The last fifty years have seen major advances in our understanding of how we perceive images. In particular, we now know quite a bit about the first steps in the processing of visual information. These steps take place first in the retina (inside the eye), then in the thalamus (deep inside the brain), and finally in the primary visual cortex (in the back of the brain). These regions constitute the early visual system. Understanding the early visual system is an essential prerequisite for the design of prostheses that bypass the eye and directly stimulate the brain. Such prostheses could be of great help to the many blind people who suffer from damage to the eye or to the nerve connecting the eye to the rest of the brain.

Neurons in the early visual system respond to light stimulation in regions of visual space known as receptive fields. While the basic mechanisms underlying these responses are fairly well understood, there are other determinants of neuronal responses that currently elude us. For example, neuronal responses are strongly influenced by stimulation outside the receptive field. Moreover, responses are strongly influenced by the previous history of stimulation. I have been investigating this kind of phenomena by recording the visual responses of neurons in thalamus and primary visual cortex.

Ultimately, our research will be truly successful, and useful, if it can be used to predict the responses of neurons to complex, rapidly varying visual stimuli such as those encountered during daily life. Most research on neuronal responses, however, is performed on simplified laboratory visual stimuli such as light/dark bars on a gray background. To overcome this limitation, we have recently begun recording neuronal responses to complex visual stimuli. These stimuli are either artificial movies taken from cartoons or natural movies filmed in a forest. We repeat the presentation of the same movie a number of times, and we record the timing of the neural impulses, or spikes, fired by the neuron. They tend to occur at certain times much more than at other times. We test our understanding of these responses by devising models that incorporate our knowledge of neuronal mechanisms. This model captures the main features of the response, but fails to predict the exact size of each burst of activity. We are currently working on improving these predictions.

For more information, visit Matteo Carandini's lab web pages.

Collaborators: Severine Durand, Valerio Mante, Vincent Bonin.