The capability of a commercial respiratory gating system based on video tracking of reflective markers to reduce motion-induced CT planning and treatment errors was evaluated. Spherical plastic shells (2.8-82 cm3), simulating the gross target volume (GTV), were placed in a water-filled body phantom that was moved sinusoidally along the longitudinal axis of the CT scanner and the accelerator for +/- 1 cm at 15-30 cycle/min. During gated CT imaging, the x-ray exposure was initiated by the gating system shortly before the end of expiration (so that the imaging time would be centered at the end of expiration); it was terminated by the scanner after completion of each slice. In nongated CT images, the target appeared distorted and often broken up. GTVs volume errors ranged 16%-110% in axial scans, and 7%-36% in spiral scans. In gated CT images, the spheres appeared 3 and 5 mm longer than their actual diameters (volume errors 2%-16%), at the respective respiration rates of 15 and 20 cycles/min. At 30 cycles/min the target appeared 1 cm longer, and volume error ranged 25%-53%. During treatment, gating kept the beam on for a duration equal to the CT acquisition time of 1 s/slice. The difference in positional errors between gated CT and portal films was 1 mm, regardless the size of residual motion errors. Because of the potential of suboptimal placement of the gating window between CT imaging and treatment, an extra 1.5-2.5 mm safety margin can be added regardless of the size of residual motion error. For respiratory rates > or = 30 cycles/min, the effectiveness of gating is limited by large residual motion in the 1 s CT acquisition time.