The laryngeal mask airway (LMA), an alternative to tracheal intubation in certain situations, has gained popularity in recent years. Initially designed for use in adults it has now become available in suitable sizes for paediatric anaesthesia. The objectives of this study were to identify the preferred site of sampling the end-tidal carbon dioxide (PETCO2) with the LMA and to determine the accuracy of this recording when compared with arterial CO2 (PaCO2). We studied 30 healthy children, age one to five years and weighing between 10 and 25 kg undergoing minor surgery requiring mask anaesthesia. In each case, after induction of anaesthesia, the LMA was inserted under direct vision to eliminate the possibility of epiglottic airway obstruction. The fresh gas flow was provided by a Jackson Rees modification of an Ayre's T-piece and was determined according to the following formula: 3 x (1000 + (100 x body weight)) LPM. Blood pressure, ECG, O2 saturation, temperature and end-tidal gas concentrations were recorded. The measures of peak PETCO2 were taken at pre-determined distances from the elbow connector down the LMA shaft. During the sampling sequence an arterial blood sample was taken for gas analysis. The PaCO2 was 63.5 +/- 9.3 mmHg (mean +/- SD). At any given sampling site, mean PETCO2 values were less than PaCO2 (P < 0.05). However, in eight patients PETCO2 values measured at the distal site were higher than the PaCO2 (negative P(a-ET)CO2 gradients).(ABSTRACT TRUNCATED AT 250 WORDS)