The neural code for ‘face cells’ is not face specific

Abstract

'Face cells' are visual neurons that selectively respond more to faces than other objects. Clustered together in inferotemporal cortex, they are thought to form a network of modules specialized in face processing by encoding face-specific features. Here we reveal that their category selectivity is instead captured by domain-general attributes. Analyzing neural responses in and around macaque face patches to hundreds of objects, we discovered graded tuning for non-face objects that was more predictive of face preference than was tuning for faces themselves. The relationship between category-level face selectivity and image-level non-face tuning was not predicted by color and simple shape properties, but by domain-general information encoded in deep neural networks trained on object classification. These findings contradict the long-standing assumption that face cells owe their category selectivity to face-specific features, challenging the prevailing idea that visual processing is carried out by discrete modules, each specialized in a semantically distinct domain.

Visual object topographic motifs emerge from self-organization of a unified representational space

“…Some theoretical accounts of these regions consider these as independent and unrelated functional modules, implicitly assuming no direct relationship between them ( Kanwisher, 2010; Zeki, 1978 ) . However, the integrated feature space of the deep neural network allows us to consider an alternate hypothesis that face- and scene-selectivity might naturally emerge as different parts of a common encoding space—one whose features are designed to discriminate among all kinds of objects more generally (Konkle & Caramazza, 2013; Bao et al, 2020; Vinken et al, 2022; Prince & Konkle, 2020; Khosla & Wehbe, 2022). If this is the case, these categories would drive responses in a localized part of the feature space, which would emerge as a localized cluster of selective responses in the SOM.…”

“…Relatedly, Huang et al, (2022) have found that information about the real-world size of objects is encoded along the second principal component of the late stages of deep neural networks ( 44 ). Further, Vinken et al, 2022 recently demonstrated that face-selective neurons in IT could be accounted for by the feature tuning learned in these same object-trained deep neural networks (( 45 ), also see ( 42 , 46 , 43 )). Thus, deep neural networks clearly operationalize a multi-dimensional representational encoding space that has information about these well-studied object distinctions.…”

Cortical Face-Selective Responses Emerge Early in Human Infancy

“…Considerable evidence indicates that the FFA plays a selective and causal role in face perception in adults. In particular, fMRI studies have shown that the FFA responds more to faces than any other stimulus yet tested (1, 71–74) (but see (75)). The FFA has been proposed to support face recognition by encoding face features, to hold information about identity, gender, and expression, and to be engaged both in face detection (i.e., distinguishing faces from other objects) and face identification (i.e.…”

Cortical topographic motifs emerge in a self-organized map of object space

“…Some theoretical accounts of these regions consider these as independent and unrelated functional modules, implicitly assuming no direct relationship between them (2, 58). However, the integrated feature space of the DNN allows us to consider an alternate hypothesis that face and scene selectivity might naturally emerge as different parts of a common encoding space-one whose features are designed to discriminate among all kinds of objects more generally (9,26,36,37,39). If this is the case, then these categories would drive responses in a localized part of the feature space, which would emerge as a localized cluster of selective responses in the SOM.…”

“…Relatedly, Huang et al (38) have found that information about the real-world size of objects is encoded along the second principal component of the late stages of DNNs. Furthermore, Vinken et al (39) recently demonstrated that face-selective neurons in IT could be accounted for by the feature tuning learned in these same object-trained DNNs; also see (36,37,40). Thus, DNNs clearly operationalize a multidimensional representational encoding space that has information about these well-studied object distinctions.…”