Accuracy in manual manipulation in a simulated craniotomy environment has been studied. Eighty-eight neurosurgeons with varying amounts of surgical experience were tested in two tasks: attempting to hold an instrument still, and trying to trace a line. Results for RMS and maximum linear and angular motion were measured in the static task. RMS and maximum error were measured in the dynamic task. Spectral analysis was also performed for both tasks. The median RMS amplitude of the position vector in three dimensions was 526 microm in the stationary task, and the median RMS amplitude of the error vector was 289 microm in the tracing task.