context and decontextualisation, respectively. Eye-tracking data were collected throughout the entire experiment to determine the spatial distribution of visual attention. Results show better overall recognition for scenes associated with stable faces and for their association. A trend also suggests that exposure to variability may improve inner-feature recognition, even though these features were fixated less in that condition. Taken together, these findings suggest that variability in facial appearance favours the encoding of inner facial features at the expense of contextual information.