|
Places cells, view cells, grid cells and transition cells...| old_uid | 1602 |
|---|
| title | Places cells, view cells, grid cells and transition cells... |
|---|
| start_date | 2006/10/10 |
|---|
| schedule | 14h |
|---|
| online | no |
|---|
| summary | In this talk, we discuss some of the issues regarding hippocampal models according to recent neurobiological results. We propose that two kinds of proprioceptive information are integrated at two different locations in the hippocampal loop corresponding to two distinct functions. First, grid cell activity recorded in the dorsocaudal medial entorhinal cortex
(EC) [Hafting et al., 2005] would be the simple result of a projection of incoming proprioceptive informations. If the structure of the connexion for a given cluster of neuron is equivalent to a modulo N operator, it is quite easy to build of a robust and high resolution code for long term path integration with very few cluster of neurons corresponding to modulor factors premiers entre eux. Broad place cell activity found in the latero dorsal EC would be the result of the integration of visual informations or view cells activity according to the field of view of the visual system. The merging of both kind of
information in the EC and in the dentate gyrus (DG) would allow the building of precise and robust "place cells" (activity preserved in the dark for instance). The phase advance of the supposed "place cells" recorded in the hippocampus proper (CA3 and CA1 regions) can be explained by a misinterpretation of the activity of those cells. In a general case, they would not be "place cells" but "transitions cells" coding for the transition between one state to the next (see [refs] for more details on the learning of such transition cells). The model of view cells, we proposed years ago to explain activity recorded in monkey can also be explained from the same model by a simple reduction of the field of view associated to a reduction of the system capability to merge multimodal information. Second, the proprioceptive information coming from a path integration on a short time scale (reset from one place to the next) would be merged in the subiculum with our CA3/CA1
"transition cells" to provide a robust feedback of the current action to the deep layer of the entorhinal cortex in order to predict the recognition of the current state. Along the talk, the crucial role of robotics experiments to determine the constraints allowing in a single
and coherent model both simple sensori-motor navigation and more complex
planning behaviors involving the building and the use of cognitive maps
will be discussed. |
|---|
| responsibles | Burle |
|---|
| |
|