Anxiety-like behavior in rats: a computational model.

This work describes a neural network model of the rat exploratory behavior in the elevated plus-maze, a test used to study anxiety. It involves three parameters: drive to explore; drive to avoid aversive stimuli; and spontaneous locomotor activity. Each network unit corresponds to a specific location in the maze and the connections, only between closest neighbors, represent the possible adjacent places to which a virtual rat can navigate. Competitive learning is used to generate a sequence of network states that correspond to the virtual rat successive locations in the maze. To evaluate the generality of the model it was also tested for two modifications of the elevated plus-maze: one with totally closed arms and the other with totally open arms. The results are compared with data obtained with rats. The simulations are consistent with experimental evidence and may provide an efficient way of describing the anxiety-like rat behavior in the elevated plus-maze. This could be useful for researching the emotional parameters involved in this anxiety animal model.