The aim of this study was to test the reproducibility of three-dimensional (3D) surface models of maxillary incisors and to propose a characterization of root morphology. The sample was comprised of pre-treatment cone-beam computed tomography (CBCT) images of fifty-five patients. The CBCTs were used to construct 3D surface models of the maxillary incisors. The reproducibility of surface models was tested by repeated construction of them by two observers. A 3D surface model that corresponded to the average of all lateral and all central incisors was generated. 3D surface distances and vector differences were calculated for each individual tooth and the average of the teeth considered. The corresponding points on the 3D surface mesh for each subgroup were compared statistically to those of the neutral subgroup using shape analysis MANCOVA and Hotelling's t-statistic (p < 0.05). Repeated construction of surface models demonstrated adequate inter-rater reproducibility. The distribution of 3D models into root morphology subgroups was: blunt (11% and 26% of the central and lateral incisors, respectively), conical (15% of the central incisors), long (27% and 20% of the central and lateral incisors, respectively), and short (15% and 4% of the central and lateral incisors, respectively). Compared to the neutral average, statistically significant differences in root morphology were found for blunt, long, conical, and short central incisors and for blunt, long, and short lateral incisors. We can conclude that 3D surface models construction for upper incisors is reproducible. 3D shape analysis using CBCT images allows a phenotypic characterization of incisor root morphology.