We evaluated tolerance of nasal cannula (NC) by 14 infants (median age, 2.6 months) and 16 children (median age, 5.5 years) with suspected obstructive sleep apnea syndrome and compared the efficacy of the NC with that of a nasobuccal thermistor in detecting obstructive apneas (OA) and obstructive hypopneas (OH) on polysomnography traces. The relationship between cannula flow and esophageal pressure was assessed in six patients. Time spent with an uninterpretable flow signal was longer when using a cannula than when using a thermistor in infants (p < 0.05) and children (p < 0.01), and it was longer in the younger patients (p < 0.05). Among the 650 OA-OH detected by either method, only 38% were detected by both, and 58% were detected by the cannula and missed by the thermistor, so that the apnea-hypopnea index was higher with cannula than with thermistor in each age group (p < 0.01). More hypopneas than apneas were detected by the cannula and missed by the thermistor (p < 0.001). Out-of-phase thoracic and abdominal motions and/or changes in the end-tidal CO(2) signal shape were associated with 86% of OH identified by cannula. In the six patients whose esophageal pressure was measured, all respiratory events identified using a cannula were associated with increased "airway resistance." Thus, the NC is more likely than the thermistor to detect OA and OH in infants and children, and this superiority is particularly marked for hypopneas.