How does familiarity impact the computational processes that support face recognition?

Le résumé fourni par la source

Recognising a familiar face is a fundamental social cognitive ability, yet the computational mechanisms that underlie this capacity remain poorly understood. Prior research has established that familiarity affects face recognition performance, but has relied primarily on manifest measures — accuracy and response time — that cannot distinguish between qualitatively different sources of performance change. Here, we used evidence accumulation modelling across a series of experiments to ask not merely whether familiarity affects face recognition, but how — specifically, which latent computational processes are modulated. Experiments 1 and 2 examined visual familiarity (briefly learned faces versus novel faces) and Experiments 3, 4 and 5 examined person knowledge (famous versus novel faces). Fitting linear ballistic accumulator models to recognition data revealed a consistent computational signature across experiments: novel faces produced substantially higher drift rates than familiar faces, accompanied by a smaller but reliable increase in response caution. The dominance of the drift rate effect localises familiarity’s primary influence to the quality of evidence accumulation — how cleanly the perceptual system extracts decision-relevant information — rather than to response bias. We interpret thes findings as reflecting a novelty pop-out effect combined with unstable memory traces for familiar faces. These findings demonstrate that computational modelling reveals dissociable mechanisms underlying face recognition that are invisible to manifest measures alone, and provide a foundation for understanding how familiarity shapes the decision processes that support social perception.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.