by Christopher Pack, PhD

Approximately half of the human brain's 100 billion neurons are involved in vision. What is it about vision that requires so much processing power? It certainly doesn’t seem difficult to recognize a face or to navigate through a crowded shopping mall, but of course this is precisely the point. Our visual systems are able to interpret the world around us in such way that we are free to think about other things. Exactly how this happens is the focus of the research carried out in the Pack Lab.

The first stop for visual information upon reaching the cerebral cortex, is a region in the back of the head called primary visual cortex or V1. Within V1 each individual neuron is responsible for analyzing the signals arriving from a tiny region of the retina. In other words, each neuron in the primary visual cortex sees the world through a sort of pinhole called a receptive field. Collectively the receptive fields in V1 cover the entire visual field, so that we have the impression that vision is composed of a continuous sampling of space. But the job of V1 is not simply to create a picture of the outside world, as there would not be anyone in the brain to examine the picture even if such a thing were possible. Rather V1 neurons are responsible for finding particular features that may be present within their receptive fields. These features are generally oriented lines or colors. Artists have long been aware that objects and even faces can be approximated from such features, indicating that they are a fundamental aspect of visual perception.

Having decomposed the visual world into simple stimulus features, area V1 sends this information to a large network of brain regions called the extrastriate cortex. These other areas have the job of reassembling the rudimentary lines and colors that V1 neurons have found into a more coherent picture of our surroundings. Thus in the extrastriate cortex one finds specialized brain regions that recognize faces and warn us about upcoming collisions. How individual neurons can perform such complex calculations is one of the deepest questions in neuroscience. It is also a question of some interest to designers of computer algorithms, which generally cannot perform as well as biological visual systems on even the simplest of tasks.

The Pack Lab combines data obtained from single neurons in the visual cortex with mathematical techniques borrowed from computer scientists and engineers to attempt to discover the algorithms used by extrastriate cortex neurons. Success in this endeavour means generating mathematical equations that can predict the responses of neurons to complicated stimuli. Sometimes these equations also make specific predictions about what people will see in particular environments. Such predictions often lead to the development of intriguing visual illusions. This is quite helpful: If we can understand why the brain perceives something that isn’t there, perhaps we can understand more generally how we see.