Simple clinical tests, like Romberg's test or a walking test, have proved to be inadequate guidelines for safe discharge after outpatient anaesthesia. A randomised study was therefore planned to compare postural stability measured by computerised posturography in 31 oral midazolam-atropine premedicated children aged 6.9 (s.e. 0.4) years who had been anaesthetised with either propofol/alfentanil/N2O or thiopentone/halothane/N2O. The sway velocity of the children was measured before premedication and 1, 2 and 3 h after the end of anaesthesia. Results show that sway velocity had returned to baseline values 3 h after the end of anaesthesia in all children who had received propofol/alfentanil/N2O and in 12 of the 15 children who had received thiopentone/halothane/N2O. The quantified version of the Romberg test performed with eyes open or closed was not impaired after anaesthesia, compared with the control values, indicating that in children poor equilibrium is not compensated by vision. The clinical recovery with respect to the times to eye opening, to responding to command or to being fully awake did not differ between the two anaesthesia methods. On the basis of recovery assessed by postural stability, propofol/alfentanil/N2O anaesthesia was not preferable to thiopentone/halothane/N2O anaesthesia after minor paediatric otolaryngological surgery.