Radiation treatment planning is performed on images that do not take variation in patient position into account. To compensate for expected variations in position of the patient, a three-dimensional expansion of the clinical target volume, or set-up margin, is added. Variations in patient position can be decreased through use of an immobilization device, allowing selection of a smaller set-up margin. The objective of this prospective study was comparison of interfractional variation in patient position between set-ups of the canine head region using palpation of bony landmarks and set-ups using a head-repositioning device. Fiducial markers were attached to the skull bones of three research dogs, and the dogs were positioned as for a typical radiation treatment of the head region using both set-up methods. A kilovoltage on-board imager was used to acquire orthogonal images and the difference between the x-, y-, and z-axis coordinates of each fiducial marker relative to the initial reference isocenter was measured. The difference in patient position for each axis coordinate was significantly lower for set-ups using the head-repositioning device than for set-ups using bony landmarks (P < 0.05). Ninety-five percent of the absolute values of the displacement vector differences were < 4.62 mm for set-up using bony landmarks, and < 1.93 mm for set-up using the head-repositioning device. A minimum set-up margin of 5-6 mm is recommended when patient set-up is based on bony landmarks and of 2-3 mm when the head-repositioning device is used.