This study investigated performance in a desktop virtual environment as a function of stereopsis and head tracking. Ten subjects traced a computer-generated wire using a virtual stylus that was slaved to the position of a real-world stylus tracked with a 6-DOF position sensor. The objective of the task was to keep the virtual stylus centered on the wire. Measures collected as the subjects performed the task were performance time, and number of times the stylus overstepped the virtual wire. The time to complete the wire-tracing task was significantly reduced by the addition of stereopsis, but was not affected by the presence of head tracking. The number of times the virtual stylus overstepped the wire was significantly reduced when head-tracking cues were available, but was not affected by the presence of stereoscopic cues. Implications of the results for performance using desktop virtual environments are discussed.