Both bottom-up and top-down processing are involved in the occipital-temporal face network, with the top-down modulation more extensively engaged when available information is sparse in the face images.

An innovative inter-subject stimulus-locked brain activation approach uncovers marked topological differences in a brain network of higher-order visual regions in individuals with a congenital impairment in face recognition compared with controls.

Consistent spatial clustering and fine-scale neural tuning to different face parts were found within the face-processing regions, which may be the neural implementation of efficient neural computation for face identification.

Diverse photographs of human faces against their natural background trigger a specific electrical response in the right hemisphere of the brain in infants aged 4–6 months.

Disrupting the activity of the medial lateral face patch (ML) using fMRI-targeted microinjections of muscimol leads to anatomically and categorically specific impairments in a naturalistic face detection task.

A generative-model-based, unsupervised learning toolbox for characterizing oscillatory bursting and brain network dynamics in univariate or multivariate time series.

Reducing visibility with higher image presentation rates increases recurrent processing demands along the visual processing pathway to resolve object recognition.

Medaka fish were able to use faces for individual recognition, and were slower to recognise inverted faces but not inverted non-face shapes.

Neuronal recordings reveal complex, heterogeneous inputs to face-selective neurons, suggesting that inferotemporal neurons do not represent objects in isolation, but are also sensitive to object relationships that reflect environmental regularities.

Mirror-symmetric view tuning in the macaque AL face patch can be explained by the spatial pooling of learned reflection-equivariant representations.