In general, I am interested in understanding the neural basis of the representation of external space, and how we move into it. Two specific problems arise regarding the perception of extra-personal space, when compared for instance with vision. First, there is no specific "space sensor": the information through which space is perceived are by essence multisensory, involving visual, somatosensory as well as vestibular inputs. Second, these informations vary constantly while we move in our environment, and despite this fact, the external world appears to be stable.

My research more particularly focuses on understanding how multisensory informations are combined to lead to the representation of the space around us, and how this representation is being updated during movement so that the world appears to be stable. In a second line of thoughts, I am also interested in understanding how visual motion of objects is processed in the brain -with single neurons as well as neuronal populations-, and how we can interact in space with moving objects.

 

The parietal cortex has been know since the 50's to be implicated in the representation of space, when neurologists like Critchley described specific spatial deficits caused by parietal lesions, namely the hemineglect syndrome. In the mid 80's, it was shown that the neglect of one hemispace by parietal patients could be completely canceled by caloric stimulation of the inner ear (thus stimulating the head movement related sensors: the vestibular system). Despite this spectacular effect, very little neurophysiological research has been dealing with that question. Thus a large part of my neurophysiological work has focused on identifying and analyzing vestibular responses in the parietal cortex. This work is realized with Werner Graf.

 

There are two major functionnal differences between pathways for the "classical" senses like vision and our "un-noticed" sense that is our sense of balance. (1) Unlike as for vision, there is no primary vestibular cortex, but a network of more than 5 cortical areas receive input from the vestibular thalamus (ventral postero-lateral and inferior nuclei). (2) As soon as the first central relay (vestibular nuclei), vestibular signals are "multiplexed" with informations from other modalities, namely visual and eye movement signals. Therefore, the "vestibular" signals reaching the cortex are in essence multisensory.

Little is known about cortical vestibular networks: neuron networks' architecture can be mapped in a large scale with the help of a self replicating tracer which is rabbies virus. Focal injections of minute quantities of this tracer can progressively unravel the active connections arriving to this location, and help getting a better picture of the head-movement system wiring.

This work is done in collaboration with Gabriella Ugolini and Werner Graf.

 

One prerequisite for a stable representation of space is that, because of delays of sensory processing, the update of the perceived space should begin before a given movement. I studied the predictive behavior, as well as the physiology, of cats while they tracked with their gaze a moving visual target, with Alexej Grantyn, Julien Petit and Alain Berthoz. In relation to this study, I also investigated the prediction of a target trajectory in relation to active and passive movement in humans, with Mark Wexler. Further studies involving a close comparison of animals and humans predictive behaviors are planed.

 

Neural signals consists of series of impulses called spikes, whose generation in response to, for example, a sensory stimulation, appears to be not entirely deterministic, at least at the cortical level. Therefore, one implicit assumption of neurophysiology is that a precise information about the events occurring in the external world is to be obtained by pooling the responses of populations of neurons, such that the intrinsic "randomness or noise" of individual cells is "cleaned up". But this is true only if the neurons are statistically independent, which is certainly not the case in the brain. Thus, a good understanding of how the brain actually works, how neuronal signals are processed by populations of neurons, and how subsequent level have access to the information transmitted by the previous level cannot be based solely on studies of single neurons. The goal of my current research is to understand how information can be transmitted optimally in a statistical sense, given the correlation structure among a population of V1-MT modeled neurons. This work is a collaboration with Alex Pouget.

 

The movement of an object in space is computed from the variations in time of its projection on the retinal surface. For large objects, local motion information (i.e. apparent direction of movement of an object point) is quite often different from the true motion of the moving element. Moreover, neurons dedicated to compute visual motion only do so in a given small region of the visual space (their receptive field). But since they are part of larger networks, their estimation of what is happening in their localized visual world is influenced by events occuring in neighboring retinal regions (non-classical surround).

This work deals with how the brain solves this infamous "aperture problem", in other words how it computes the true motion of an object from localized ambiguous estimates. This work is realized with Gene Stoner.