Abstract
Face recognition ability varies substantially across the population, yet the cognitive and computational processes that give rise to this variation remain poorly understood. Here, we combined approaches from experimental, differential and mathematical psychology to test how individual differences in face recognition ability relate to latent parameters of the decision-making processes involved in face recognition. Participants (N = 300) completed a face memory task with learned and novel faces, together with independent measures of face recognition ability (Cambridge Face Memory Test) and general object recognition ability (car recognition). We modelled task performance with an evidence-accumulation model that translates accuracy and response times into latent parameters, including drift rate (the signal-to-noise ratio of the evidence) and threshold (response caution). Higher face recognition ability was associated with greater drift rate discrimination between correct and incorrect response accumulators in the face memory task, for both learned and novel faces and to a similar degree, consistent with expertise effects in other domains. This association remained largely unchanged when accounting for car recognition ability, suggesting that it is at least partly tied to face processing itself. Higher face recognition ability was also associated with greater response caution, which may reflect better calibration to task demands. Although the associations were modest in size, these findings suggest that individual differences in face recognition ability are reflected in distinct computational mechanisms, namely more efficient extraction of information from faces and a more cautious decision policy. This approach enables a more explicit and computational basis for cognitive theories of face recognition to be developed, as well as for testing whether these mechanisms operate similarly across the full spectrum of face recognition abilities.