How neurons in humans encode information about the outside world and how this processing changes when the brain is diseased are central questions in neuroscience and medicine. Historically, microelectrode recordings of single-unit neuronal activity have been confined to animal preparations. Recently, it has become possible to obtain single-unit recordings in humans undergoing deep brain stimulation surgery. In this study, we recorded neuronal activity from the sub-thalamic nucleus (STN) of the basal ganglia of patients with Parkinson's disease (PD). In parallel, identical experiments were conducted on a healthy primate, providing a rare opportunity to analyze STN neuronal activity recorded in both the disease and healthy state during the same behavioral tasks. We developed point process models of STN neurons to capture neural spiking dynamics as a function of extrinsic stimuli and the neuron's own spiking history. Our models quantify, for the first time, pathological signatures in PD neural activity such as bursting, 10-30Hz oscillations, and loss of directional plurality, which may directly relate to motor disorders observed in PD patients such as bradykinesia, resting tremor, and rigidity.