Nos tutelles

Nos partenaires


Accueil du site > Séminaires LNC > séminaires à venir (upcoming seminars)

Elizabeth Thomas

The Brain Behavior Gap : The Importance of Computational ModelingCognition,

Action et plasticité sensorimotrice (INSERM-Université de Bourgogne)

One of the biggest challenges in neuroscience, if not the biggest, is bridging the gap between the neuronal and behavioral levels. How is it possible that the wet, messy brain is able to judge intervals of time, or distinguish trees from non-trees ? The same challenge of bridging these levels of organization confronts us when trying to understand neuronal malfunctions. Many of the difficulties in trying to bridge the brain-behavior gap come from the fact that both the organization and functions of the brain are distributed. This gives rise to solutions at the higher network or behavioral level that are variable and/or redundant.

I will describe how computational neuroscience can be used to confront these problems by means of i) computational models and ii) computational tools. Computational models, in general, attempt to mimic salient neural characteristics and can be used to test hypotheses, probe mechanisms and, most importantly, to make predictions. I will describe a model we developed to understand how timing information can be represented in a distributed neuronal network. The model demonstrates, counterintuitively, that the dissociation observed in prospective and retrospective timing under cognitive load does not necessarily imply a physical separation of these two types of timing in the brain. In contrast, computational tools do not attempt to mimic neural processes but, rather, deal with the problem of recordings taken from a distributed neural system. These recordings include multiunit cellular recordings, evoked potentials, EMGs, or fMRi. I will illustrate, using self-organizing maps, how a neuronal population of approximately 50 neurons in the inferior temporal cortex encodes tree and non-tree categories. I will also very briefly mention how various computational techniques can be used to extract relevant variables from noisy data, automatically determine EMG onset times, etc.