DELAY ACTIVITY IN THE MEDIAL PREFRONTAL CORTEX OF THE RAT DISCRIMINATES BETWEEN STIMULUS SEQUENCES THAT PREDICT FUTURE REWARD

S.L. Cowen *; M.L. Ruiz Luna; J.P. Martin; S.N. Burke; B.L. McNaughton

NSMA , Univ of Arizona, Tucson, AZ, USA

The medial prefrontal cortex (anterior cingulate, prelimbic and infralimbic

cortices) has been implicated in the acquisition of reinforcement learning tasks. Furthermore, this region receives inputs from multi-modal integration areas throughout the cortex, the hippocampus, and from the VTA, a region that is strongly implicated in error-driven learning. Given these observations, we hypothesized that neurons in this region may selectively encode stimulus sequences that predict the delivery of future reward. To test this hypothesis, simultaneous activity from multiple neurons in the medial prefrontal cortex of the rat was recorded as the animal performed a paired associate discrimination task. A delay of 700 milliseconds separated the presentation of the first and second stimulus of the pair. In all of the test conditions, the first stimulus indicated a 50% probability of reinforcement. In one test condition, the presence of the second stimulus further improved the animal's current prediction of reinforcement. In the second condition, the second stimulus provided no additional information. Although neural responses to the presentation of the first stimulus of the pair were weak or non-existent in all conditions, responses during the delay period that preceded the presentation of the second stimulus, and responses during the presentation of the second stimulus were significantly discriminative between individual stimulus sequences when the second stimulus of the sequence predicted reward. These results suggest that the medial prefrontal cortex is involved in the representation of sequences that predict rewarding outcomes and that these representations are activated prior to the presentation of the expected stimulus. This activity may serve to facilitate the rapid comparison between expected and observed stimuli so that appropriate actions can be quickly initiated. From a reinforcement learning perspective, these neurons, in essence, provide a “prediction” from which prediction error can be computed.

Support Contributed By: MH046823

Key words: medial prefrontal cortex, reinforcement learning, attention, associative learning