Deep brain simulation (DBS) is a powerful new therapeutic approach for patients with Parkinson's disease. However, patient selection is critical for a valuable therapeutic result. Dopa sensitivity of the target symptoms, severe disability and low neurosurgical risks are among the major criteria for this indication. Other criteria like age or cognition must still be addressed in future prospective studies. The preferred target for DBS in PD is the subthalamic nucleus for various good reasons. However, prospective studies for this procedure are lacking and some clinical problems may be more easily solved with targeting the internal pallidum or the thalamus. Despite major progress in this field, much work remains to be done.