Complementary roles of the rat prefrontal cortex and striatum in reward-based learning and shifting navigation strategies

Abstract : Many mammals can behave according to different navigation behaviors, defined as " strategies " which, although not systematically requiring conscious processes, depend on the specific task they are required to solve. In certain cases, if a visual cue marks the goal location, the agent can rely on a simple stimulus-response (S-R) strategy. In contrast, other tasks require the animal to be endowed with a representation of space that allows it to locate itself and to locate goals in the environment. In order to efficiently navigate, the animal not only should be able to learn and exhibit these types of strategies, but it should also be able to select which strategy is the most appropriate to a given task conditions in order to shift from one strategy to the other to optimize outcomes. The present work employs a multidisciplinary approach (e.g. behavior, neurophysiology, computational neuroscience and autonomous robotics) to study the roles of the rat prefrontal cortex and striatum in learning and shifting navigation strategies, and their possible application to robotics. It aims more particularly at investigating the respective roles of the medial prefrontal cortex (mPFC) and of different parts of the striatum (DLS :dorsolateral ; VS: ventral) in these processes, and the nature of their interactions. The experimental work presented here consisted in : (1) studying the role of the striatum in S-R learning by : (a) analyzing electrophysiological data recorded in the VS of rats performing a reward-seeking task in a plus-maze; (b) designing an Actor-Critic model of S-R learning where VS is the Critic which drives learning, whereas DLS is the Actor which memorizes S-R associations. This model is applied to robotics simulations, and compared with existing models in a virtual plus-maze; (2) studying the role of mPFC in strategy shifting by means of electrophysiological recordings in the mPFC of rat performing a task requiring such kind of shifts. The principal results of this work suggest that : (1) In the S-R framework: (a) as in primates, the rat VS shows a reward anticipation activity coherent with the Actor-Critic theory; (b) these reward anticipations can be combined with self-organizing maps in an Actor-Critic model that gives a better performance than previous models in a virtual plus-maze, and that shows generalization abilities potentially applicable for the field of autonomous robotics; (2) the rat mPFC seems to play a role when the animal's current strategy has poor reward yields, prompting learning of another strategy. Moreover, population activity in mPFC changes rapidly in correspondence with shifts in the animal's task-solving strategy, possibly underlying the contribution of this brain area to flexible selection of behavioral strategies. In conclusion the results are discussed in the framework of previous behavioral, physiological and modeling studies. We propose a new architecture of the rat prefronto-striatal system, where sub-territories of the striatum learn concurrent navigation strategies, and where the medial prefrontal cortex helps decide at any given moment which strategy dominates for behavior.
Complete list of metadatas
Contributor : Mehdi Khamassi <>
Submitted on : Wednesday, April 18, 2012 - 8:49:55 PM
Last modification on : Friday, March 22, 2019 - 1:31:50 AM


  • HAL Id : tel-00688927, version 1


Mehdi Khamassi. Complementary roles of the rat prefrontal cortex and striatum in reward-based learning and shifting navigation strategies. Cognitive Sciences. Université Pierre et Marie Curie - Paris VI, 2007. English. ⟨tel-00688927⟩



Record views


Files downloads