OBJECTIVE The subthalamic nucleus (STN) has recently become the surgical target of choice for the treatment of medically refractory idiopathic Parkinson's disease. A number of anatomic and physiological targeting methods have been used to localize the STN. We retrospectively reviewed the various anatomic targeting methods and compared them with the final physiological target in 15 patients who underwent simultaneous bilateral STN implantation of deep brain stimulators. METHODS The x, y, and z coordinates of our localizing techniques were analyzed for 30 STN targets. Our final targets, as determined by single-cell microelectrode recording, were compared with the following: 1) targets selected on coronal magnetic resonance inversion recovery and T2-weighted imaging sequences, 2) the center of the STN on a digitized scaled Schaltenbrand-Wahren stereotactic atlas, 3) targeting based on a point 13 mm lateral, 4 mm posterior, and 5 mm inferior to the midcommissural point, and 4) a composite target based on the above methods. RESULTS All anatomic methods yielded targets that were statistically significantly different (P < 0.001) from the final physiological targets. The average distance error between the final physiological targets and the magnetic resonance imaging-derived targets was 2.6 +/- 1.3 mm (mean +/- standard deviation), 1.7 +/- 1.1 mm for the atlas-based method, 1.5 +/- 0.8 mm for the indirect midcommissural method, and 1.3 +/- 1.1 mm for the composite method. Once the final microelectrode-refined target was determined on the first side, the final target for the contralateral side was 1.3 +/- 1.2 mm away from its mirror image. CONCLUSIONS Although all anatomic targeting methods provide accurate STN localization, a combination of the three methods offers the best correlation with the final physiological target. In our experience, direct magnetic resonance targeting was the least accurate method.