To clarify the mechanisms of rigid and semi-rigid mandibular repositioning devices (MRDs) in obstructive sleep apnea syndrome (OSAS), seven and 13 patients received rigid and semi-rigid MRDs, respectively. Each patient underwent polysomnographic and computed tomographic examinations at the initial consultation and after symptom improvement. Three-dimensional models of the upper airway (hard palate level to epiglottic base) were reconstructed by image processing software (Mimics version 14.2) to measure airway morphology. The mean age and body mass index were 58.1 years and 24.8 kg/m2, respectively, in the rigid MRD group and 57.9 years and 23.2 kg/m2, respectively, in the semi-rigid MRD group. The apnea-hypopnea index significantly improved (P < 0.05, Wilcoxon signed-rank test) from 22.0 to 8.9 and 20.5 to 11.5 events per hour of sleep in the respective groups. The cross-sectional areas measured at the epiglottic tip (from 2.0 to 2.6 cm2) and hard palate (from 2.6 to 3.3 cm2) levels also increased in the respective groups (P < 0.05). However, airway volume, cross-sectional area measured at the uvular tip level, and anteroposterior and transverse diameters of the airway were not significantly different. In conclusion, both types of MRDs improve respiratory status, but they affect different parts of the airway.