A prototype armoured laryngeal mask airway (LMA) was compared with tracheal intubation (ETT) for anaesthesia for adenotonsillectomy. Fifty-five children were randomised into the LMA group and 54 into the ETT group. During insertion of the LMA, peripheral oxyhaemoglobin desaturation (SpO2) < 94% occurred in ten patients (18.2%) and in seven patients (13%) during tracheal intubation (NS). After opening the Boyle-Davis gag, airway obstruction occurred in ten patients (18.2%) in the LMA group and in three patients (6%) in the ETT group (P = 0.07). In five patients (9%) the LMA was abandoned in favour of tracheal intubation. In all others (91%), when the need for adequate depth of anaesthesia was realized, a satisfactory airway was achieved more rapidly than with tracheal intubation (P < 0.001), and maintained throughout surgery. Manually assisted ventilation was required in all patients in the ETT group, mean duration 373 +/- 385 sec, and in 26 patients (52%) in the LMA group, mean duration 134 +/- 110 sec, P < 0.001. Mean end-tidal CO2 (PetCO2) was 45.5 +/- 6.21 mmHg in the ETT group and 46.6 +/- 6.09 in the LMA group (NS). The LMA did not limit surgical access. Heart rate, MAP and blood loss in the LMA group were 110 +/- 21, 74 +/- 9 mmHg and 1.92 +/- 1.22 ml.kg-1 respectively, compared with 143 +/- 13 (P < 0.001), 85 +/- 12 mmHg (P < 0.001) and 2.62 +/- 1.36 ml.kg-1 (P < 0.05) with tracheal intubation. Fibreoptic laryngoscopy at the end of surgery in 19 patients in the LMA group revealed no blood in the larynx.(ABSTRACT TRUNCATED AT 250 WORDS)