A NEURAL ARCHITECTURE FOR RULE-GUIDED BEHAVIOR: A SIMULATION OF PHYSIOLOGICAL EXPERIMENTS

Abstract

The ability to use abstract rules is very important in applying our experience to new situations. To examine the system governing such rule-guided behavior, we developed a recurrent neural network model of rule-guided behavior and simulated a physiological experiment involving a rule-guided delayed task (Wallis, Anderson, & Miller, 2001). Our model was constructed using neural system identification (Zipser, 1992), and a fully recurrent neural network model was optimized to perform a rule-guided delayed task. The model’s hidden layer contained rule-selective units, and an examination of connection weights substantiated the postulate that rule-selective neurons maintain encoded rule information and contribute to rule-guided responses indirectly. The simulation results predicted functional interactions among neurons involved in various task-related activities. The similarity between the behavior of the model units and biological neurons shows that the brain uses mechanisms like those of the model, and that ample mutual connections in the prefrontal cortex are the basis for promoting flexible learning.