This thesis presents a model of the basal ganglia using a continuous time recurrent neural network (CTRNN) to simulate its pathways and learning. The model reproduces the response of globus pallidus neurons seen in previous work. An error function is used to calculate the difference between striatum and globus pallidus activity for weight updates between the subthalamic nucleus and globus pallidus. Experiments adding noise and using a sigmoidal update rule are also presented. The thesis discusses representing the model with equations versus code, noting tradeoffs between readability and reproducibility. A CTRNN library was implemented in Python to simulate the basal ganglia model.