Detailed analysis of the upper airway has not been performed in children with obstructive sleep apnea. We used magnetic resonance imaging and automatic segmentation to delineate the upper airway in 20 children with obstructive sleep apnea and in 20 control subjects (age, 3.7 +/- 1.4 versus 3.9 +/- 1.7 years, respectively). We measured mean and minimal cross-sectional area, length, and volume of: (1) the total airway; (2) regions along the adenoid, tonsils, and where adenoid and tonsils overlap; and (3) 10 segments at 10% increments along the airway. The mean cross-sectional area of the total airway of the obstructive sleep apnea group was significantly smaller in comparison with the control group, 28.1 +/- 12.6 versus 47.1 +/- 18.2 mm2, respectively (p < 0.0005). Minimal cross-sectional area and airway volume were smaller in this group, 4.6 +/- 3.3 versus 15.7 +/- 12.7 mm2 (p < 0.0005), and 1,129 +/- 515 versus 1,794 +/- 846 mm3 (p < 0.005), respectively. Regional analysis suggested that the upper airway in children with obstructive sleep apnea is most restricted where adenoid and tonsils overlap. Segmental analysis demonstrated that the upper airway is restricted throughout the initial two-thirds of its length and that the narrowing is not in a discrete region adjacent to either the adenoid or tonsils, but rather in a continuous fashion along both.