Purchase this article with an account.
Elena Waidmann, Kenji W Koyano, Julie J Hong, Brian E Russ, David A Leopold; Local image features dominate responses of AM and AF face patch neurons. Journal of Vision 2019;19(10):259b. doi: https://doi.org/10.1167/19.10.259b.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The fMRI-defined face patches of the macaque inferotemporal cortex contain neurons that are known to respond more strongly for faces than for other images. Previous studies have measured the tuning of such neurons to parametrized internal facial features. Here we investigated the relative contribution of internal and contextual head, body, and background information in the anterior medial (AM) and anterior fundus (AF) face patches. We developed a stimulus paradigm in which we systematically swapped image elements to investigate the relative influence of internal facial features (eyes, mouth), external facial features, attached bodies, and background scenes. Each test stimulus was a photorealistic composite image containing a combination of the image parts, presented to each of three monkeys briefly during a passive viewing task requiring strict fixation. We recorded local populations of single neurons from AM in one monkey and from AF in two other monkeys, using chronically-implanted microwire bundles. We found that most AM neurons were strongly modulated by either internal or external facial features, but few were modulated by both. For those sensitive to internal features, variation of the upper face (eyes), but not the lower face (nose-mouth) was most important. Although no AM neurons were selective for body parts, AF neurons tended to be weakly modulated by variation in bodies. AF neurons at one recording site were predominantly sensitive to details of the external hair, but not to inner facial features. At the second AF site, neurons were primarily modulated by the lower face, with many selectively firing for an open mouth. None of the neurons in either area were strongly modulated by background scene. In sum, the parts-swapping paradigm demonstrates that face patch neurons exhibit idiosyncratic specialization for internal and contextual facial features that is to a large extent shared within a 1mm3 volume of tissue.
This PDF is available to Subscribers Only