Monkey Vision

The macaque visual system is one of the most astonishing pieces of machinery in the universe, consisting of over three dozen different areas specialized for different aspects of vision.  We have been focusing our attention on a set of six regions in the temporal lobe called “face patches,” that show stronger blood flow to faces compared to non-face objects in fMRI scans (Figure 1).  Our lab has shown that these regions are strongly and specifically connected to each other, and the regions are functionally distinct. The system is a remarkable gift from Mother Nature for studying object perception–in the vast jigsaw puzzle of object perception, we’ve found a set of pieces that belong together.

One reason the face patch system is especially interesting is that its nodes span multiple levels of processing complexity, from early form detection to social decision making. We are currently studying the system at three different levels:

  Macaque Face Patch

Figure 1. The macaque face patch system. Six face patches are shown on an inflated right hemisphere of the macaque brain. 

1) Understanding the interaction between face patches and earlier retinotopic cortex. 

A large body of evidence suggests that object detection should aid object segmentation. For example, in the classic “Dalmation dog” display, knowing that a dog is present is essential to perceiving the dog pop out from the background        (Figure 2). By recording simultaneously from face patches and retinotopic cortex, we can dissect the neural circuits linking object detection to segmentation, to should shed light on basic questions about object representation, including how object identity is bound to object position.  This opportunity is unprecedented: previously, it was not possible to access populations of posterior inferotemporal cells coding a specific object shape, thus it was impossible to explore in detail the coordination of object representations between posterior inferotemporal and retinotopic cortex.

The Dalmation Dog

Figure 2. What is happening between posterior IT and V2/V4 when one suddenly perceives the Dalmation dog?

2) Understanding the nature of face representations within different face patches.

What is the “face space” being represented in each face patch?  At a phenomenological level, we have discovered a special group of cells in the most anterior face patches that respond extremely sparsely and selectively to only a few individuals, invariant to view.  What features are these cells detecting? More generally, can we understand the features being coded by individual cells in such a way that we can predict their response to an arbitrary face?  What are the transformations between face patches, i.e., how does the face representation in one patch become transformed to that in a downstream patch (a sensible question because we know the patches are strongly and specifically connected to each other anatomically)?  How does activity in each patch contribute to face perception?

3) Understanding the interaction between face patches and the “social brain.”

Face patches exist because faces are so socially important to primates. Observing someone make a millisecond glance at a watch can tell us all we need to know about a person. Many debilitating brain illnesses are marked by abnormal face perception, including autism and schizophrenia. The face patch system gives us a unique entrée into the social brain, through study of coordinated activity between the face patches and their downstream targets in the medial temporal lobe, prefrontal cortex, claustrum, pulvinar, and amygdala. How does dynamic activity in this ‘extended face system’ enable real-time integration of information about the mood and intentions of those around us