The discriminability of local cues determines the strength of holistic face processing

Face perception is thought to result from the dynamic interplay between holistic and featural modes of processing. What determines the engagement of each mode is currently unknown. Here, we investigated whether the discriminability of local feature cues is a critical determinant of holistic/featural processing engagement. We estimated the strength of holistic processing based on observers’ failure to discriminate target features independently of the context of distracter features in a congruency paradigm. Feature discriminability was manipulated by varying the dissimilarity of target features parametrically, using morphing. We observed that the size of the congruency effect decayed monotonically as a function of the dissimilarity of the target features. In other words, the more similar the target features the stronger the holistic processing. A correlation analysis confirmed that local feature discriminability reliably predicted holistic engagement at upright orientation. In contrast, when a clear local feature difference was detected, perceptual contamination by the other surrounding features was prevented. This evidence firmly suggests that the interplay between holistic/featural processing depends on the discriminability of the signal provided at the local featural level.

► Face perception relies on the dynamic holistic/featural processing interplay.
► Local feature discriminability determines holistic/featural processing engagement.
► Holistic engagement negatively correlates with local feature dissimilarity.