The investigators carried out threshold determinations on 16 children and 6 adults in wakefulness, under general anaesthesia (we used chloral hydrate anaesthesia) and in sleep (stage II-III and stage REM). Falling asleep (stage I and initial stage of anaesthesia respectively) the latencies of the individual components of the acoustically evoked potentials are prolonged in mean of 30 msec. Simultaneously the amplitude of N1 significantly decreases and N2 becomes a prominent point (Fig. 1). The generation mechanisms of wave N2 are obviously different from those of wave N1. Its input-output curve takes a very steep course (Fig. 5) and the shortening of latencies increases with growing intensity of stimulus too (Fig. 4). Amplitude histogrammes demonstrated the dependency of the form of the acoustically evoked potential on the degree of synchronisation of EEG activity. While in the case of desynchronisation N1 appears more markedly, N2 does in the case of synchronisation. The mean deviation of the ERA threshold totals plus 3.8 +/- 6.9 dB (n = 41) under chloral hydrate anaesthesia, plus 4.9 +/- 6.7 dB (n = 37) in natural sleep in contrast to the wakefulness. With a 99% confidence there occur confidence intervals ranging from + 1 to + 7 dB and from +2 to +8 dB respectively. In identifying the threshold potentials error I (existing potential not recognized) occurred in 15-20%, error II (random wave seen as potential) in 20% of these studies. All these experiments showed significant lower variances for the latencies compared with variancies of amplitudes. The variance of amplitudes is smallest in children (Table 1) under general anaesthesia as well as in adults in wakefulness (Table 2). For the practical performance of ERA chloral hydrate is recommended for studies on children. A uniform EEG-state as well as a uniform depth of sleep are basic conditions for ERA during sleep, sedation or under anaesthesia. These conditions must constantly be controlled by EEG, EOG and EMG.