PURPOSE: To study the potential impact of the combined use of CT and MRI scans on the Gross Tumor Volume (GTV) estimation and interobserver variation. METHODS AND MATERIALS: Four observers outlined the GTV in six patients with advanced head and neck cancer on CT, axial MRI, and coronal or sagittal MRI. The MRI scans were subsequently matched to the CT scan. The interobserver and interscan set variation were assessed in three dimensions. RESULTS: The mean CT derived volume was a factor of 1.3 larger than the mean axial MRI volume. The range in volumes was larger for the CT than for the axial MRI volumes in five of the six cases. The ratio of the scan set common (i.e., the volume common to all GTVs) and the scan set encompassing volume (i.e., the smallest volume encompassing all GTVs) was closer to one in MRI (0.3-0.6) than in CT (0.1-0.5). The rest volumes (i.e., the volume defined by one observer as GTV in one data set but not in the other data set) were never zero for CT vs. MRI nor for MRI vs. CT. In two cases the craniocaudal border was poorly recognized on the axial MRI but could be delineated with a good agreement between the observers in the coronal/sagittal MRI. CONCLUSIONS: MRI-derived GTVs are smaller and have less interobserver variation than CT-derived GTVs. CT and MRI are complementary in delineating the GTV. A coronal or sagittal MRI adds to a better GTV definition in the craniocaudal direction.