Deep brain stimulation (DBS) is an established therapy for cardinal motor signs and medication-related complications in Parkinson’s disease (PD). Current DBS therapy is limited to “open-loop” neurostimulation: the neurostimulator cannot sense the brain signals nor the behavior it is modulating. It applies continuous pulse trains of fixed frequency, amplitude, pulse width, and pattern, and cannot adjust such parameters in response to neural activity, or the patient’s state of activity or behavior. It applies a one size fits all set of stimulation parameters no matter what the dominant symptom of the patient. Adaptive or closed-loop (cl)DBS uses real time physiological signals to inform the neurostimulator when and how to change stimulation, to provide optimal therapy and minimize adverse effects. Successful subthalamic (STN) clDBS in PD will require discovery of patient specific neural and behavioral features (control variables) that reflect the disease, state of activity and/or dominant symptom, along with control policy algorithms that will change stimulation parameters in a manner that modulates the variable and improves motor function in a patient specific manner. In this talk I will discuss our research into relevant neural and physiological signals that may be useful for clDBS for specific PD motor symptoms, and how we have used a dual threshold control policy algorithm based on patient specific therapeutic windows in the first feasibility studies on clDBS for PD using a fully implanted neurostimulator.