Virtual fMRI brown bag: November 20
Please join us for a talk given by J. Brendan Ritchie, a postdoctoral fellow working in the lab of Dr. Leslie Ungerleider at the NIH.
Time: 12:00-1:00pm
Place: Zoom
Lost in face space? New frontiers in face perception and its neural basis
Abstract
If you asked someone to guess what facet of visual perception preoccupies vision scientists “faces” would likely not be their first answer. Yet, as any vision scientist can tell you, faces are “special”. At least, they are for the visual system. Our capacity to distinguish the identities of faces is so impressive because, treated as a visual stimulus, the properties of any two faces are overall extremely similar to each other and yet we are able to recognize familiar faces across identity preserving transformation of viewpoint or changes in facial expression. Naturally, this has generated a great deal of research on how the visual system is able to encode information about face identities when faced with such variation. One fruitful theoretical approach has been to characterize face identities as occupying a space where dimensions reflect possible parametric variation in the form and position of face features. This approach also gels naturally with neuroimaging research and artificial neural network models, each of which also trade in multidimensional representations that can be characterized in terms of an encoding space.
Research on face identity is perhaps necessarily preoccupied with the representation of human faces. However, our ability to detect faces in our environment is certainly not restricted to our conspecifics. We can readily detect faces of animals that have radically different facial morphology than our own or experience face pareidolia when we perceive arrangement of features of inanimate objects as being-face like. These instances of face detection cannot easily be accommodated by the enticing face space approach. I review recent neuroimaging and behavior studies in humans and monkeys on face perception of animal faces and face pareidolia, which suggest these varied forms of face detection may be far more fundamental to understanding face perception than commonly appreciated. An important consequence of this is that any ecologically valid model of face perception will have to be able to account for how the same system of representations can encode information not just about human face variation, but face features of animals and objects that are very much unlike our own.